Gaussian process regression for survival time prediction with genome-wide gene expression

Area of Science:

• Genomics
• Biostatistics
• Cancer Research

Background:

• Predicting cancer patient survival from genome-wide gene expression is challenging due to weak and abundant gene effects.
• Existing methods often struggle with accuracy in identifying relevant gene expression patterns.

Purpose of the Study:

• To develop a novel Gaussian process accelerated failure time model for predicting cancer survival.
• To accurately predict survival time using genome-wide or pathway-wide gene expression data.
• To handle various forms of censored survival data and integrate multi-omics data.

Main Methods:

• Proposed a Gaussian process accelerated failure time model.
• Utilized a Monte Carlo expectation-maximization algorithm for joint imputation of censored log-survival time and parameter estimation.
• Applied the method to genome-wide gene expression data.

Main Results:

• Demonstrated the performance and advantages of the proposed method through simulations and real data analysis.
• Successfully analyzed kidney renal clear cell carcinoma data from 513 patients.
• The method effectively predicted survival time and accommodated censored outcomes.

Conclusions:

• The Gaussian process accelerated failure time model offers an effective alternative for cancer survival prediction using gene expression.
• The method shows advantages over existing approaches, particularly in handling complex genomic data and censored outcomes.
• The developed R package provides a practical tool for researchers in cancer genomics and biostatistics.