i 

 

 

 

 

 

Title: The generation of Plasmodium 
falciparum Hsp90 selective inhibitors 

using rational drug design 
 

 

T Matlhodi  
orcid.org0000-0001-9763-0567 

 

 

Dissertation accepted in fulfilment of the requirements for the 
degree master of science in biochemistry at the  

North West University 
 

 

Supervisor: Dr F Mokoena 
Co-supervisor: Dr N Gumede    

  

Graduation ceremony: May 2024 
 



 

ii 

 
DECLARATION 

I, Thato Matlhodi, hereby declare that this dissertation, titled: “The generation of Plasmodium 

falciparum Hsp90 selective inhibitors using rational drug design”, is the result of my research and 

academic work undertaken towards the completion of my Master of Science (MSc) degree at 

North-West University. 

 

I confirm that this dissertation represents my original work and that it has not been previously 

submitted for any other degree, diploma, or academic qualification. Any sources of information 

used in the dissertation have been duly acknowledged through proper citation and referencing, in 

accordance with the guidelines provided by my university. 

 

I further declare that all the primary data and findings presented in this dissertation are authentic, 

and any contributions made by others to this research have been duly recognized through 

acknowledgment or co-authorship, as applicable. 

 

Throughout the process of conducting this research, I have adhered to the highest ethical 

standards. I have obtained the necessary permissions, where required, to access and use any 

data, materials, or resources employed in this study. I have also maintained confidentiality and 

ensured the privacy of any individuals or organizations involved in this research. 

 

I  declare that my dissertation supervisors, Dr Fortunate Mokoena and Dr Njabulo Gumede, have 

provided guidance and support throughout the research process. Their expertise and valuable 

feedback have significantly contributed to the successful completion of this dissertation. 

 

I understand that any attempt to misrepresent the content or veracity of this dissertation may 

result in the imposition of penalties, including the withholding of my degree. 

 

By submitting this dissertation, I affirm that I have read, understood, and complied with all the 

relevant regulations and requirements set forth by my university and the department responsible 

for my MSc program. 

 

 

Thato Matlhodi 

27325903 

19/05/2023  



 

iii 

DEDICATION  

I dedicate this dissertation to my beloved baby girl, Letlotlo Bokgabo Mongale. You have been 

my constant source of inspiration and motivation throughout this academic journey. As I worked 

tirelessly to complete this MSc degree, I was driven by the desire to show you that with 

determination and perseverance, dreams can become a reality. You have filled my life with 

boundless love and joy, and I dedicate this achievement to you, my precious daughter. May this 

serve as a testament to the power of hard work and the infinite possibilities that lie ahead for you. 

 

I would also like to dedicate this dissertation to my parents, Khumo and Puoetsile Matlhodi. Your 

unwavering faith in my abilities and unwavering support have been invaluable throughout my 

academic pursuits. You have instilled in me a strong work ethic and taught me the importance of 

never giving up on my dreams. Your constant encouragement and belief in my potential have 

been the guiding lights that have led me to this point. This achievement is as much yours as it is 

mine, and I am forever grateful for your love, guidance, and sacrifices. 

 

To Letlotlo Bokgabo Mongale and my parents, Khumo and Puoetsile Matlhodi, I dedicate this 

dissertation. May it serve as a testament to the power of love, determination, and unwavering 

support. Your presence in my life has made all the difference, and I am honored to have you by 

my side on this remarkable journey. 



 

iv 

ACKNOWLEDGEMENTS 

I would like to express my sincere gratitude to the National Research Foundation (NRF) and the 
NWU Masters’ bursary for providing the financial support that made it possible for me to pursue 
this degree. Their funding has been instrumental in enabling me to focus on my studies and 
research, and I am deeply appreciative of their investment in my academic journey. 
 
I extend my heartfelt appreciation to my supervisor, Dr Fortunate Mokoena, for her invaluable 
guidance, support, and belief in my abilities. Her expertise, dedication, and motivation have been 
essential in shaping the direction and success of my MSc research. I am truly grateful for the 
opportunities provided and the wisdom shared throughout this process. 
 
I would also like to acknowledge and thank my co-supervisor, Dr. Njabulo Gumede, for his 
contributions and assistance as a co-supervisor. His input and expertise have enriched my 
research and broadened my perspective, and I am thankful for his valuable insights and guidance. 
 
To my parents, I am forever grateful for your unwavering support, love, and faith in me. Your 
encouragement and belief in my abilities have been the driving force behind my accomplishments. 
Your constant presence and reassurance have provided the foundation upon which I have built 
my academic journey. Thank you for always being there for me. 
 
I extend my appreciation to my lab mates, Tlhalefo Ntseane and Ofentse Mafethe. We have 
endured challenges together, celebrated successes, and shared countless hours of hard work 
and dedication in the Parasite Drug Discovery Lab. Your friendship, collaboration, and mutual 
support have made this journey more meaningful and enjoyable. I hope that we all continue to 
thrive and succeed in our future endeavors. 
 
To my partner, Reabetswe Paballo Mongale, thank you for being my rock and constant source of 
inspiration. Your unwavering support, love, and understanding have sustained me throughout this 
demanding period of my life. You have been my pillar of strength, lifting my spirits when they were 
low and believing in me when I doubted myself. I am grateful to have you by my side. 
 
I would like to acknowledge my best friends, Tlotlo Montwedi and Kgalalelo Letimela, for their 
support, encouragement, and friendship. Your presence and belief in my abilities have been 
invaluable. Thank you for standing by me and cheering me on every step of the way. 
 
To all those mentioned above and to the countless others who have supported me along this 
journey, thank you from the bottom of my heart. Your contributions, encouragement, and belief in 
me have made this accomplishment possible. I am forever grateful for your presence in my life.   



 

v 

OUTPUTS 

The findings of this study were showcased at the: 

 

South African Society of Biochemistry and Molecular Biology SASBMB conference in January 

2022, receiving recognition for their contribution to the field of bioinformatics and biochemistry.  

 

 Matlhodi Thato, Gumede Njabulo, Choene Mpho and Mokoena Fortunate. the use of 

machine learning for generation of Plasmodium falciparum Hsp90 selective inhibitors. 

27th congress of South African biochemistry and molecular biology. Virtual event (23-26 

January 2022) 

 

Furthermore, the study featured a poster presentation at an H3D symposium held in October 

2022. 

 Matlhodi Thato, Gumede Njabulo, Choene Mpho and Mokoena Fortunate. the use of 

machine learning for generation of Plasmodium falciparum Hsp90 selective inhibitors. 

27th congress of South African biochemistry and molecular biology.  Webersburg estate, 

Stellenbosch (25-28 October 2022).  

 

The work is being prepared for publication   

 Thato Matlhodi, Lisema Patrick Makatsela, Tendamudzimu Harmfree Dongola, Addmore 

Shonhai, Njabulo Joyfull Gumede and Fortunate Mokoena. Auto QSAR-based Active 

learning docking for hit identification of potential inhibitors of P. falciparum Hsp90 as 

antimalarial agents.  

  



 

vi 

ABSTRACT 

Malaria, which is mainly caused by Plasmodium falciparum parasite, remains a devastating 

disease that poses a significant burden on the public health systems and socio-economic 

development of countries globally. This is particularly true in regions with limited resources, such 

as sub-Saharan Africa. Drug-resistant strains of the P. falciparum parasite have significantly 

restricted effective treatment options, necessitating the development of new and effective 

antimalarial drugs that can overcome drug resistance. The molecular chaperone, heat shock 

protein 90 (Hsp90), plays a critical role in parasite survival and proliferation. Therefore, it is a 

promising target for the development of new antimalarial drugs. The generation of selective 

inhibitors is important for target specificity, therapeutic efficacy, safety, and drug tolerability. Using 

rational drug design strategies, this study aimed to generate selective inhibitors of P. falciparum 

Hsp90 (PfHsp90) through molecular modelling, facilitated by an active learning approach, 

followed by biochemical and biophysical assays.. This study used a combination of computational 

techniques, namely reaction-based enumeration, active learning, induced fit docking, molecular 

dynamics simulations, absorption, distribution, metabolism, excretion, and toxicity (ADME/T) 

predictions, and molecular mechanics with generalized born surface area (MM-GBSA) to 

generate a total of 10 000 drug-like design ideas. Furthermore, a large compound library from 

chEMBL was used to identify compounds with high affinity and selectivity for the target protein. A 

total of 13 models were built using active learning autoQSAR, which achieved good statistical 

results of R2 = 0.69, Q2 = 0.61, and RMSE = 0.44, on average for all models. The current study 

then assessed the ability of a subset of the purchased compound to inhibit the ATPase activity of 

PfHsp90 using an ATPase assay and protein-ligand interaction measured using surface plasmon 

resonance (SPR). Compounds FTN-T4 and FTN-T10 brought the most robust inhibition of 

PfHsp90’s ATPase activity, greater than that of the positive control harmine, which is in correlation 

with the in-silico data, while the majority of the compounds displayed a similar inhibition of the 

ATPase activity to harmine. In the biophysical study, compound FTN-T2 displayed moderate 

binding with a KD of 7µM, while FTN-T10 and FTN-T5 displayed binding affinities comparable to 

that of ADP, at KD values of 13 µM and 19 µM, respectively. The results presented in this study 

demonstrate the great potential of active learning models to generate novel P. falciparum Hsp90 

inhibitors. The identified compounds were found to disrupt the ATPase activity of PfHsp90, form 

sufficient interactions between the protein and ligand, possess a good safety profile and 

selectivity, and represent promising candidates for further lead optimization. By targeting 

PfHsp90, these inhibitors offer potential avenues for overcoming drug resistance and improving 

the efficacy of antimalarial therapies. 

