Comparison of sequence-based vs structure-based pangenome analyses of the streptomyces

The pangenome of a genus is the complete set of gene families found across all species and strains within that genus, including core families shared across all genomes and accessory families with variable distribution. Identifying these families is essential for studying bacterial evolution and functional conservation, yet many computational approaches rely solely on sequence similarity and therefore have difficulty grouping remote homologs that have diverged substantially in sequence. In this thesis, we develop a structure-based framework that uses predicted protein structures to define gene families and examine its ability to detect conserved core families in Streptomyces, a genus that produces many clinically used antibiotics. By comparing structure-based and commonly used sequence-based methods, we find that structural similarity can recover homologous relationships that are not apparent from sequence alone, leading to the identification of 211 additional core gene families with low sequence similarity. These findings show that incorporating structural information can broaden core genome inference by capturing functionally conserved but sequence-diverged families. The workflow presented here is broadly applicable and provides an alternative approach for pangenome analysis in Streptomyces and other bacterial genera.