FuncAnnoClust Web Application for Analyzing Prokaryotic Genome Annotations Using Multivariate Statistics and Machine Learning

Abstract

Comparative functional analysis of prokaryotic genomes is a key area of bioinformatics, but the practical implementation of such research often faces technical barriers. For example, automatic genome annotation using the widely used Rapid Annotations using Subsystems Technology platform is complicated by the fact that standard downloads do not contain information about the system and categorical levels of the subsystems database hierarchy. Filling this gap requires either specialized scripting or labor-intensive manual data processing, limiting the pool of specialists capable of applying these approaches. This paper presents an interactive web platform that automates the comparative analysis of prokaryotic functional profiles based on annotations of the specified system. The developed solution reconstructs the full hierarchical classification structure and integrates three functional modules: classification of genomic functions according to the database hierarchy, calculation of aggregated function values within selected categories and systems, and statistical analysis using dimensionality reduction, clustering, and group significance assessment methods. High computational speed is achieved through the implementation of optimized table structures. The platform was tested on twenty genome assemblies of bacteria of the class Saccharimonadia, reconstructed from metagenomic data. Clustering across the full set of systems demonstrated consistency with the results of phylogenetic analysis, calculation of average nucleotide identity, and profiling of orthologs from the Kyoto Encyclopedia of Genes and Genomes (KEGG). Analysis of specialized functional categories revealed environmentally driven adaptation strategies in bacteria. The platform eliminates technical barriers to working with automated annotation results, ensuring transparency and reproducibility of comparative genomics for researchers without programming skills. The source code is published in an open repository at https://github.com/DariaGI/FuncAnnoClust.