PheSeq, a Bayesian deep learning model to enhance and interpret the gene-disease association studies

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genotype-phenotype association studies often yield results lacking robustness and clear interpretations.
• Existing methods struggle to effectively integrate diverse phenotype data for enhanced genetic analysis.

Purpose of the Study:

• To introduce PheSeq, a Bayesian deep learning model designed to improve the robustness and interpretability of genotype-phenotype association studies.
• To leverage phenotype descriptions for more accurate gene-disease association identification.

Main Methods:

• Developed and implemented PheSeq, a Bayesian deep learning framework.
• Integrated phenotype descriptions into the model for enhanced data perception.
• Applied the PheSeq model to case studies involving Alzheimer's disease, breast cancer, and lung cancer.

Main Results:

• Identified 1024 priority genes associated with Alzheimer's disease.
• Discovered 818 and 566 priority genes for breast cancer and lung cancer, respectively.
• Achieved moderate positive rates and high recall rates through data fusion, enhancing gene-disease association interpretation.

Conclusions:

• PheSeq significantly enhances the interpretation and robustness of gene-disease association studies.
• The model's ability to integrate phenotype data provides valuable insights for complex diseases.
• PheSeq offers a powerful tool for prioritizing genes in major diseases like Alzheimer's and various cancers.