Keywords: Malaria; Plasmodium falciparum; Heat shock protein 90; active learning  

 



 

vii 

TABLE OF CONTENTS  

DECLARATION ........................................................................................................................ II 

DEDICATION ........................................................................................................................... III 

ACKNOWLEDGEMENTS ........................................................................................................ IV 

OUTPUTS ................................................................................................................................. V 

ABSTRACT ............................................................................................................................. VI 

CHAPTER 1 16 

1. INTRODUCTION ................................................................................................................. 16 

Hypothesis 19 

Aim and objectives. ............................................................................................................... 19 

1.1 Aim: .................................................................................................................. 19 

1.2 Objectives: ........................................................................................................ 19 

CHAPTER 2 20 

2. LITERATURE REVIEW ....................................................................................................... 20 

2.1 Malaria 20 

2.2 Plasmodium falciparum life cycle .................................................................................. 21 

2.3 Uncomplicated and severe form of malaria ................................................................... 23 

2.4 Past and present antimalarial treatments ...................................................................... 24 

2.5 Molecular chaperones as drug targets .......................................................................... 27 

2.6 Hsp90 29 

2.7 Mechanism of Hsp90 ATPase activity ............................................................................ 30 

2.8 Small molecule inhibitors of Hsp90 ............................................................................... 31 

2.9 P. falciparum Hsp90 (PfHsp90) as a drug target ........................................................... 32 

2.10 Drug discovery .............................................................................................................. 33 

2.10.1 Rational drug design ..................................................................................................... 34 

2.10.2 Application of molecular docking in drug discovery ....................................................... 35 



 

viii 

2.10.3 Molecular dynamic simulation ....................................................................................... 36 

2.10.4 Molecular Mechanics/Generalized Born Surface Area (MMGBSA) ............................... 37 

2.10.5 Machine learning (ML) in drug discovery ...................................................................... 38 

2.10.6 Quantitative structure activity relationship (QSAR) ....................................................... 38 

2.11 ATPase activity assay ................................................................................................... 39 

2.12 Biophysical methods in early drug discovery ............................................................. 40 

2.12.1 Surface Plasmon resonance ......................................................................................... 40 

CHAPTER 3 42 

3. MATERIALS AND METHODS ............................................................................................ 42 

3.1. In silico methods ............................................................................................................ 42 

3.1.1 Computational software programme ............................................................................... 42 

3.1.2 Protein retrieval and preparation ..................................................................................... 44 

3.1.3Ligand preparation ........................................................................................................... 44 

3.1.4 Induced fit docking (IFD) ................................................................................................. 44 

3.1.5 Reaction-based enumeration .......................................................................................... 45 

3.1.6 Classical glide SP ligand docking ................................................................................... 46 

3.1.7Receptor grid generation ................................................................................................. 47 

3.1.8 Active learning model generation .................................................................................... 47 

3.1.9 In silico ADME predictions .............................................................................................. 48 

3.1.10 Free binding energy calculations .................................................................................. 49 

3.1.11 Molecular dynamic simulation ....................................................................................... 50 

3.2 Biochemical methods...................................................................................................... 50 

3.2.1 Transformation ............................................................................................................... 50 

3.2.2 Plasmid extraction .......................................................................................................... 50 

3.2.3 Restriction enzyme digest ............................................................................................... 51 

3.2.4 Protein expression .......................................................................................................... 51 

3.2.5 Protein purification .......................................................................................................... 51 

3.2.6 Protein concentration determination ............................................................................... 52 

3.2.7 ATPase activity assay..................................................................................................... 53 

3.3. Biophysical methods...................................................................................................... 53 

3.3.1 Surface plasmon resonance (SPR)................................................................................. 53 

CHAPTER 4 55 



 

ix 

4.1 RATIONAL DESIGN RESULTS AND DISCUSSIONS. ..................................................... 55 

4.1.1 Reaction-based enumeration ....................................................................................... 55 

4.1.2 Active learning model enhanced by ligand docking. ................................................. 57 

4.1.4 Induced fit docking. ...................................................................................................... 60 

4.1.5 ADME/T property predictions ...................................................................................... 65 

4.1.6 Molecular dynamics simulations. ................................................................................ 69 

4.1.3 Binding free energy calculations................................................................................. 76 

CHAPTER 5 78 

5.1 BIOCHEMISTRY RESULTS AND DISCUSSIONS ............................................................ 78 

5.1.1 Confirming the integrity of the pET28b (+)-PfHsp90-NTD construct. ........................ 78 

5.1.2 Recombinant PfHsp90 expression .............................................................................. 78 

5.1.3 Protein purification ....................................................................................................... 79 

5.1.4 ATPase assay ............................................................................................................... 81 

CHAPTER 6 84 

6.1 BIOPHYSICAL ANALYSIS ............................................................................................... 84 

6.1 Surface plasmon resonance (SPR) ................................................................................ 84 

CHAPTER 7 88 

CONCLUSION ........................................................................................................................ 88 

Future recommendations...................................................................................................... 89 

BIBLIOGRAPHY ..................................................................................................................... 90 

APPENDIX ............................................................................................................................ 102 

1.1.3Reagents ....................................................................................................................... 102 

1.1.4 Plasmid 103 

1.1.5 Transformation ............................................................................................................. 103 

1.1.6 Pure-yield plasmid extraction kit ................................................................................... 103 



 

x 

1.1.7 Restriction digest .......................................................................................................... 105 

1.1.8 Agarose gel electrophoresis ......................................................................................... 106 

1.1.9 Sodium dodecyl-polyacrylamide gel electrophoresis (SDS-PAGE) ............................... 106 

1.1.10 Western blotting .......................................................................................................... 108 

RESULTS 109 

2.1 Molecular dynamic simulations .................................................................................... 111 

 112 

  



 

xi 

LIST OF TABLES  

Table 1: Pathfinder generated reaction pathways -Schrödinger Suite 2021-3 ......................... 45 

Table 2: IFD results of the reference compound 7 represented as CP-7 and compound 10 

as CP-10. ..................................................................................................... 55 

Table 3: The best Auto QSAR models were used to train the listed design concepts to 

predict their docking scores. ......................................................................... 58 

Table 4: induced fit docking results of the purchased compounds showing Docking score, 

glide energy, glide emodel of Induced fit docked complexes. ....................... 61 

Table 5: in silico physio-chemical property predictions by Qikprop .......................................... 66 

Table 6: Binding free energy calculation for investigated compounds using Prime MM-

GBSA ........................................................................................................... 76 

Table 7: Kinetics constants measured by SPR for the interaction between tested 

compounds and immobilized PfHsp90.......................................................... 84 

  



 

xii 

LIST OF FIGURES  

 

Figure 1: Life cycle of P falciparum.. ....................................................................................... 22 

Figure 2: Chemical structures of the past antimalarials. .......................................................... 25 

Figure 3: Diagrammatic sketch of the full length P. falciparum Hsp90. .................................... 29 

Figure 4: Model of the Hsp90 ATPase cycle showing the obligatory sequential hydrolysis 

and conformational switching. ...................................................................... 30 

Figure 5: in silico methodology flow diagram .......................................................................... 43 

Figure 6: Oxadiazole-1 selected pathway. .............................................................................. 46 

Figure 7: reference compounds.. ............................................................................................ 55 

Figure 8: Details the ligand interaction diagrams for compound 10 and 7 in the PfHspo90 

N-terminal domain ATP binding site.. ........................................................... 56 

Figure 9: Auto QSAR active learning scatter plot for model 1 showing the performance of 

the QSAR KPLS model’s predicting activity for experimental binding 

affinity for the test set. .................................................................................. 59 

Figure 10: Auto QSAR active learning plots for model 2 ......................................................... 59 

Figure 11: Auto QSAR active learning plots for model 13 with an R2 of 0.7460 and a Q2 of 

0.6182 .......................................................................................................... 60 

Figure 12: PfHsp90 (PDBcode:3K60) interaction with six selected compounds. ..................... 64 

Figure 13: Compound structures analysed by Qikprop for the in silico physio-chemical 

property predictions. ..................................................................................... 69 

Figure 14: Protein-ligand RMSD plots of all the frames in the trajectory for PfHsp90-NTD 

and Compounds complexs at a 50ns MDS trajectory and FTN-T2 at 

150ns. ............................................................. Error! Bookmark not defined. 

Figure 15: 2D Protein-ligand interaction diagrams of all the frames in the trajectory for 

PfHsp90-NTD and Compounds complexes at a 50ns MDS trajectory and 

FTN-T2 at 150ns.. ........................................................................................ 74 

https://d.docs.live.net/5aa5c10f71ce73f5/Desktop/Thato_Dissertation_11062023.docx#_Toc138418466
https://d.docs.live.net/5aa5c10f71ce73f5/Desktop/Thato_Dissertation_11062023.docx#_Toc138418466


 

xiii 

Figure 16: Restriction enzyme digest analysis for pET28b (+)-PfHsp90-NTD plasmid 

construct, ..................................................................................................... 78 

Figure 17: PfHsp90-NTD expression on BL21(DE3) E. coli cells analysis. .............................. 79 

Figure 18: SDS-PAGE analysis of PfHsp90 protein purification samples. ............................... 80 

Figure 19: ATPase activity of PfHsp90 against our inhibitors. Compounds were added in 

increasing concentrations (0.25–20 μM) to a fixed amount of PfHsp90 

(2 μM) to investigate its effect on the ATPase activity of PfHsp90 with 

Harmine as our positive control. ................................................................... 83 

Figure 20: Sansogram from the Analysis of the interaction of FTN-T2, FTN-T5, FTN-T10, 

geldanamycin and Harmine with PfHsp90-NTD by surface plasmon 

resonance spectroscopy. ............................................................................. 86 

 

LIST OF ABREVIATIONS  

 2D    Two dimensional 

 3D    Three dimensional 

 ACTs     Artemisinin-based combination therapies 

 ADME     Absorption, distribution, metabolism, and excretion  

 ADP     Adenosine diphosphate 

 AL     Active learning  

 ATP    Adenosine triphosphate 

 BSA     Bovine serum albumin  

 CD     Circular dichroism  

 CDC    Centre for Disease Control  

 CTD    C-terminal domain 

 DMSO    Dimethyl sulfoxide 

 DNA    Deoxyribonucleic acid 

 DTT     Dithiothreitol 

 EC50    Half maximal effective concentration 

 EGFR     Epidermal growth factor receptors  

 GA     Geldanamycin 



 

xiv 

 GIST     Gastrointestinal stromal tumours  

 GUI     Graphical user interface  

 HBA    Hydrogen bond acceptor 

 HBD    Hydrogen bond donor 

 HEPES   4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid 

  HER2     Human epidermal growth factor receptor 2 

 HIV    Human immunodeficiency virus 

 HSP     Heat shock proteins 

 Hsp90    Heat shock protein 90  

 IC50    Half-maximal inhibitory concentration 

 IFD    Induced fit docking 

IPCC     Intergovernmental Panel on Climate Change 

 IPTG    Isopropylthio-β-galactoside 

 ITC     Isothermal titration calorimetry   

 KCl    Potassium chloride     

 KDa     Kilo Daltons  

 KIT     Receptor tyrosine kinase 

 LogP     n-Octanol/water partition coefficient 

  MD     Middle domain 

 MDS    Molecular Dynamics Simulations 

 MgCl2    Magnesium chloride 

 ML     Machine learning 

 MM-GB/SA    Molecular mechanics with generalised Born and surface             

area solvation 

 MW    Molecular weight 

 NHS    N-hydroxysuccinimide 

 NMR     Nuclear Magnetic Resonance  

  NTD    N-terminal domain 

 OPLS4    Optimized potential for liquid simulations 4 

 PAINS    Pan assay Interfering Structures 

 PBS     Phosphate buffer saline  

  PCR     Polymerase chain reaction 



 

xv 

  PDGFRA    Platelet Derived Growth Factor Receptor Alpha 

 PfCRT    Chloroquine resistance transporter  

  PfHsp90    P. falciparum Heat shock protein 90 

 Pi    Inorganic phosphate 

 PMSF    Phenylmethylsulfonyl fluoride 

 QSAR     Quantitative structure-activity relationship  

 RDD     Rational drug design 

 RNA    Ribonucleic acid 

 ROS     Reactive oxygen species 

 SASBMB   South African society of biochemistry and molecular 

biology 

 SDS-PAGE    Sodium dodecyl-sulfate polyacrylamide gel electrophoresis 

 SMARTS    Smiles arbitrary target specification  

 SP     Standard precision 

 SPR     Surface plasmon resonance  

 TPR     Tetratricopeptide repeat  

 TPSA    Topological polar surface area 

 WHO    World health organization  

 XP     Extra precision 

 YT    Yeast tryptone 

 



