Revolutionizing Cancer Genomics: AI-Driven Interpretation of Nanopore Sequencing Signals

Abstract

Nanopore sequencing technology enables the detailed analysis of biomolecules such as DNA and RNA by detecting variations in ionic current as these molecules traverse nano-scale pores. This method captures nuanced patterns, including sequence context and modification signatures. This research investigates the application of artificial intelligence (AI), particularly deep learning models such as SquiggleNet, NanoDeep, and DeepMod2, in analyzing nanopore sequencing data. The performance metrics of these models are notable, with SquiggleNet achieving an accuracy of 90.8% and an AUC score of 0.817, along with a recall rate of 72.5%. NanoDeep demonstrated an AUC score of up to 0.925 on simulated datasets and an accuracy of 84.9%, while DeepMod2 exhibited varying AUC scores with the highest being 0.903 and recall rates reaching 92.9%. The findings underscore the transformative potential of AI on enhancing clinical diagnosis and customizing medical treatment through rapid and precise biomarker detection. However, despite these promising results, further research and validation are necessary to confirm the efficacy and robustness of these AI models in diverse clinical settings. This research highlights the significant advancements that AI can bring to the field of genomics, especially in cancer diagnostics, but emphasizes the need for continued development and real-world testing to fully realize their potential.