Sparse Partial Least Squares Methods for Joint Modular Pattern Discovery

Area of Science:

• Genomics
• Pharmacogenomics
• Computational Biology

Background:

• The relationship between genomic factors and cancer drug response is not fully understood.
• Patient responses to anticancer treatments vary due to genetic and somatic alterations.
• Identifying complex genomic factor-drug response relationships in pharmacogenomics data is challenging.

Purpose of the Study:

• To develop a framework for identifying multiple-to-multiple relationships between genomic factors and drug response.
• To uncover joint modular patterns in large-scale gene expression and drug response data.
• To reveal coordinated gene-drug associations.

Main Methods:

• Introduction of a sparse partial least squares (SPLS) framework.
• Application of network-regularized penalty within the SPLS framework.
• Analysis of large-scale pairwise gene-expression and drug-response data.

Main Results:

• Identification of joint modular patterns linking genomic factors and drug response.
• Discovery of coordinated gene-drug associations.
• Demonstration of SPLS framework's capability in analyzing complex biological data.

Conclusions:

• The developed SPLS-based methods are powerful tools for uncovering associations between different feature types.
• SPLS methods can be applied to various biological problems, including expression quantitative trait loci (eQTL) analysis.
• This framework advances the understanding of genomic influences on cancer drug response.