Multidimensional molecular changes-environment interaction analysis for disease outcomes

TL;DR

This paper introduces a penalization-based method for analyzing interactions between multidimensional molecular data and environmental factors in complex diseases, improving prediction and understanding of disease outcomes.

Contribution

It develops a novel statistical approach to jointly analyze multiple molecular data types and environmental interactions, addressing limitations of previous single-type analyses.

Findings

Outperforms existing methods in simulations

Provides stable and accurate predictions on TCGA cancer data

Identifies biologically meaningful molecular-environment interactions

Abstract

For the outcomes and phenotypes of complex diseases, multiple types of molecular (genetic, genomic, epigenetic, etc.) changes, environmental risk factors, and their interactions have been found to have important contributions. In each of the existing studies, only the interactions between one type of molecular changes and environmental risk factors have been analyzed. In recent biomedical studies, multidimensional profiling, under which data on multiple types of molecular changes is collected on the same subjects, is becoming popular. A myriad of recent studies have shown that collectively analyzing multiple types of molecular changes is not only biologically sensible but also leads to improved estimation and prediction. In this study, we conduct M-E interaction analysis, with M standing for multidimensional molecular changes and E standing for environmental risk factors, which can accommodate multiple types of molecular measurements and sufficiently account for their overlapping information (attributable to regulations) as well as independent information. The proposed approach is based on the penalization technique, has a solid statistical ground, and can be effectively realized. Extensive simulation shows that it outperforms multiple closely relevant alternatives. In the analysis of TCGA (The Cancer Genome Atlas) data on lung adenocarcinoma and cutaneous melanoma, sensible findings with superior stability and prediction are made.