 

16 
 

CHAPTER 1  

1. INTRODUCTION 

Malaria is an infectious disease that is transmitted to humans by female Anopheles mosquito 

vectors. The population most at risk of malaria infection is mainly children under the age of 5 and 

pregnant women (Steketee et al., 2001; Gontie et al., 2020), with an estimated 619 000 individuals 

succumbing to the deadly disease in 2021 (WHO, 2022). Most of the fatalities associated with 

malaria are attributed to what is considered the deadliest malarial species, Plasmodium 

falciparum (Ismail et al., 2002; Battle et al., 2019; Weiss et al., 2019). The emergence of drug-

resistant parasitic strains has greatly hindered the fight against malaria, limiting the effectiveness 

of the current treatment options (Malaguarnera and Musumeci, 2002).  

 

Current treatment options for malaria include vaccines, insecticides, and therapeutic drugs. 

However, each strategy has its own drawbacks. For instance, the most advanced malaria vaccine 

is Mosquirix (RTS) S/AS01 (Nadeem et al., 2022), provides partial protection against P. 

falciparum. However, its efficacy is limited, and multiple doses are required to achieve a moderate 

level of protection. Furthermore, the duration of protection offered by vaccination is short 

(Laurens, 2020; Syed, 2022). Common strategies for controlling malaria-transmitting mosquitoes 

include insecticide-treated bed nets and indoor residual spraying (Okumu and Moore, 2011; Pryce 

et al., 2022). While these interventions have proven effective in reducing malaria transmission, 

they face challenges, such as insecticide resistance in mosquitoes. Over time, mosquitoes have 

developed resistance to commonly used insecticides, which limits the effectiveness of this 

approach (Okumu and Moore, 2011).  Drugs such as artemisinin-based combination therapies 

(ACTs) as the first-line treatment for uncomplicated malaria (WHO, 2022), face the major 

challenge of drug resistance, which has significantly hindered the effectiveness of these drugs, 

leading to treatment failure and increased mortality rates of P. falciparum parasites. Although 

vaccines, insecticides, and drugs are important tools in the fight against malaria, their limitations, 

and challenges, such as limited efficacy, insecticide resistance, and drug resistance, highlight the 

need for continuous research and development of novel and more effective strategies for malaria 

prevention and treatment. 

 

Environmental stress caused by the cyclic transition from the mosquito vector to the human host 

of the P. falciparum parasite, leads to the expression of an arsenal of molecular chaperones, such 

as heat shock protein 90 (Hsp90), Hsp70, Hsp40 and others, to ensure parasite survival and 



 

17 
 

proliferation. These environmental stresses include changes in the pH, temperature, and nutrient 

supply (Pavithra et al., 2004; Uwimana et al., 2020). Hsp90 is one of the most important and 

highly expressed members of the molecular chaperones, constituting approximately 2% of the 

total proteome (Fink, 1999; Garnier et al., 2002; Hunter et al., 2014). Hsp90 is a key regulator of 

vital biological processes, including stress control, protein folding, DNA repair, development, 

immunological response, and signalling pathways (Aligue et al., 1994; Pratt, 1998; Pearl, 2005; 

Prodromou, 2016). Hsp90 interacts with a wide range of proteins referred to as client proteins. 

Some of the Hsp90 client proteins are transcription factors, growth receptors, and kinases 

(Prodromou, 2016), making Hsp90s a desirable pharmacological target for a variety of illnesses, 

including cancer, neurological disorders, and infectious diseases such as malaria, owing to their 

crucial role in cell survival. 

 

P. falciparum Hsp90 (PfHsp90) plays a critical role in assisting with the folding of newly 

synthesized proteins, regulating the actions of transcription factors and protein kinases, and acts 

as a buffer to mitigate the cellular toxicity caused by misfolded or aggregated proteins in response 

to stress, growth, and development during its host-vector cyclic infection stages (Banumathy et 

al., 2003; Gitau et al., 2012). Targeted gene mutations of some client proteins of the molecular 

chaperone have been shown to give rise to the development of drug resistance. For example, the 

Shihanas study (2013) showed that PfHsp90 interacts directly with the chloroquine resistance 

transporter (crt) (Shaninas et al., 2013; Stofberg et al., 2021). Furthermore, multidrug resistance 

genes, such as crt, are located on the same gene cluster as Hsp90, which plays a central role in 

protein quality control (Zininga, 2015). Ashly et al. (2015) also linked artemisinin resistance to an 

enhanced stress response, implicating PfHsp90. Therefore, PfHsp90 is an attractive drug target 

for the development of effective and dependable antimalarial drugs that inhibit P. falciparum 

infection and growth while reducing the possibility of developing resistance to treatment.  

 

Hsp90 is a dimeric protomer composed of three domains: the N-terminal domain (NTD), which 

binds ATP, connected to the middle domain (MD) through a charged linker region, and the C-

terminal domain (CTD) (Jackson, 2013). NTD inhibitors of Hsp90 act by binding to the ATP-

binding pocket, thereby preventing the binding of ATP to Hsp90 and inhibiting ATPase activity. 

This, in turn, prevents conformational changes required for Hsp90 to interact with its client 

proteins and promote their folding and stability. Most of the Hsp90 inhibitors target the NTD by 

competing with ATP for binding and therefore, abrogating the enzymatic activities of the molecular 

chaperone (Chène, 2002; Huang et al., 2017; Wang et al., 2019; Shin et al., 2020; Banerjee et 

al., 2021). 

 



 

18 
 

Using known human Hsp90 inhibitors to investigate the function and essentiality of PfHsp90 over 

the years, several studies have validated PfHsp90 as a drug target for the development of novel 

antimalarial drugs that would most likely circumvent resistance mechanisms (Kumar et al., 2003; 

Banumathy et al., 2003; Pallavi et al., 2010; Shahinas et al., 2013). Critical findings have been 

obtained using these known inhibitors, including differences in the binding affinities of the 

inhibitors to human Hsp90 and PfHsp90. Geldanamycin (GA) was found to have a higher affinity 

for PfHsp90 than for human Hsp90 (Kumar et al., 2003). Secondly, PfHsp90 has higher ATPase 

activity than its human homolog, making it more susceptible to inhibition (Pallavi et al., 2010; 

Shahinas et al., 2013). Thirdly, GA was also able to halt the development of P. falciparum cultures 

in the intraerythrocytic cycle of the parasite between the ring and trophozoite stages (Banumathy 

et al., 2003), drastically reducing the number of merozoites released into the bloodstream. The 

elucidation of the PfHsp90 crystal structure by Corbett and Berger in 2010 has been an important 

breakthrough in validating PfHsp90 as a drug target by providing more avenues to identify unique 

conformational differences between PfHsp90 and Hsp90s from other organisms.  

 

The identification of this drug target offers an advantage in the implementation of rational drug 

design strategies, such as those documented by Egan (2005), Plouffe et al. (2008), Sharman et 

al. (2012), Wang et al. (2016), Everson et al. (2021), and Tamari (2022), over traditional 

phenotypic screens. Using this approach, inhibitors with high potency and modest affinity have 

been identified (Wang et al., 2016; Everson et al., 2021). Rational drug design strategies form the 

basis of multidisciplinary approaches required for the drug development process (Mandal et al., 

2009), overcoming the challenges encountered with traditional screening strategies, including 

very high costs, long turnaround times, and specificity of target proteins (Reddy and Parrill, 1999). 

Selective targeting is essential for drug discovery because of its increased efficacy and reduced 

toxicity. Broadly, rational drug design strategies can be classified into two categories: firstly, small-

molecule development with desired qualities for targets, biomolecules (proteins or nucleic acids) 

with established functional functions in cellular processes, and 3D structural information. The 

second is the development of small molecules with predefined target properties (Mandal et al., 

2009). In recent years, to accelerate and improve drug discovery and development, machine 

learning (ML) techniques have been increasingly integrated into rational drug design. Therefore, 

this study shows the synergy between active learning-based Glide docking tools, and Pathfinder 

reaction-based enumeration tools, highlighting their combined effectiveness in the context of drug 

discovery. 

 

The present study implemented a machine learning approach, active learning (AL), which is a 

classification under supervised ML, to design novel inhibitors with a starting point of previously 

described molecules with binding affinity towards PfHsp90 (Everson et al., 2021). The active 



 

19 
 

learning approach helps develop highly accurate quantitative structure-activity relationship 

(QSAR) models, effectively exploring wide chemical spaces with docking and deep learning 

(Dixon et al., 2016; Yang et al., 2021), followed by validation of novel inhibitors through 

biochemical and biophysical methods. 

Research question 

Can active learning-based Glide docking be utilized to generate novel inhibitors with 

distinct chemical scaffolds and enhanced affinity for PfHsp90, resulting in the disruption 

of PfHsp90's enzymatic properties and demonstrate notable binding affinity towards 

it?Aim and objectives 

1.1 Aim:  

The aim of this study was to generate small molecules with selective inhibition of PfHsp90 using 

a rational drug design coupled with biochemical and biophysical validation techniques.  

 

1.2 Objectives: 

The key objectives of this study were to: 

 

 generate inhibitors of PfHsp90 using Pathfinder reaction-based enumeration from known 

inhibitors taken from literature, 

 identify compounds that bind to PfHsp90 through active learning-based Glide docking and 

Auto QSAR techniques, 

 assess the generated inhibitors ADME (absorption, distribution, metabolism, and 

excretion) properties and safety profile in silico, 

 evaluate the conformational stability of the interactions between PfHsp90 and the 

generated compounds by molecular dynamics simulations, 

 conduct heterologous overexpression and purification of PfHsp90-NTD, 

 evaluate the inhibitory capabilities of generated compounds using ATPase assays, 

 assess the biophysical interactions between selected inhibitors and PfHsp90 using 

surface plasmon resonance (SPR).  

