NanoDeep

NanoDeep applies deep learning to perform real-time adaptive sampling of nanopore sequencing data to enrich microbial DNA and distinguish microbial from human sequences.

Key Features:

• Deep Learning Architecture: Uses a convolutional neural network (CNN) with squeeze-and-excitation modules to analyze raw squiggle data from native DNA sequences and rapidly classify microbial versus human genomic material.
• Adaptive Sampling Capabilities: Implements adaptive sampling by reversing the voltage across individual nanopores to selectively enrich or deplete specific DNA molecules within a sequencing library.
• Efficiency and Fidelity: Improves sequencing efficiency and fidelity for bacterial genomes relative to standard nanopore settings, enhancing data quality for complex microbial communities.

Scientific Applications:

• Microbial Sequencing: Classifying bacterial reads in mixed human and microbial libraries and enriching bacterial content for microbiome analysis.
• Metagenomics: Enriching metagenomic sequences from gut samples to aid detection of unknown microbiota within complex biological samples.

Methodology:

Processes raw squiggle signals from nanopore sequencers and applies a CNN with squeeze-and-excitation to identify and classify genomic origin in real time, enabling dynamic adaptive-sampling adjustments including reversal of nanopore voltage to enrich or deplete target molecules.