The human gut microbiome influences host physiology, but biological insights depend on the taxonomic resolution used to characterize microbial communities. This thesis introduces operational subspecies units (OSUs), a reproducible intermediate resolution between species and strains. First, we developed HuMSub, a cohort-independent catalog of human gut microbial subspecies and the panhashome method for rapid species- and subspecies-level quantification. Applied to colorectal cancer, OSU-level profiling identified disease-associated signals missed at the species level and improved machine-learning prediction. We then developed HuMSub2, substantially expanding genome coverage while improving OSU delineation and abundance estimation. Applied to immune checkpoint inhibitor response, HuMSub2 identified responder-associated subspecies overlooked by species-level analyses and partially improved prediction performance. Together, HuMSub and HuMSub2 demonstrate that biologically relevant microbiome variation often resides below the species level and establish OSU profiling as a scalable, reproducible framework for linking microbial diversity to disease and treatment response.