 

 



 

20 
 

CHAPTER 2  

2. LITERATURE REVIEW 

2.1 Malaria  

Recent estimates indicate that approximately 247 million people were affected and 619 000 

succumbed to the deadly vector-borne parasitic disease of malaria globally in the year 2021 

(WHO, 2022). Children under the age of five and expectant mothers are the most severely 

affected population members in areas with high transmission rates, such as sub-Saharan Africa 

(Steketee et al.,2001; van Eijk et al., 2015; Gontie et al., 2020). In the malaria-endemic regions 

of Africa, where an estimated 30 million women become pregnant (WHO, 2003), malaria infection 

increases the risk of maternal anaemia and delivers a low birth weight (LBW) baby (<2,500g), 

which in turn, is a significant risk factor for infant mortality (Guyatt and Snoe, 2004).  

 

The main etiological agents of malaria are Plasmodium species that are transmitted to the human 

host via a bite from an infected female Anopheles mosquito (Sandosham, 1984; Beck-Johnson 

et al., 2013). Human malaria is mainly caused by six Plasmodium species, including P. falciparum, 

P. vivax, P. malariae, P. knowlesi, P. ovale (spp crustisi and wallikeri) (Schofield and Grau, 2005; 

Ashley et al., 2018), with the greatest threat posed by P. falciparum and P. vivax. Globally, 

approximately 95% of all infections are caused by P. falciparum and P. vivax (Battle et al., 2019; 

Weiss et al., 2019). P. falciparum and P. vivax are quite distinct organisms based on their 

morphological characteristics, ecology, and biology. P. falciparum has a high frequency in all 

tropical areas of the world (Manske et al., 2012), with the highest prevalence in sub-Saharan 

Africa, whereas P. vivax is mostly found in Asia, the Middle East, and South America, but rarely 

in select regions of Africa (Rougeron et al., 2021). Additionally, these two parasites exhibit 

biological differences, including features of various life cycle stages, morphology, mean 

parasitemia, and duration, among which the two are particularly noticeable. P. vivax preferentially 

invades young erythrocytes and can present a quiescent form in the liver that can cause relapses 

(dormant hypnozoites), which is not the case for P. falciparum (Bourgard et al., 2018). Virulent P. 

falciparum can lead to life-threatening complications such as cerebral malaria, severe anaemia, 

and organ failure (Mohanty et al., 2003; Kochar et al., 2006; Doumbo et al., 2010). The symptoms 

of uncomplicated P. falciparum malaria include high fever, chills, headache, muscle pain, fatigue, 

nausea, vomiting, and diarrhoea (Centers for Disease Control and Prevention (CDC), 2022). In 

severe cases, patients may experience seizures, coma, and respiratory distress (Idro et al., 2010).  

In contrast, P. vivax malaria is usually less severe and dormant for a long period (Dayanand et 



 

21 
 

al., 2018). However, P. vivax malaria can also cause relapse of the disease, which is uncommon 

in P. falciparum (White, 2011).  

 

Malaria risk has always been geographically specific, predominantly occurring in tropical and 

subtropical regions (Gallup and Sachs, 2001), because temperature and aridity are favourable for 

the spread of Anopheles vectors (Guerra et al., 2008). Factors such as environmental stability, 

climate, altitude, vegetation, and poor implementation of control measures, play a major role in 

disease incidence (Autino et al., 2012). Several studies employing global climate change 

projections and population-based studies have identified climate change as one of the major 

contributing factors to the distribution of malaria in high-risk areas such as sub-Saharan Africa 

(IPCC, 2014; Watts et al., 2020; Sweileh, 2020). Environmental factors such as the rise in 

temperature, precipitation, and extreme weather events, including floods, heat waves, and 

storms, are influenced by climate change, thereby having a direct impact on the epidemiology of 

malaria, as they are favourable for the survival and transmission of the mosquito vectors of 

malaria (Kulkarni et al., 2022). Therefore, urgent efforts are needed to identify effective, novel, 

and affordable antimalarials with alternative modes of action to control the spread of mosquito 

vectors. 

 

2.2 Plasmodium falciparum life cycle  

The life cycle of P. falciparum involves two hosts, as infection occurs between the human host 

and mosquito vector (Summarized in Figure 1). Human host infection begins when an infected 

female Anopheles mosquito consumes a blood meal by biting the host. In this way, sporozoites 

already present on the salivary glands of the mosquito are inoculated into the bloodstream of 

humans (Figure 1). Approximately 45 minutes after the bite, sporozoites travel through the blood 

circulation and enter liver parenchymal cells (hepatocytes) (Figure 1). Each sporozoite starts an 

asexual reproductive phase inside the hepatocytes (pre-erythrocytic liver stage), which takes 

approximately 1-2 weeks resulting in the formation of schizonts that contain thousands of 

merozoites (Figure 1). Merozoites are released into the bloodstream when mature schizonts 

burst. For P. falciparum, the hepatic phase of parasite development (hepatic schizogony) lasts for 

approximately five days. Merozoites quickly infiltrate red blood cells after entering the bloodstream 

to begin asexual reproduction (erythrocytic schizogony) (Figure 1). Once red blood cells are 

invaded, the P. falciparum parasite enters the ring stage for a period of 20-24 hours, where 

merozoites develop inside the red blood cells into trophozoites and schizonts (Sherman, 2005; 

Bannister and Mitchell, 2003), which then burst to release the next generation of merozoites, 

followed by infiltration of additional red blood cells and completion of the erythrocytic cycle 



 

22 
 

(Bannister and Mitchell, 2003; Kaussis et al., 2009). When a schizont ruptures, parasites and 

erythrocytic material are released into the bloodstream, initiating the pathophysiology of malaria 

and the onset of symptoms. 

 

 

 

Figure 1: Life cycle of P falciparum. 1, migration of sporozoites through the blood stream after a bite from the infected 
female Anopheles mosquito to invade hepatocytes. 2, the sporozoites differentiate and mitotically divide into thousands 
of merozoites which are subsequently released and enter the intra-erythrocytic cycle (asexual blood stage), while a 
sub-population of the merozoites switch to sexual development gametocytes. 3, the gametocytes are ingested by 
female Anopheles mosquito and undergo development, which results in sporozoites in the salivary glands of the 
mosquito. Figure adapted from Tankeshwar., 2021.  

 

Male and female gametes are produced by a subpopulation of the merozoites that would have 

switched to sexual development (Josling and Llinas, 2015; Bousem and Drakely, 2016). Five 

morphologically recognizable stages take place for the formation and maturation of gametocytes. 

These distinctive transitional stages transmit malaria to the mosquito via a blood meal. Once 

ingested by mosquitoes, each individual gametocyte forms 1 female macrogamete or up to 8 male 

microgametes. In the mosquito midgut, the fusion of gametes results in the formation of a zygote 

that develops into a motile ookinete that can penetrate the midgut wall to form oocysts. The 



 

23 
 

oocysts enlarge over time and burst to release sporozoites that migrate to the mosquito salivary 

gland, rendering the mosquito infectious to human beings (Figure 1) (Bousem and Drakely, 2016).  

2.3 Uncomplicated and severe form of malaria 

The onset of symptoms greatly depends on factors such as the degree of immunity, type of 

infecting parasite species and density of parasite inocula, all having a great impact on the duration 

of the incubation period (Bartoloni & Zammarchi, 2012). For P. falciparum, the incubation period 

lasts from 9 to 30 days and uncomplicated malaria initial symptoms are non-specific and common 

to all fever infections. These include flu-like manifestations, often rendering clinical diagnosis 

unreliable (Bartoloni & Zammarchi, 2012). It is precisely for this reason that sensitive molecular 

diagnostic tools such as Polymerase chain reaction (PCR), should be employed as crucial 

detection tools of malaria (Erdman and Kain, 2008). The most common sign of malaria infection 

is high fever (CDC, 2022). Prodromal symptoms, such as chills, body-aches, headache, cough, 

diarrhoea, and vomiting, may appear two days prior to the onset of fever (Warrell, 2017). The 

fever generally manifests irregularly and fluctuates over time accompanied by shaking and light 

chills (Mayo clinic, 2023). The fever in P. falciparum malaria may occur every 48 hours, but is 

usually irregular, showing no distinct periodicity (Crutcher and Hoffman, 1996). 

 

In the case of severe malaria, symptoms affect the central nervous system (cerebral malaria), 

pulmonary system (respiratory failure), renal system (acute renal failure), resulting in severe 

anaemia (Bunnang, 1988). Any malaria patient must be evaluated and get the correct treatment 

as the development of these complications can be swift, and because the illness can be deadly 

at this phase (Day and Dondorp, 2007). Frequent observations are necessary to assess for the 

first indications of systemic problems. On the other hand, cerebral malaria presents as acute lung 

injury, which can develop into breathing difficulties (25% of patients), acidosis, typically presenting 

as acute tubular necrosis, and acute kidney injury (Taylor et al., 2012). 

 

The advancement of the most prevalent fatal consequences of malaria (cerebral malaria, 

respiratory failure and acute renal failure) and development of severe malaria from uncomplicated 

malaria are greatly dependent on immunological and epidemiological factors. Factors such as 

lack of prompt response to the early symptoms in children, lack of knowledge of the early malaria 

symptoms, use of wrong malaria medication and dosage, closeness and accessibility of health 

centres (Anumudu et al., 2007). In the absence of immune compromise and parasite resistance 

to available drugs, the frequency of malaria infection can be reduced, and the development of 

severe malaria and its consequences monitored, if caregivers are aware of the symptoms and 

treat them promptly with appropriate therapies. 



 

24 
 

2.4 Past and present antimalarial treatments 

Even though malaria infection numbers have reduced from 245 million in 2000 to 230 million in 

2015 across the 108 countries that were malaria endemic in 2000, recent reports have shown 

that malaria cases have surged since 2016, with the largest yearly increase of 13 million cases 

reported between 2019 and 2020, during the first year of the COVID-19 pandemic (WHO, 2022). 

Emphasizing the importance of avoiding complacency with current treatment and preventive 

techniques (Cheney, 2018; Mwangi, 2018; Tse et al., 2019). Antimalarial drugs may also result in 

a prolonged incubation period, due to their effects on the parasite’s multiplication rate, although, 

ineffective for treating malaria (Bartoloni & Zammarchi, 2012). 

