Integration Analysis of Three Omics Data Using Penalized Regression Methods: An Application to Bladder Cancer

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Investigating complex diseases requires integrating diverse omics data (genomic variants, DNA methylation, gene expression).
• Challenges in omics integration include data heterogeneity, high dimensionality, multicollinearity, and complex result interpretation.
• Innovative statistical methods are crucial for overcoming these integration hurdles and advancing disease mechanism discovery.

Purpose of the Study:

• To propose and validate a novel permutation-based method (MaxT algorithm) for assessing significance and multiple testing correction in omics data integration.
• To explore associations between single nucleotide polymorphisms (SNPs), DNA methylation (CpGs), and gene expression in bladder tumor samples using penalized regression (LASSO, ENET).
• To develop a flexible and computationally efficient approach for multi-omics data analysis.

Main Methods:

• A three-step analysis workflow was employed: (1) SNP/CpG selection within a 1Mb window around gene probes, (2) application of LASSO and Elastic Net (ENET) regression models (SNP, CpG, and Global) to assess associations, and (3) significance testing using the permutation-based MaxT method.
• The proposed method was applied to bladder tumor samples measuring common genetic variants, DNA methylation, and gene expression.
• A simulation study and replication analysis using The Cancer Genome Atlas (TCGA) dataset were performed to validate the method's performance and findings.

Main Results:

• The integrated analysis identified 48 genes significantly associated with both SNPs and CpGs.
• Of these, 36 genes (75%) were successfully replicated in an independent dataset (TCGA), demonstrating robustness.
• The proposed method demonstrated reduced computational time and flexibility for analyzing multiple omics data types.

Conclusions:

• The developed permutation-based MaxT method combined with penalized regression offers an effective strategy for omics data integration in complex disease research.
• The findings highlight the importance of integrating genetic variants and DNA methylation data to understand gene expression regulation in bladder cancer.
• This approach facilitates the discovery of novel insights into complex genetic mechanisms underlying diseases.