Robust identification of gene-environment interactions for prognosis using a quantile partial correlation approach.

Area of Science:

• Genetics and Epidemiology
• Biostatistics
• Computational Biology

Background:

• Gene-environment (G-E) interactions are crucial for understanding complex diseases.
• Prognostic G-E interactions, particularly with survival data, are understudied and present unique challenges.
• Existing methods struggle with long-tailed or contaminated survival outcomes.

Purpose of the Study:

• To develop a robust statistical approach for identifying G-E interactions in disease prognosis.
• To address limitations of current methods in handling censored and non-standard survival data.
• To improve the accuracy of prognostic G-E interaction identification.

Main Methods:

• Developed a novel approach using censored quantile partial correlation (CQPCorr).
• Leveraged quantile regression for a strong statistical foundation.
• Incorporated weights for censoring and partial correlation to isolate interactions while controlling for main effects.

Main Results:

• The proposed CQPCorr method demonstrated superior accuracy in simulations compared to existing techniques.
• Analysis of TCGA lung cancer and melanoma data yielded biologically relevant findings.
• Results differed from those obtained using alternative G-E interaction methods.

Conclusions:

• The CQPCorr technique offers a robust and statistically sound method for G-E interaction analysis in prognostic survival data.
• This approach effectively handles censored and potentially non-standard survival outcomes.
• The findings highlight the utility of CQPCorr for uncovering novel prognostic G-E interactions in cancer research.