Probability-based collaborative filtering model for predicting gene-disease associations.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate prediction of pathogenic human genes remains a significant challenge in current research.
• Leveraging extensive gene-disease data and biological experiments can enhance prediction accuracy.
• Computational methods offer a time- and cost-efficient alternative for gene-disease association studies.

Purpose of the Study:

• To propose a novel computational model for predicting pathogenic human genes.
• To enhance the accuracy and efficiency of gene-disease association predictions.
• To provide a tool for identifying new disease-related genes.

Main Methods:

• Development of a probability-based collaborative filtering model (PCFM).
• Integration of human and nonhuman species data, including latent factorization, heterogeneous regularization (average and personal), vector space similarity, and Pearson correlation coefficient.
• Comparison with four state-of-the-art prediction approaches.

Main Results:

• The proposed PCFM demonstrated superior performance compared to four existing state-of-the-art methods.
• PCFM achieved higher accuracy in predicting pathogenic human genes.
• The model effectively integrates diverse datasets for improved predictive power.

Conclusions:

• The PCFM is a valuable tool for predicting disease-associated genes.
• The model is particularly useful for identifying relationships involving novel human genes or diseases with limited existing data.
• PCFM offers a promising computational approach for advancing genetic disease research.