 

The treatment of malaria depends on several factors, including the severity of the infection, the 

parasite strain responsible for the infection, and whether the patient presents with symptoms of 

severe or uncomplicated malaria (Bartoloni and Zammarchi, 2012). WHO has recommended 

artemisinin-based combination therapies (ACTs) as first-line treatment option for P. falciparum 

malaria, as well as supportive care measures such as hydration and management of fever and 

other symptoms (WHO, 2022). Severe or complicated malaria treatment may need hospitalization 

and reinforcement through administration of intravenous fluids (such as normal saline, dextrose 

saline, dextrose, Hartmann's solution, Ringer's lactate, but not including human albumin or 

plasma substitutes such as gelatin) and oxygen therapy (Hodgoson and Angus, 2016). The risk 

of parasitic drug resistance is a growing concern in the treatment of malaria. WHO has identified 

drug resistance as one of the key challenges to malaria control and elimination efforts (WHO, 

2020). Gene mutations in the P. falciparum parasite leads to the development of drug resistance 

if antimalarial medications are taken improperly or often, which might result in treatment failure 

(Shibeshi et al., 2020). It is challenging to control and eradicate malaria in afflicted areas because 

of this resistance's fast expansion (White, 2004; Okeke et al., 2005; Buabeng et al., 2007; 

Shibeshi et al., 2020) 

 

Quinine was the first chemically pure alkaloid successful therapy for malaria in the 1820’s that 

comes from the bark of the cinchona tree (Cinchona pubescens), introduced to Europe in the 17th 

century from Peru (Greenwood, 1992), and since then additional organic and synthetic chemicals 

have been developed [Figure 2], (Tse et al., 2019). Quinine works by inhibiting haemoglobin 

breakdown in Plasmodium parasites, which is a comparable mode of action to chloroquine 

(Blasco et al., 2017). Regrettably, mutations hindering proper drug transport have rendered this 

treatment less effective over time (Capela et al., 2019), as parasite developed resistance to these 

medications, decreasing their efficacy. As a result, they are no longer in use. 



 

25 
 

 

Figure 2: Chemical structures of the past antimalarials that were used.  

 

Since its discovery in the 1820’s, quinine was one of the most effective antimalarial treatments 

(Reviewed by Achan et al., 2011). The first case of resistance was reported in the 1980’s (Bunnag 

et al., 1996), and as of 2006, quinine was no longer used as the first line therapy for malaria. 

 

During the 1930s, the antimalarial chloroquine was developed and became widely used as an 

antimalarial during the 1960’s, having fewer side effects (Loeb et al., 1946). Both quinine and 

chloroquine are quinoline containing drugs, targeting the heme degradation pathway of the 

parasite. The compounds use a weak base mechanism to kill the malaria parasite by inhibiting 

haemoglobin degradation from occurring in the acidic food vacuoles of the intraerythrocytic 

trophozoites (Slater, 1993; Fitch 2004; Bray et al., 2005). The first cases of chloroquine resistance 

arose in South-East Asia and South America within a decade of its introduction.  

 



 

26 
 

Ever since then, it has expanded slowly but unabatedly throughout endemic regions, eventually 

reaching Africa in the late 1970s (Bray et al., 2005; Mita et al., 2009; Packard, 2014). Past studies 

on P. falciparum, the most dangerous form of human-infecting malaria parasite, have revealed 

that point mutations in the chloroquine resistance transporter (PfCRT) protein are crucial for the 

development of chloroquine resistance (White, 2004; Bray et al., 2005). Usage of chloroquine 

was discontinued from the treatment of P. falciparum infection in almost all endemic regions due 

to the global spread of resistance (Balikagala et al., 2020). 

 

Artesunate is an antimalarial drug that belongs to the artemisinin class of compounds. It is one of 

the active ingredients in artemisinin-based combination therapies (ACTs), which are the 

recommended first-line treatment for uncomplicated malaria caused by P. falciparum. The 

mechanism of action of artesunate include the formation of reactive oxygen species (ROS) in the 

malaria parasite (Gopalakrishnan and Kumar, 2015). Artesunate is triggered by heme, which is 

released from the parasite's hemoglobin during red blood cell digestion (Tilley et al., 2016). The 

activated artesunate then combines with the heme to form reactive oxygen species (ROS), which 

damage the parasite's cell membrane, proteins, and DNA, ultimately leading to death 

(Gopalakrishnan and Kumar, 2015). Artemisinins are currently the established antimalarial agents 

with a good safety profile. However, artesenuate needs to be complimented by other drugs to be 

effective, as the key objective when treating malaria is not only eliminating parasites but also 

ensuring complete clearance within a reasonable period without any chances of recurrence. 

Artesenuate is fast acting and possess a short half-life. Therefore, alone, artesenuate cannot 

provide long-lasting immunity against malarial parasites (Borrmann et al., 2003), additional 

medications are often necessary for enhanced efficacy duration and prevention of reoccurrence 

post treatment cycles completion. Clinical trials with artemisinin monotherapies have shown that 

these compounds are taking considerably longer to clear malaria infections in Southeast Asia 

typically twice as long as observed a decade ago. 

  

Drugs like mefloquine or lumefantrine have been identified as viable options to facilitate prolonged 

action while inhibiting the emergence of malarial drug resistance due to their long half-life (Ruwizhi 

et al., 2022). Artesunate targets intra-erythrocytic asexual blood-stage malaria parasites, and 

generally partner drugs in the ACT complemented by targeting multiple stages of the parasite's 

life cycle, increasing the likelihood of complete parasite clearance, and lowering the risk of 

recurrence (Tilley et al., 2019). This is significant because the Plasmodium parasite has a 

complicated life cycle that includes several stages, including the liver, blood, and transmission 

phases (Nosten and White, 2007). WHO advised the use of ACTs as the first-line therapy for P. 

falciparum malaria in all endemic countries in 2005 (Nosten and White, 2007; WHO, 2015). 

Malaria-related morbidity and death have decreased because of the use of artemisinin-based 



 

27 
 

combination therapy in place of the inefficient, failed treatments (Ashley et al., 2014). There are 

about 67 countries that are affected by P. falciparum malaria, 41 of which are in Africa, and have 

adopted the ACTs as a treatment option (Bosman and Mendis, 2007).  

 

Extensively used ACTs include artesunate-amodiaquanine, artemther-lumefantrine, 

dihydroartemisinin-piperaquine, artesunate-mefloquine and arteunate plus sulfadoxine-

pyrimethamine (Ouji et al., 2018). The WHO (2015) also provided evidence-based guidelines for 

the treatment of severe malaria. Individuals with severe malaria must be administerd artesunate 

, either intravenously or intramuscularly until they are able to take the medication orally (Esu et 

al., 2014). There have been issues with the solubility of artesunate, hence it is mainly taken as a 

tablet (Blasco et al., 2017). Then the patient can be given ACTs treatment, which must be 

completed within three days. However, clinical trials with artemisinin monotherapies have shown 

that these compounds are taking considerably longer to clear malaria infections in Southeast Asia 

typically twice as long as observed a decade ago (Corey et al., 2016). Improper usage or 

incompletion of the ACT is likely to result in the resistance to ACTs by the P. falciparum parasite 

(Onyango et al., 2012). The fundamental reason that ACTs may contribute to the establishment 

of resistance is that artemisinin derivatives have a short half-life, which means they are quickly 

cleared from the body. Some parasites may be able to survive the initial artemisinin treatment 

and develop resistance to the drug because of this. If these resistant parasites are allowed to 

proliferate, they can spread and become dominant, resulting in treatment failure and the 

establishment of artemisinin-resistant malaria parasite strains (Nsanzabana, 2019). Although 

ACTs have been efficient in tackling malaria, there is already emerging evidence of increasing 

treatment failures in parts of south Asia (Ashley et al., 2018) and Rwanda (Rosenthal, 2021). In 

addition, concerns regarding the cost and sustainability of ACTs in Africa have been raised 

(Oladipo et al., 2022). In this perspective, it is critical to rapidly identify, develop and validate 

targets and molecules, which exhibit different structural and mechanistic modalities to the current 

front-line therapies. By targeting molecular chaperones as they are critical for the survival of the 

parasite, it may be possible to develop therapies that are less prone to resistance than existing 

front-line therapies, such as ACTs. Therefore, the identification and validation of molecular 

chaperones as a drug target represents an important step in the development of new, effective 

antimalarial drugs.  

2.5 Molecular chaperones as drug targets 

Molecular chaperones are proteins capable of interacting with and stabilizing non-native protein 

structures or newly formed polypeptides, preventing aggregation while promoting the formation 

of correct and functional conformations (Hendrik and Hartl, 1993; Ellis. 2006). These wide range 



 

28 
 

of proteins act as catalysts in the refolding of stress denatured or newly formed polypeptides, 

controlling the assembly of oligomeric proteins, transporting proteins and facilitating protein 

degradation, and protein–protein interactions (Fink, 1999; Edkins and Boshoff, 2014).  

 

Heat shock proteins (Hsp) are members of the molecular chaperone family, functioning to 

overcome the effects brought about by cellular stress including elevated temperatures or a 

crowded environment (Edkins and Blatch, 2012). Eukaryotic Hsps are both constitutively 

expressed under normal conditions andheat inducible (Donnelly et al., 2008). Hsp90 alpha in 

human is the house-keeping gene, therefore, constitutively expressed under normal cellular 

homeostasis. However, Hsp90 beta is the stress protein, expressed following exposure to stress 

ie. It is nicknamed the cancer Hsp90. During cellular stress, the expression of these Hsps is 

upregulated, preventing the aggregation of the stress denatured proteins, assisting them to re-

fold, therefore, maintaining a functional proteome (Derry, 2018). All eukaryotic organisms possess 

the same ability to respond to stress, and this capability entails the use of molecular chaperones 

to help the organism cope with the impacts of stress. Adapting to the differences in host and 

vector conditions is particularly very important for parasitic organisms as it can be very stressful 

for the organism (Kumar et al., 2007). Hsps are classified and named according to their function 

and molecular weight on kilo Daltons (KDa), for instance Hsp90 has a molecular size of 

approximately 90 kDa (Zininga and Shonhai, 2019). The diverse family of Hsps include small heat 

shock proteins (Hsp), Hsp40, Hsp70, Hsp90 and Hsp110 (Lindquist and Craig, 1988; Kim et al., 

2013). Hsps are highly desirable therapeutic targets because of their roles in human disease and 

cellular stress responses, particularly Hsp90 and Hsp70.  

 

Since chaperones are typically overexpressed in cancer cells and have also been linked to 

malignant cell maintenance (Calderwood et al., 2006), Hsp90, for instance, is well established as 

a therapeutic target in cancer and neurological illnesses. It is currently the subject of research in 

the context of parasites. The crucial role played by these proteins in the survival and development 

of these parasites in different physiological habitats (human host and mosquito vector), is what 

makes them such ideal drug targets. In P. falciparum, Hsps are responsible for pathogenesis, 

regulation of infection, thermal protection of the parasite and additionally, the exportation of 

parasite proteins to the red blood cells (Shonhai, 2010). The major types of Hsps found in P. 

falciparum include, PfHsp40, PfHsp70 and PfHsp90. With PfHsp90 and PfHsp70 being the most 

prominently expressed, regulating transcription factors and protein kinases while also helping to 

fold newly synthesized proteins (Gitau et al., 2012; Kim et al., 2013).  



