Beyond single markers: bacterial synergies identified by Multidimensional Feature Selection reveal conserved microbiome disease signatures

Abstract

The gut microbiome encodes disease-relevant information not only in the abundance of individual taxa and functions, but in the way they co-occur and interact. Yet metagenomic analyses have largely relied on univariate approaches that evaluate features in isolation, systematically overlooking the combinatorial signals that arise from meaningful, disease-informative patterns of microbial co-occurrence. Here, we introduce a framework based on the Multidimensional Feature Selection (MDFS) algorithm to identify synergistic feature pairs - combinations of taxa and functions in which the information carried by a feature significantly increases in the presence of a partner. This includes features which carry no detectable individual information and may therefore be overlooked by univariate analyses. We first validated the approach on a meta-analysis of colorectal cancer (CRC) cohorts - one of the most competitive microbiome classification benchmarks available - using a leave-one-cohort-out cross-validation framework. Classifiers built solely on our synergy-derived features achieved performance comparable to the state-of-the-art benchmark (AUC = 0.84 vs 0.85). This confirmed that we were able to capture an informative disease signal while revealing microbial interactions not provided by univariate methods. A subset of the pairs was repeatedly selected by the models during feature selection, which indicated their stability. Post-selection effect size analysis revealed their consistent differences between cases and controls. To learn more about diseases beyond CRC, we extended our framework to 13 additional datasets spanning inflammatory bowel disease, cardiometabolic disease, liver cirrhosis and atherosclerotic cardiovascular disease. We identified 80,790 synergistic pairs in eight of them as well as a set of candidate markers recurring across independent cohorts. Overall, synergistic pairs carried substantially more information about disease status than their individual constituents. Our framework therefore established meaningful microbial co-occurrence patterns which would not be possible with univariate methods alone. The framework is freely available at https://github.com/Kizielins/MDFS_synergies. Importance Most microbiome studies search for individual gut bacterial species associated with disease. However, bacteria do not act in isolation, and their combined presence or relative balance may be far more informative than any single microbe considered alone. This study presents a computational framework which identifies pairs of gut microorganisms whose co-occurrence or relative abundance carries significantly more information about disease status than their importance alone. Applied to stool metagenomic data from patients with colorectal cancer, as well as individuals with other conditions, we demonstrate that these synergistic interactions are: 1) widespread; 2) reproducible across independent patient cohorts; 3) and resolve disease-relevant microbial relationships which single-feature analyses do not report. Our framework offers a more complete view of how the gut microbiome is altered in disease and provides a principled basis for identifying robust, interaction-based biomarkers.