Systems, media, methods, and kits disclosed herein can be used to analyze human microbiota for the detection of a condition (e.g., a disease or condition). Further, the systems, media, methods, and kits disclosed herein can utilize machine learning algorithms to analyze samples with high accuracy. In an aspect, a classifier capable of distinguishing a population of subjects based on microbiome composition may comprise: a plurality of microbiome-associated features associated with two or more classes of subjects inputted into a machine learning model, wherein the features comprise the microbiome species and abundance of microbiome elements, wherein the features are derived from a taxonomic community composition analysis of a cell-free nucleic acid sample in a population of subjects; wherein the features contribute to a classifier sensitivity of greater than 50% and a classifier specificity of greater than 85% to distinguish the population of subjects into two or more classes.