Repository navigation
The increasing application of multi-omics technologies has heightened the demand for accurate cross-omics association analysis. However, current analyses strategies often depend on correlation metrics, annotation databases, or exhibit suboptimal performance when handling sparse and noisy data. To address this, we present OmicsBridge, a cross-omics association analysis algorithm that integrates graph regularization with latent factor decomposition to jointly model count data (amplicon sequencing and metagenomic data) and mass spectrometry data. OmicsBridge operates independently of annotation resources, remains stable with sparse high-noise data, and captures both global patterns and specific microbe–metabolite associations. Validation on public datasets and our own multi-omics data confirmed its robustness and interpretability. In a rhizosphere study of soil-borne crop diseases, OmicsBridge identified key microbe Rubrivivax and its interacting shikimic acid, which enhanced the mechanistic interpretability of the study. Moreover, these interactions were subsequently experimentally validated in situ. To extend usability, over 200 functions, including OmicsBridge, have been encapsulated within the R package EasyMultiOmics for data mining and visualization, and are publicly available as open-source resources.