Channel modelling for DNA-based data storage

Abstract

DNA-based storage offers a high-density data storage solution, but the synthesis and sequencing processes cause high error rates. Accurate models of the errors and error distributions are required due to the high synthesis cost. Currently, available channel models fail to represent all aspects of the errors introduced during the DNA data storage process. This study proposes new models capable of representing DNA data storage channel error bursts. A Markov model is used alongside substring characteristics derived from a MinION dataset. The methods for creating the proposed models are described. The proposed models are compared to currently available models. We show that the proposed models represent string-specific characteristics, error runs, bursts, and error-free runs better than similar existing models.