High-accuracy SNV calling for bacterial isolates using deep learning with AccuSNV.

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Accurate detection of bacterial mutations is crucial for understanding microbial evolution, transmission, and antimicrobial resistance.
• Existing single nucleotide variant (SNV) detection tools struggle with bacterial genome complexity, leading to high false positive rates and requiring manual filtering.

Purpose of the Study:

• To develop a high-precision, automated tool for bacterial SNV calling using deep learning.
• To overcome the limitations of traditional SNV detection methods that process samples individually.

Main Methods:

• Developed AccuSNV, a novel deep learning tool employing a convolutional neural network (CNN).
• AccuSNV integrates alignment information across multiple bacterial samples to identify patterns and enhance precision.
• Evaluated AccuSNV against seven existing SNV callers using simulated and real-world bacterial datasets.

Main Results:

• AccuSNV demonstrated superior performance compared to seven other SNV calling tools in both simulated and real-world datasets.
• The tool achieved high precision and accuracy in identifying single nucleotide variants across various sequencing depths and bacterial species.
• AccuSNV successfully identified known SNVs in curated bacterial datasets.

Conclusions:

• AccuSNV offers a robust and automated solution for bacterial SNV detection, significantly improving accuracy and reducing manual effort.
• The tool's ability to leverage multi-sample information enhances its precision for complex bacterial genomes.
• AccuSNV provides user-friendly downstream analysis modules, making advanced genomic analysis accessible to a wider range of users.