London Calling 2023: Squiggle analysis for metagenomic viability inference

Metagenomic approaches enable unbiased whole microbial community characterizations but cannot differentiate between living and dead microbes, which is crucial for virulent pathogen detection. Traditional methods for identifying living microbes are labor-intensive and time-consuming. This project aims to develop a computer-based framework using nanopore sequencing to predict microorganism viability from raw metagenomic squiggle data. Nanopore sequencing measures ionic current fluctuations in signal traces (“squiggle”) in real-time as single-strand nucleotides pass through membrane-embedded nanopores. Squiggles can detect atomic changes that reveal functionally important genomic characteristics. We hypothesize that DNA from dead microorganisms gets exposed to environmental damage and lacks DNA repair mechanisms, thereby generating squiggle signal that is distinct from DNA in living organisms. We extracted DNA from living and dead bacteria, obtained squiggle data via nanopore sequencing, and trained deep neural networks to explore differences in squiggle data from such AI predictions.