Researcher(s)
- Nolan Vasil, Computer Science, University of Delaware
Faculty Mentor(s)
- Shawn Polson, Computer Science, University of Delaware
Abstract
Understanding viral impacts on microbial communities requires accurate identification and functional characterization of viral populations from genomic data. One effective strategy leverages replication proteins encoded in viral genomes, which play central roles in phage biology and strongly influence infection phenotypes such as replication speed, burst size, and infection strategy (virulent versus temperate). These phenotypic traits, in turn, shape microbial host dynamics and nutrient cycling across ecosystems. A combination of in vitro (e.g., enzyme biochemistry assays), in vivo (e.g., phage mutagenesis and infection assays), and in silico (e.g., bioinformatics and phylogenetics) approaches has demonstrated that replication proteins—including Family A DNA polymerase, ribonucleotide reductase, and helicase—and replication modules (co-occurring protein sets within viral genomes) provide a powerful basis for predicting infection phenotypes from sequence data. However, applying this framework at scale currently requires complex, manually integrated bioinformatic workflows, which limits reproducibility, hinders standardization across studies, and restricts broader adoption by the research community.
To address these challenges, this project develops a standardized, automated analysis pipeline implemented in Nextflow, a workflow management system designed for reproducible and scalable execution across local workstations, high-performance computing (HPC) clusters, and cloud computing environments. Using containerized Docker and Singularity environments, the modular workflow integrates Bash and Python scripts to automate identification of viral replication proteins, annotation of genomic feature (genofeature) metadata (including biochemical function, active sites, and conserved motifs), generation of protein embeddings to support clustering, detection of replication modules on contigs representing viral populations, and production of statistical summaries and visualizations for downstream interpretation.
By reducing manual intervention and enforcing standardized analysis steps, this automated framework improves efficiency, scalability, and reproducibility in viral metagenomic analyses. More broadly, it lowers technical barriers to applying replication protein-based approaches, enabling more consistent phenotype prediction and advancing our ability to link viral genomic content to ecological function across environments.