Advancing reproducibility in multi-omics studies

Abstract

Multi-omics studies provide opportunities to investigate complex biological systems, but their interpretation can be complicated by numerous sources of variation. Here, we discuss the biological, analytical, and computational factors that can affect the reproducibility of omics studies. We introduce practical recommendations to improve the reproducibility of computational workflows for omics data analysis and apply these to a workflow analyzing the NTR multi-omics data set. Next, we analyze a healthy adult cohort to characterize major sources of molecular variation in the healthy population and identify associations with health-related phenotypes, including biological age. In a second study, multi-omics profiling of neuronally differentiated patient-derived iPSC lines revealed major sources of variation related to metabolic and mitochondrial activity and neuronal differentiation status. Accounting for these factors enabled the identification of disease-associated molecular features and the rediscovery of a known NANS-CDG metabolic biomarker. Finally, we discuss broader challenges and future perspectives in multi-omics research and provide practical recommendations, based on the studies in this thesis, for obtaining robust and biologically meaningful signals from multi-omics data sets.