Title: Multi-omics integrated analysis by means of graphical models
Authors: Ilaria Bussoli - University of Padova (Italy) [presenting]
Abstract: Thanks to the advances in technology and bioinformatics of the last decade, a large amount of biological data coming from various experiments in metabolomics, genomics and proteomics is available. However, as it is the case with omics disciplines, the complex information content of such experiments introduces a challenge of its own, hence forming biologically relevant conclusions requires specialized forms of data analysis. One of the many problems in this area is how to integrate or model the information provided by differential gene expression and differential gene co-expression under different phenotypic subsets with the one delivered by transcriptome variations and DNA variant. To solve this, these biological records and the topology of the affected biological pathways are incorporated and modelled through conditional Gaussian regression models and chain graph models for mixed variables (both continuous and discrete). Topological properties of collapsibility and decomposability of the graph underlying each biological pathway are assessed to reduce dimensionality. On the obtained sub-regressions, initial testing on relevance of DNA variants (described as binary variables) on differential co-expression of genes is conducted. A direct visualisation of the direction and amplification of biological signals is implemented on the interested biological pathways.