Peptide-Centric Computational Frameworks for Taxonomic and Functional Interpretation in Metaproteomics

Abstract

Metaproteomics provides direct molecular evidence of the expression of specific proteins in microbial communities by identifying and quantifying peptides from complex biological samples. However, the interpretation of metaproteomic data remains challenging because experimentally observed peptide evidence must be converted into biological conclusions across multiple analytical layers, including reference database construction, protein inference, shared peptide ambiguity, taxonomic assignment, functional annotation, statistical testing, and visualization of hierarchical outputs. This thesis develops peptide-centric computational frameworks for taxonomic and functional interpretation in metaproteomics, with the goal of transforming metaproteomic peptide identifications into taxonomically resolved, functionally interpretable, statistically supported, and visually organized microbiome profiles. First, a metagenomics-informed hamster gut metaproteomics workflow was established for the study of SARS-CoV-2 infection. In this application-driven study, metagenomic assembly, annotation, binning, and database generation provided the reference foundation for downstream metaproteomic analysis and enabled deep profiling of microbial and host-associated proteins in young and old hamsters. This work demonstrates the importance of host- and sample-specific reference construction for metaproteomic studies of complex microbiome systems. Second, MetaUmbra was developed to address genome-level presence inference from metaproteomic peptides. By evaluating unique and shared peptide evidence within a formal statistical framework, MetaUmbra provides genome-level p-values and false discovery rate-controlled support for interpreting peptide lists against user-defined genome panels. This framework addresses the ambiguity introduced by shared peptides and helps distinguish statistically supported genome-level evidence from raw peptide matching. Third, MetaX was developed as an integrated peptide-centric metaproteomics analysis platform. MetaX combines database construction and updating, peptide-level taxonomic and functional annotation, Operational Taxon-Function analysis, data preprocessing, downstream statistical analysis, and visualization tools within a unified workflow. By supporting comparative analysis across peptides, proteins, taxa, functions, and taxon-function units, MetaX provides the central analytical framework for taxonomic and functional interpretation in this thesis. Fourth, MetaTree was developed to extend hierarchical visualization and multi-group comparison of MetaX-derived and other hierarchical outputs. MetaTree focuses on topology-consistent visualization, aligned comparison across samples and groups, comparison matrices, and publication-ready figure export. Together, these studies establish a connected set of peptide-centric computational frameworks that link metaproteomic peptide evidence to reference database construction and peptide identification, genome-level inference, taxon-function analysis, statistical interpretation, and hierarchical visualization. Collectively, this thesis advances metaproteomics from peptide and protein identification toward integrated biological interpretation of complex microbial communities.