By controlling gene expression, cells adapt to their circumstances and make the best use of available resources.The abundances of RNA, proteins, lipids, and metabolites can now be measured thanks to recent technological advancements.The states of the various layers in a biological system are reflected in these incredibly complex datasets.Multi-omics is the combination of these various approaches and data to have a better understanding of the condition of the biological system.As mass spectrometry technology continues to become more accessible, multi-omic analyses of the proteome and metabolome are becoming increasingly popular.However, it is still difficult to derive insight from the combination of this data.
In the present study machine learning and model interpretation were combined to find relationships between molecules in various omic levels.The relationships that were found demonstrated protein control (ProC) over metabolites.Citrate control proteins were shown to be novel protein regulators when they were mapped into existing genetic and metabolic networks.
Additionally, five gene functions could be predicted by grouping the magnitudes of ProC across all metabolites; each of these predictions was confirmed experimentally. It was correctly predicted that two uncharacterized genes, YJR120W and YDL157C, would influence mitochondrial translation. Additionally, the SDH9, ISC1, and FMP52 genes’ expected and confirmed functions were confirmed.Users can get results exploration and MIMaL analysis of their own multi-omic data on a website.
This website’s source code can be found at https://github.com/qdickinson/mimal-website.
Reference:
Dickinson Q. et. al. (2022) Multi-omic integration by machine learning (MIMaL). Bioinformatics. 38(21): 4908-4918.