Automating Tobamovirus Contig Identification with Snakemake and Machine Learning

Emerging tobamoviruses pose a growing threat to vegetable and ornamental crops. We re-analyzed publicly available sequencing datasets to uncover hidden tobamoviral sequences. We developed a custom, scalable Snakemake pipeline for contig discovery and combined bioinformatics with machine learning to classify short and fragmented viral contigs. Feature importance analysis highlights that alignment quality cannot be reliably assessed by identity alone, and benefits from incorporating gap-related and length-based metrics. These preliminary results show that machine-learning–assisted classification substantially expands virome coverage and provides a flexible framework for future virus discovery.