 

29 
 

2.6 Hsp90 

The most widely occurring molecular chaperones in eukaryotic cells is the Hsp90 (Picard, 2002). 

The ATP dependent molecular chaperone of Hsp90 plays critical roles in fundamental processes 

of cell signalling and, cell cycle control. Hsp90s are the central hub of proteostasis (Bieble et al. 

2019), which are conserved across all organisms, displaying >50% sequence similarity between 

E. coli and human Hsp90s (Chen et al., 2006).  

 

Hsp90 isoforms are located in the cytosol (Hsp90α and Hsp90β), the mitochondria (Hsp75/ TRAP-

1), and the endoplasmic reticulum (GRP94/GRP96). There are specialised forms of Hsp90s found 

in different cellular organelle with specialised sets of client proteins. In humans, Hsp90 has also 

been found to localize in plasma membrane or extracellular matrix (Edkins and Blatch, 2012). P. 

falciparum harbours an apicoplast localizing the isoform of Hsp90 (Acharya et al., 2007). The 

cytosolic Hsp90 isoform is responsible for properly folding over 300 protein substrates, termed 

clients, including protein kinases, transcription factors, and receptors critical for maintaining 

protein homeostasis and regulating vital cellular processes (Acharya et al., 2007). All members 

of the Hsp90 family are structurally configured as homodimers comprising of three main domains 

(Figure 3), the N-terminal domain (NTD) which is the primary binding site for ATP/ADP (Figure 3 

in red), responsible for the provision of requisite energy for client folding (Banerji, 2009). A 

charged linker region that connects the NTD with the middle domain (MD) (Figure 3 in green), 

which varies in amino acid sequence across different organisms (Jackson, 2013). The MD 

functions as a discriminator of different client proteins (Hawle at al., 2006). Lastly, the C-terminal 

domain (CTD), which is adjacent to the MD, is the site of dimerization (Figure 3 in blue). Cytosolic 

Hsp90s possess an end C-terminal EEVD motif which acts as a binding site for tetratricopeptide 

repeat (TPR) containing co-chaperones (Prodromou et al., 2000). 

 

Figure 3: Diagrammatic sketch of the full length P. falciparum Hsp90, with the N-terminal (NTD) ATP binding site 
(Amino acid 1-275) shown in Red rectangle, the Middle domain (MD) from amino acid 275 – 444 depicted in  green 
circle and then the C-terminal domain (CTD) form amino acid 444-677 in blue circle. The 4 vertical lines on the NTD 
represents the ATP binding residues  Adapted from Kabakov, 2021. 

 



 

30 
 

2.7 Mechanism of Hsp90 ATPase activity 

The ATP-dependent Hsp90 possess an NTD that serves as the ATP binding site and an ATPase 

cycle as depicted in Figure 4. ATP binding and hydrolysis is critical for the functioning of Hsp90 

(Obermann et al., 1998; Panaretou et al., 1998). ATP binding also influences conformation of 

Hsp90, promoting the rearrangement of the client binding region (Csermely et al., 1993; Grenert 

et al.,1997). A study by Weikl et al. (2000) provided evidence that conformational changes occur 

in the Hsp90 molecule upon binding of ATP. Upon client protein binding, Hsp90 would be trapped 

in the ATP bound state. The ATP molecule will subsequently undergo hydrolysis (Figure 4, 1st 

hydrolysis), releasing ADP and Pi. ADP induces a conformational change of the Hsp90 molecule 

from a widely open V-shaped molecule to the NTD dimerized closed state, while the conversion 

to the compact ADP state will displace the bound client, followed by reopening to reset the 

chaperone (Figure 4) (Elnatan et al., 2017). 

 

Figure 4: Model of the Hsp90 ATPase cycle showing the obligatory sequential hydrolysis and conformational switching. 
Coloured in grey are the protomers. 2 ATP molecules bind to an Hsp90 molecule apo state (top left), which induces a 
dimerized closed state of the NTD. During the closed state, the buckled protomer hydrolysis the 1st ATP molecule 
releasing Pi, which then drives a conformational switch of the straight protomer bound to ATP, to a buckled 
conformation, while the previously buckled protomer (now ADP), straightens. This rearrangement facilitates the 
remodelling of the client. As a result, the now buckled promoter hydrolyses the second ATP. Finally, the ADP/ADP 
dimer re-opens, releasing nucleotides and resetting TRAP1 to the apo state. 

 

 



 

31 
 

In the past decade, Hsp90 has become a major therapeutic target for cancer, there has also been 

increasing interest in it as a therapeutic target in neurodegenerative disorders, and in the 

development of anti-virals and anti-protozoan infections (Jackson, 2012). There has been great 

success with regards to targeting Hsp90 in cancer research, where it has already been studied 

extensively (Kim et al., 2009; Edkins, 2016; Park et al., 2020). Currently there are 18 

compounds/inhibitors of Hsp90 at various phases of cancer clinical trials. AUY922 (also known 

as Luminespib) is a clinically approved Hsp90 inhibitor that has been demonstrated to have 

anticancer efficacy in preclinical and clinical investigations (Jensen et al., 2008). AUY922 is a 

small chemical inhibitor that binds to Hsp90's ATP-binding site, reducing its activity and 

encouraging the degradation of client proteins that rely on Hsp90 for stability (Jensen et al., 2008). 

AUY922 has showed encouraging outcomes in clinical studies for treating several forms of 

cancer, including gastric cancer (Park et al., 2018). Another example is jeselhy® (pimitespib) an 

orally administered small molecule inhibitor that specifically targets the α and β isoforms of Hsp90 

to treat gastrointestinal stromal tumours (GIST), which gained its first approval in June 2022 in 

Japan (Hoy, 2022). Inhibition of Hsp90 leads to antitumor effect by destabilization and reduction 

of proteins involved in cancer growth and survival, including receptor tyrosine kinase (KIT), 

platelet derived growth factor receptor alpha (PDGFRA), human epidermal growth factor receptor 

2 (HER2), and epidermal growth factor receptors (EGFR) (Thaiho Pharma, 2022). 

 

The laudable efforts of Hsp90 targeting in eukaryotic organisms has inspired repurposing of the 

compounds for parasitic disease. However, several lessons have been learned the from currently 

available Hsp90 inhibitors, especially from cancer research. For instance, geldanamycin (GA), 

although highly effective and with selective affinity towards cancerous cells with minimal effect on 

normal cells, has been reported to have hepatotoxic side effects (Neckers et al., 1999). This has 

limited its pharmaceutical application. Hence the molecular structure of geldanamycin has been 

used as a scaffold to develop new smaller derivatives, which are able to retain the potency 

towards Hsp90, without (or with fewer) adverse effects. Some of the derivatives currently in 

anticancer clinical trials include, tenespimycin (17-AAG) in phase 3 (Selleckchem, 2023; 

https://www.selleckchem.com/products/17-AAG(Geldanamycin).html and Alvespimycin (17-

DMAG) HCL in phase 2 (https://www.selleckchem.com/products/17-AAG(Geldanamycin).html ). 

2.8 Small molecule inhibitors of Hsp90 

Previous studies have shown that inhibiting the ATPase activity at the N-terminal domain of the 

Hsp90 is an effective strategy for disrupting the functions and interactions with other client 

proteins (Usmani and Chiosis, 2012). A study by Theoderaki and Caplan (2011), demonstrated 

that the inhibition of Hsp90 with N-terminal inhibitors leads to the degradation of client proteins. 

https://www.selleckchem.com/products/17-AAG(Geldanamycin).html
https://www.selleckchem.com/products/17-AAG(Geldanamycin).html


 

32 
 

To date, a number of synthetic and natural Hsp90 targeting inhibitors have been evaluated 

(Neckers 2003; Drysdale et al., 2006; Taldone et al., 2008), including the most common GA and 

its derivatives such as 17-allylamino-17-demethoxygeldanamycin (17-AAG), 17-dimethylamino-

ethylamino-17-demethoxygeldanamycin (17-DMAG) (Pallavi et al., 2010), which are already in 

phase 3 clinical trials despite the fact that it can only be administered intravenously because of 

solubility concerns (Koca et al., 2016).  

With regards to targeting PfHsp90, these inhibitors have been found to be cytotoxic to human 

cells as they are unable to be selective of PfHsp90 over the human Hsp90 (Shonhai, 2010), 

rendering them ineffective as antimalarials. This calls for other alternative small molecule 

inhibitors with high specificity for PfHsp90 over the human Hsp90. The objective of the current 

study was to generate novel small inhibitors that selectively target PfHsp90 to serve as 

antimalarial agents.     

 

2.9 P. falciparum Hsp90 (PfHsp90) as a drug target 

PfHsp90 is an attractive drug target for the development of novel antimalarial drugs as the P. 

falciparum parasite's ability to survive and proliferate is greatly dependent on the molecular 

chaperone. Molecular chaperones are expressed by the parasite to effectively cope with severe 

environmental stresses, including changes in pH, temperature, and nutrient supply, during its 

cyclic transition from the mosquito vector to the human host (Pavithra et al., 2004; Uwimana et 

al., 2020). The cellular stress induces the abundant expression of PfHsp90 in the parasite 

(Shahinas et al., 2013). Functional interruption of PfHsp90 using the NTD inhibitors from anti-

cancer agents have previously been explored (Kumar et al., 2003; Banumathy et al., 2003; Pallavi 

et al., 2010; Shahinas et al., 2013; Murillo-Solani et al., 2017; Posfai et al., 2018). These studies 

demonstrated the essentiality of PfHsp90 in the survival and proliferation of the P. falciparum and 

have provided evidence to investigate PfHsp90 as a drug target. A study by Banumathy et al. 

(2003) showed that geldanamycin can inhibit PfHsp90 in blood stage parasites, thereby 

preventing growth and transition from the ring to the trophozoite stage of the parasite. PfHsp90 

was validated as a drug target with pre-clinical studies wherein 17-AAG, PUH-71 and harmine 

were able to cure P. berghei infected mice in a pre-clinical trial (Pallavi et al., 2010). Another study 

that aimed to show the effect of geldanamycin on P. falciparum replication and morphology by 

Kumar et al., (2003) gave evidence that treatment with geldanamycin resulted in a more robust 

inhibition to PfHsp90 than its human counterpart, the human Hsp90, where the parasite’s growth 

was strongly inhibited by GA with an IC50 of 200 nM in human erythrocyte culture. 

