BIOSTATISTICS MEETS OMICS IN PRECISION BIOLOGY

Abstract

Multi-omics data can be combined with biostatistical approaches, and have been an important component of precision biology since they enable the discovery of biomarkers and by extension deciphering of mechanisms of action of diseases.  We applied an advanced multi-layered omics characterisation technique encompassing genomes, transcriptomics, proteomics and metabolomics of groups of individuals that are clinically characterised in the context of this research.  To ensure data were of high utility and could be compared, they passed through rigorous preprocessing, statistical normalisation and dimensionality reduction.  The Differential expression of the phenotypic groupings identified relevant genes, proteins, and metabolites. Quite a number of them played a role in the regulation of the immune system, alteration of metabolism and cell to cell signalling.  Canonical correlation analysis integrative modelling and machine learning identified that the various forms of omics data were correlated with molecular signatures that can be exploited in prediction. During cross-validation, the accuracy rate was above 90%, with which the models were to classify the data.  The pathways and hub molecules identified based on network-based studies are relevant to the disease and regulation processes.  Techniques of ensemble learning performed better than single-algorithm models, which indicates the applicability of integrative analytics.  We employed visualisation as a way of demonstrating that results were robust and with biological implications. These were the examination of the multi-omics correlations, prediction model accuracy, and pathway enrichment.  These findings demonstrate that integrating biostatistics and multi-omics establish an end-to-end-reliable clinically applicable framework performing precision biology. It has a direct impact on the identification of biomarkers, the development of personalised medicine and the development of targeted therapies.