A novel multivariate curve resolution based strategy for multi-omic integration of toxicological data

Abstract

Multi-omics has become a current subject of research in computational bioinformatics and chemometrics to trigger a more accurate understanding of biological processes. The Multivariate Curve Resolution Alternating Least Squares (MCR-ALS) approach has been demonstrated to be a powerful chemometric method convenient for the analysis of multisets with multiple components with strongly overlapped contributions. However, MCR-ALS has yet to be tested to analyze multi-omic data fused at the mid-level. Hence, this study seeks to assess the efficacy of MCR-ALS when analyzing mid-level fused data, both independently and in conjunction with a complementary variance decomposition technique (Joint and Individual Variation Explained, JIVE), for the evaluation of multi-omic data within the ecotoxicology field. Particularly, we have evaluated the ecotoxicological effects of tributyltin zebrafish embryos by concatenating epigenomics, transcriptomics, and metabolomics datasets. The developed pipeline demonstrated the importance of the pre-processing and variance decomposition step for improving the MCR-ALS capability to describe the multi-omics data. Our results showed that applying the combination of JIVE and MCR-ALS to the mid-level multi-omic fused data provided a more accurate interpretation of toxicological data. Therefore, integrating these chemometric tools into the analysis of multi-omic data in ecotoxicology could allow a more comprehensive analysis of this complex data, boosting the understanding of the toxicological mechanisms of pollutants.