 



 

33 
 

Harmine is capable of selectively binding PfHsp90 over human Hsp90 (Shihanas et al., 2012), 

providing the evidence that selective inhibitors of PfHsp90 can be developed. The challenge with 

regards to designing selective inhibitors of PfHsp90 over human Hsp90 is the highly conservative 

nature of the molecular chaperones (sequence homology of about 69%) (Corbett and Berger, 

2010). Biochemically, PfHsp90 has been demonstrated to possess a higher ATPase activity than 

its human homolog, making it more susceptible to inhibition (Pallavi et al., 2010; Shahinas et al., 

2013). Overall, the observed difference suggest that the two chaperones have distinct enzymatic 

properties and these, can be exploited biochemically for inhibitor design. As PfHsp90 is a 

validated drug target in malaria, the process of identifying and developing its inhibitors involve hit 

discovery, hit-to-lead optimization, and preclinical testing. 

 

2.10 Drug discovery  

Drug discovery is the process involving a confluence of scientific disciplines, including chemistry, 

biochemistry, biology, pharmacology, toxicology, and computational science to identify and 

develop new therapeutic entities that can effectively treat or cure diseases by targeting specific 

molecular pathways or biological processes (Zhou and Zhong, 2017).  

 

In target-based drug discovery, the drug discovery pipeline typically starts off by the identification 

and validation of a potential therapeutic target (protein or enzyme) that plays key role in the 

development or mechanism of action of the disease (Hughes et al., 2011). Followed by 

identification of a small potent compounds can selectively bind with high affinity. Hit identification 

employ various strategies to screen enormous libraries of chemicals for molecules that interacts 

with the target (Schenone et al., 2013). These compounds are then optimized through a process 

of iterative testing and modification to improve their potency, selectivity resulting in a lead 

candidate. Once a lead molecule has been identified, it must go through a series of in vitro and 

in vivo preclinical testing in animal models to determine its safety and effectiveness. If the results 

are good, the molecule will undergo safety and efficacy testing in increasingly larger groups of 

patients in human clinical trials (Hughes et al., 2011). 

 

There are other different approaches for drug discovery including traditional drug design, which 

is based on screening thousands of natural and synthetic compounds for activity (Reddy and 

Parrill, 1999). This approach includes the oldest known technique, the serendipity method (Ban, 

2022). Examples in the history of pharmacy for drugs discovered by the serendipity method 

include the most popular story about penicillin, a drug that saved millions of lives during the 

Second World War and for which Fleming, Florey and Chain received the Nobel Prize in 1945. 



 

34 
 

Once a potential drug compound was selected by this process of screening, medicinal chemists 

would then synthesize hundreds of related compounds to develop the safest, most effective drug 

for patients use (Reddy and Parrill, 1999). However, the costs and risks associated with this 

process are enormous. This phenotypic screening technique’s shortcoming is that it does not 

indicate why a compound is active or inert, or how it may be improved. It also does not guarantee 

that an active compound is specific for a particular human target protein (Reddy and Parrill, 1999). 

A lack of such specificity can be a major cause of unpleasant side effects, potentially halting a 

drug's clinical development. Other approaches include, chemical modification of established 

drugs or natural compounds (Guo, 2017), screening of databases, virtually or by high throughput 

(HTS) assay to discovering new drugs (Smith, 2002; Shoichet, 2004; Butkiewicz et al., 2017).  

 

Nowadays, the most sophisticated method for drug discovery is rational drug design, which is 

both the smartest and least expensive form of drug discovery (Flower, 2002; Young, 2009; 

Doytchinova, 2022). The drug discovery process is a long and complex one, often taking more 

than a decade and costing billions of dollars. However, successful drug discovery can lead to life-

changing treatments and cures for a wide range of diseases. 

 

2.10.1 Rational drug design  

Rational drug design (RDD) is a process in which drugs are developed by using knowledge of a 

target molecule's structure and function (Batool et al., 2019). The process involves using 

information about the molecular target, such as its 3D structure, to design a molecule that 

interacts specifically with that target to produce a therapeutic effect (Kholodenko, 2000). RDD 

aims to create more effective and specific drugs with fewer side effects, compared to traditional 

phenotypic screening approaches (Mahapatra and Karuppasamy, 2022). The molecular structure 

of the target is used to predict how a drug molecule will interact with it, and the design process 

can involve computer simulations, molecular modelling and synthesizing new compounds to test 

their efficacy (Sharp, 2002). 

 

The success of RDD depends on the availability of detailed information about the target molecule, 

when this information is available, the process can lead to the development of novel and highly 

effective drugs (Wang and Gainza, 2003). In the absence of a crystal structure or experimental 

data for the target protein, homology modelling can provide valuable insights into the protein's 

structure and aid in the design of potential drugs (Vyas et al., 2012). RDD strategies form part of 

the different approaches required for the process of drug discovery, with the objective of 

overcoming the challenges around the traditional drug design process (Mahapatra and 

Karuppasamy, 2022). The process is challenging as it is complex, expensive and time-



 

35 
 

consuming, involving numerous factors that must be taken into consideration (Reddy et al., 1999). 

Biomolecules play crucial roles in the development of a disease by interacting with one another 

through protein-protein or protein nucleic acid interactions, spreading signalling events or 

changes in the metabolic processes, therefore, altering the bioactivities carried out by these 

biomolecules (Baltoumas et al., 2021). Targeting these biomolecules could be advantageous. 

This can be done by either by binding to and inhibiting their activity using small molecules, whose 

competitive binding affinities would be higher than that of their natural ligands that bind on the 

active site on the biomolecule, or by inhibiting the biomolecular interactions with each other, 

preventing crosstalk between the molecules (Fuller et al., 2009). RDD has significantly advanced 

in recent years due to the advent of computational tools and computer aided methodologies such 

as virtual screening of libraries and machine learning approaches (Henry, 2001; Miller 2009). The 

successful application of rational drug design strategies is seen in the development of drugs such 

as saquinavir (Invirase) and ritonavir (Norvir) in HIV treatment, that work by targeting the protease 

enzyme involved in the replication of HIV, inhibiting its activity, and preventing the production of 

new viruses (Santos et al., 2015).  

 

2.10.2 Application of molecular docking in drug discovery 

Molecular docking is a computational technique that simulates the interaction between a ligand 

and a receptor to predict their binding mode and affinity (Meng et al., 2011). Molecular docking 

has become an indispensable tool in early-stage drug discovery for hit identification, lead 

optimization, and virtual screening of large compound libraries (Morris et al., 2009). Predictions 

of the binding modes and binding affinities by docking is attributed to docking algorithms that 

essentially identify favourable orientation of a ligand in the binding site of the protein. Glide being 

one of the most widely used estimate binding affinities (Friesner et al., 2004). Glide estimates 

binding affinities through a combination of molecular docking and scoring methods. It utilizes a 

grid-based algorithm to predict the binding conformation of a ligand within the active site. The 

glide scoring function considers various factors such as steric clashes, hydrogen bonding, 

hydrophobic interactions, and electrostatic interactions to evaluate the binding affinity of the ligand 

(Friesner et al., 2004). The starting point for glide scoring is the empirically based ChemScore 

function of Eldridge et al. (1997) which can be written as: 

 ∆𝐺𝑏𝑖𝑛𝑑 = 𝐶0 + 𝐶𝑙𝑖𝑝𝑜 ∑ 𝑓(𝑟𝐼𝑟) + 𝐶ℎ𝑏𝑜𝑛𝑑 ∑ 𝑔(∆𝑟)ℎ(∆𝑎) + 𝐶𝑚𝑒𝑡𝑎𝑙 ∑ 𝑓(𝑟𝐼𝑚) + 𝐶𝑟𝑜𝑡𝑏𝐻𝑟𝑜𝑡𝑏                       (1) 

 

Chemscore defines the summation in the second term including all ligand-atom/receptor-atom as 

lipophilic, whereas the summation in the third term includes all ligand-receptor hydrogen bonding 

interactions. Functions such as f, g, and h (equation 1) produce a complete score 1.00 for lengths 



 

36 
 

or angles that fall within conventional boundaries and a half score 1.000.00 for distances or angles 

that are beyond those limitations but fall within greater threshold values (Friesner et al., 2004). 

 

While docking is generally successful in reproducing experimentally determined ligand poses, 

accuracy can be affected by factors such as protein flexibility and ligand conformational variability 

(Kitchen et al., 2004). Therefore, to improve docking accuracy computational softwares like the 

Schrödinger’s induced fit docking (Halgren et al., 2004; Friesner et al., 2004; Friesner et al., 2006) 

and molecular dynamics simulations (Amber et al., 2005) have been designed to take into 

consideration the flexibility of the protein and ligand.    

 

Schrödinger's Induced fit docking approach is taken to address the issue of receptor flexibility 

during ligand binding by combining docking techniques. Using glide to account for ligand flexibility 

and the prime refinement module accounting for receptor flexibility where minimization is used to 

sample side-chain degrees of freedom in the receptor while enabling modest backbone 

movements (Sherman et al., 2006). Therefore, the methodology compensates for both modest 

backbone relaxations and substantial side-chain conformational changes in the receptor 

structure. 

 

Docking has been successfully applied in various stages of drug discovery for instance in virtual 

screening, enabling the screening of large compound libraries to identify potential lead 

compounds. Virtual screening methods, such as high throughput docking, enable the prioritization 

of compounds based on their predicted binding affinities (Ferreira et al., 2015). Docking plays a 

crucial role in lead optimization by guiding chemical modifications of compounds to improve 

binding affinity and selectivity. Structure-based design approaches, including fragment-based 

docking and de novo design, aid in the design of potent and specific ligands (Hajduk et al., 2007).  

 

2.10.3 Molecular dynamic simulation 

Molecular dynamic simulation (MDS) is an important application in drug discovery studies 

particularly in ligand docking applications as it enables the study of conformations of a molecule 

or complex that are thermaly accessible. Furthermore, MDS gives insights into the natural 

dynamics of biomolecular interactions in a solution at different timescales (Hansson et al., 2002). 

MDS can be applied in the study of protein ligand complexes, mixed solvents, organic solids and 

synthetic macromolecular complexes to analyse molecular and condensed-phase systems that 

are dynamic in nature. Coupled with other computational tools MDS seeks to address challenges 

faced in drug discovery studies where the dynamic nature of the proteins cannot be ignored, such 

as ligand induced conformational changes in the receptors active site, by capturing dynamic 



 

37 
 

events of scientific interest, making it a fundamental computational tool in drug discovery 

(Schrödinger, 2022).  

