Integrative Co-Expressed Gene Network Analysis for Bladder Cancer Subtyping Using Multi-Omics Data

Traditional stratification methods for muscle-invasive bladder cancer (MIBC)
often rely on a single omics approach to derive subtypes, such as gene expression
or mutation profiles. This project introduces a novel co-expressed network method
that allows the integration of multiple data types at different network levels, aiming
to improve MIBC stratification.

Cluster analysis is performed on the gene expression of TCGA’s MIBC cohort
to establish the starting point in this project. Two of the five identified groups are
Basal groups with heterogeneous Interferon-γ(IFN-γ) responses. This is consistent
with the in vitro research at JBU, which is used to divide the Basal into three
subgroups further. The patients with the lowest IFN-γ response have the poorest
survival prognosis, showing the potential to identify new groups.

The integrative network approach prioritises transcription factors (TF) through
selective edge pruning and integrates the mutation burden via edge weights. A
subset of 98 TF is found to play an important role in tissue differentiation, and
their expression is used to stratify the MIBC. This revealed a Basal group of 20
samples with one of the lowest survival rates found in the literature. The group
exhibits squamous markers, has a low immune response, and has statistically signif-
icant expressed genes involved in tumour aggressiveness, or that can be treatment
targets.

The network approach is refined and used to build a healthy co-expressed
network, which is then used to stratify MIBC. The method identified communities
that can be traced to splitting the healthy dataset into new groups, revealing new
insights into bladder biology. Mutation integration revealed strongly and highly
correlated genes with various biological roles, such as cell proliferation or chromatin
remodelling.

The integrative co-expressed network approach identified distinct gene subsets
and new MIBC groups, underscoring the power of networks as a bioinformatics
tool for understanding bladder biology.