Linear combination test for gene set analysis of a continuous phenotype

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene set analysis (GSA) methods are crucial for identifying associations between gene sets and phenotypes in gene expression studies.
• Existing GSA methods are primarily designed for binary phenotypes, posing challenges for continuous phenotypes common in clinical settings.
• Dichotomizing continuous phenotypes can lead to inaccurate sample classification and compromise downstream enrichment analysis.

Purpose of the Study:

• To develop novel GSA methods capable of directly analyzing continuous phenotypes without artificial classification.
• To enhance the power and rigor of GSA by incorporating correlation structures within gene sets and pathways.
• To address the limitations of existing GSA approaches for continuous phenotypes in gene expression data.

Main Methods:

• Extended the Linear Combination Test (LCT) from binary to continuous phenotypes.
• Incorporated covariance matrix estimation for continuous phenotypes within the GSA framework.
• Compared the proposed LCT method (and a modification, LCT2) against two existing GSA methods for continuous phenotypes using simulations and a real microarray dataset.

Main Results:

• The proposed LCT methods demonstrated superior performance compared to the two other publicly available GSA methods for continuous phenotypes.
• Simulation studies and analysis of a real microarray dataset supported the efficacy of the LCT approach.
• The findings highlight the advantages of LCT in handling continuous phenotypes directly.

Conclusions:

• The LCT methods offer a robust and efficient approach for gene set analysis with continuous phenotypes.
• The study provides valuable insights into improving GSA for complex biological data.
• Free R-codes for LCT are available, facilitating broader application in research.