Constrained maximum entropy models to select genotype interactions associated with censored failure times

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Identifying genotype interactions is crucial for understanding complex disease risks.
• Existing methods often struggle to incorporate the temporal aspect of disease occurrence.
• Modeling disease progression over time requires advanced statistical approaches.

Purpose of the Study:

• To develop a novel screening method for genotype interactions associated with disease risks over time.
• To extend the maximum entropy conditional probability model for temporal disease occurrence.
• To identify novel genetic associations in type 1 diabetes patients.

Main Methods:

• Utilized a maximum entropy conditional probability model extended for disease occurrence intervals.
• Employed the Expectation-Maximization (EM) algorithm for censored data.
• Implemented stepwise greedy search and minimum description length for constraint selection.
• Validated the method through extensive simulations and a genome-wide association study (GWAS).

Main Results:

• The proposed method effectively categorizes disease outcomes into risk groups using quantile-dependent intervals.
• Performance is influenced by sample size, number of genotypes, and minor allele frequencies.
• Outperformed existing methods like likelihood ratio test, Lasso, and a binary maximum entropy model.
• Identified novel one-genotype and two-genotype interactions associated with neuropathy in type 1 diabetes.

Conclusions:

• The novel screening method accurately identifies genotype interactions related to disease risk over time.
• The approach provides a robust framework for analyzing temporal disease data in genetic studies.
• This method offers improved performance compared to traditional statistical and machine learning techniques.