 

The principle behind MDS is that given the positions of all atoms in a biomolecular system, the 

force exerted on each atom by all other atoms can be calculated. Thus, making it possible to 

predict each atom's spatial position as a function of time using Newton's equations of motion. To 

be more specific, one calculates the forces on each atom periodically while stepping through time, 

updating each atom's location and velocity using those forces generating a trajectory. In essence, 

the generated trajectory is a three-dimensional movie that depicts the system's atomic-level 

configuration at each point over the simulated time period (Hollingsworth and Dror, 2018). 

2.10.4 Molecular Mechanics/Generalized Born Surface Area (MMGBSA) 

MMGBSA is a computational approach used to estimate binding free energies and analyze 

protein-ligand interactions (Genheden and Ryde, 2015). The overall approach includes energy 

calculations for each snapshot, molecular dynamics simulations to sample the conformational 

space, and statistical analysis to determine the average binding free energy. By breaking down 

the free energy of binding into several energy components such van der Waals interactions, 

electrostatic interactions, solvation energies, and entropy contributions, MMGBSA is able to 

compute the free energy of binding (Genheden and Ryde, 2015). MMGBSA can be applied to a 

wide range of studies including drug binding affinity Prediction, protein-ligand binding free energy 

calculations and protein-protein interaction studies (Wang et al., 2018). 

The strengths of MMGBSA contributing to its value drug discovery and protein-ligand interaction 

studies includes its computation al efficiency, making it less demanding compared to more 

rigorous methods such as free energy perturbation or thermodynamic integration (Virtanen et al., 

2015). This efficiency allows for the analysis of large datasets and enables rapid screening of 

potential ligands, saving valuable time and computational resources. Additionally, MMGBSA has 

demonstrated reasonable accuracy and good correlation with experimental binding affinities 

(Wang et al., 2019; Forouzesh and Mishra, 2021; Taylor and Ho, 2023).  

 

Hybrid approaches that combine MMGBSA with other methods, such as quantum mechanics 

calculations or enhanced sampling techniques, have shown promise in improving accuracy and 

accounting for specific phenomena (Massova and Kollman, 2000; Jerome, 2015). Enhanced 

sampling techniques, such as molecular dynamics or metadynamics, can enhance 

conformational sampling and provide more accurate free energy estimates. Additionally, the 

integration of machine learning approaches has been explored to improve MMGBSA predictions 

by capturing complex relationships between molecular features and binding affinities (Dong et al., 

2021; Ji et al., 2021). 



 

38 
 

2.10.5 Machine learning (ML) in drug discovery  

Machine learning (ML) is playing an increasingly important role in drug discovery. ML algorithms 

analyze large amounts of data to identify patterns and make predictions, which can be used to 

inform drug discovery efforts (Hansen, 2016). For example, ML can be used to predict the activity 

of potential drug compounds, to prioritize compounds for further testing, and to predict the toxicity 

and side effects of drugs (Chen et al., 2018). Drug discovery aims at finding novel molecules with 

chemical characteristics for the treatment of disease. With that objective in mind, recent advances 

in technology have enabled the use of ML tools and techniques in drug discovery to the research 

process while reducing risk and expenditure during clinical trials (Dera et al., 2021). One popular 

approach in drug discovery is to use ML algorithms to analyze high-throughput screening data, 

where the activity of many compounds is measured against a particular target (Chen et al., 2019).  

 

Another approach is to use ML to analyze chemical and biological data to predict drug-target 

interactions and guide the design of new compounds (Xu et al., 2020). The application of ML in 

drug discovery is rapidly advancing, and it is expected to have a significant impact on the speed 

and efficiency of drug discovery efforts in the coming years (Kohavi et al., 2018). These 

techniques are aimed at improving the decision-making process regarding pharmaceutical data 

with applications such as quantitative structure activity relationship (QSAR) generating predictive 

models, de novo drug design and hit discovery, retrieving more accurate outcomes. ML is a 

category that falls under artificial intelligence tools, utilizing numerous algorithms to interpret and 

attain knowledge from data (Chandra and Hareendran, 2021).  

2.10.6 Quantitative structure activity relationship  

Through mathematical systems, QSAR possesses the ability to predict with a high level of 

accuracy, the physicochemical and biological fate properties that a de novo compound will have 

based on the knowledge of its chemical structure and from previous experimental studies by 

numerically linking the chemical structures with their biological activities (Carracedo-Reboredo et 

al., 2021). To theoretically anticipate the biological activity, QSAR models integrate computational 

and statistical methodologies, permitting the development of potential novel molecules without 

having to go through trial-and-error process of organic synthesis (Carracedo-Reboredo et al., 

2021). This method is cheaper and faster as it exists in a virtual environment, relating chemical 

structure to biological activity, without the need for certain resources such as equipment, 

reagents, instruments, and kits (Carracedo-Reboredo et al., 2021). To be able to carry out a 

QSAR study, one needs molecular structures of different compounds with a common mechanism 

of action, biological activity of each ligand in the study and physicochemical properties (Coronin 

and Schultz, 2003).  



 

39 
 

 

For the QSAR models to be able to predict the biological activities of ligands that do not yet exist, 

is by generating equations to calculate descriptors with a high probability of pharmacological 

success (Carracedo-Reboredo et al., 2021). There are different QSAR methods that can be used 

in drug discovery to train models that predict the activity or chemical properties of compounds. 

An overview of some popular QSAR techniques is provided below:  

 

 Molecular descriptors, which are numerical representations of molecular attributes 

including size, shape, and chemical properties, are frequently used in QSAR models. 

Based on chemical structures, these descriptors may be computed and utilized as input 

features in machine learning models (Khan, 2016). 

 Multiple Linear Regression (MLR), is a well-known QSAR technique that uses linear 

equations to connect chemical descriptors with activity data. In order to build a predictive 

model, it makes the assumption that there is a linear relationship between the descriptors 

and the activity (Santiago et al., 2018). 

 An ensemble learning technique called Random Forest (RF) mixes many decision trees 

to produce a prediction model. The final forecast is achieved by averaging the predictions 

made by each individual tree, each of which is trained on a distinct sample of the data 

(Svetnik et al., 2003; Haffar et al., 2023). 

 Neural networks: Convolutional and recurrent neural networks, as well as other deep 

neural networks, can be used for QSAR modeling. These networks may capture intricate 

connections between molecular characteristics and activities because they are made up 

of several layers of linked nodes (neurons) (Alzubaidi et al., 2021). 

 Active learning QSAR is a methodology that involves the selection of informative data 

points for annotation by an oracle (expert) to train a predictive model. With the goal of 

reducing the amount of labeled data required while maintaining or improving model 

performance (Reker and Schneider, 2015; Tu et al., 2023). 

2.11 ATPase activity assay 

ATPase activity assays are biochemical methods used to measure the hydrolysis of adenosine 

triphosphate (ATP) by ATPase enzymes (Bartolommei et al., 2013; Rule et al., 2016). ATPase 

activity assays measures the release of inorganic phosphate (Pi) during ATP hydrolysis by 

ATPase enzymes (Bartolommei et al., 2013). These assays have been widely employed in the 

study of Hsp90, to investigate the role of Hsp90 (Rowlands et al., 2010), also modulation of Hsp90 

by monitoring ATP hydrolysis in the presence or absence of modulators, the impact on Hsp90 

function can be assessed, providing insights into its allosteric regulation and potential therapeutic 



 

40 
 

targeting (Obermann et al., 1998; Dixit and Verkhivker, 2012). Even in High-throughput screening 

campaigns to identify small molecule modulators of Hsp90 ATPase activity (Rowlands et al., 

2004). Over the years these assays have been adapted for automation, allowing the rapid 

evaluation of compound libraries for potential Hsp90 inhibitors and activators.  

 

2.12 Biophysical methods in early drug discovery   

Biophysical methods are an integration of techniques from physics, chemistry and biology, used 

in the study of function, structure, and interactions of biological molecules, forming part of the key 

components in early drug discovery (Holdgate et al., 2019).  The most used biophysical methods 

include, X-ray crystallography, which is used as a starting point for many drug discovery studies 

by determining the 3D structure of molecules by crystallizing the molecule of interest and then 

exposing it to X-ray radiation, allowing the study of specific interactions of a protein target and 

ligand (Smyth and Martin, 2000; Carvalho et al., 2009). Isothermal titration calorimetry (ITC), is a 

method most used in drug design to provide a comprehensive thermodynamic and kinetic profile 

of ligands binding, by measuring interactions between protein and small molecules to determine 

the binding affinity, enthalpy, and entropy (Linkuvienė et al., 2016; Su and Xu, 2018). Nuclear 

magnetic resonance (NMR) spectroscopy is an essential screening method in structure-based 

drug discovery for studying ligand binding to protein targets, offering crucial structural information 

on protein-ligand interactions, which can be used to help optimize weak-binding hits into high-

affinity leads (Pellecchia et al., 2002; Gossert and Jahnke, 2016). Surface plasmon resonance 

Spectroscopy (SPR), SPR is a real-time optical method for studying biomolecular interactions. It 

monitors binding events by measuring changes in the refractive index of a medium near a metal 

surface without the requirement for labelling (Narayan and Carroll, 2017). Protein-protein 

interactions, small molecule binding, and DNA-protein interactions can be studied using SPR 

(Olaru et al., 2015). Circular dichroism (CD) spectroscopy, is extensively used in drug discovery 

for evaluating the stereochemistry of chiral molecules and proteins, by providing information on 

protein secondary structure, folding, and conformational changes (Greenfield, 2006). There are 

many other biophysical techniques which can be employed in early drug discovery, but since our 

study aimed to investigate small molecule binding, hence SPR was employed.  

2.12.1 Surface Plasmon resonance  

SPR is a powerful biophysical method extensively employed in early drug discovery to investigate 

biomolecular interactions. The principle of SPR is based on the interaction between light and free 

electrons at the interface of a metal film and a dielectric medium. The phenomenon occurs when 

the incident light's wavelength matches the resonance condition, resulting in the generation of 



 

41 
 

surface plasmons (Deng et al., 2017; Prabowo et al., 2018). These surface plasmons can be 

detected by monitoring the changes in the refractive index of the medium near the metal surface, 

which is indicative of molecular interactions (Hinman et al., 2018). SPR can be used in the study 

of protein-protein interactions (Douzi, 2017), small molecule interactions (Frostell et al., 2013), 

DNA and RNA Interactions (Licatalosi, and Jankowsky, 2020). SPR can provide valuable 

information about conformational changes in biomolecules upon ligand binding (Paynter and 

Russell, 2002).  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



 

42 
 

 

 

 

 

 

CHAPTER 3  

3. MATERIALS AND METHODS 

3.1. In silico methods 

3.1.1 Computational software programme 

The following tasks were performed using the graphical user interface (GUI) Maestro v12.9 in 

Schrödinger Suite 2021-3. All molecular modelling calculations were performed on various 

modules available in Maestro