Machine Learning-Driven Quantum Sequencing of Natural and Chemically Modified DNA

Abstract

Simultaneous identification of natural and chemically modified DNA nucleotides at molecular resolution remains a pivotal challenge in genomic science. Despite significant advances in current sequencing technologies, the ability to identify subtle changes in natural and chemically modified nucleotides is hindered by structural and configurational complexity. Given the critical role of nucleobase modifications in data storage and personalized medicine, we propose a computational approach using a graphene nanopore coupled with machine learning (ML) to simultaneously recognize both natural and chemically modified nucleotides, exploring a wide range of modifications in the nucleobase, sugar, and phosphate moieties while investigating quantum transport mechanisms to uncover distinct molecular signatures and detailed electronic and orbital insights of the nucleotides. Integrating with the best-fitted model, the graphene nanopore achieves a good classification accuracy of up to 96% for each natural, chemically modified, purine, and pyrimidine nucleotide. Our approach offers a rapid and precise solution for real-time DNA sequencing by decoding natural and chemically modified nucleotides on a single platform.