FunSAV: predicting the functional effect of single amino acid variants using a two-stage random forest model

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single amino acid variants (SAVs) are common genetic variations linked to human diseases.
• Understanding the functional impact of SAVs is crucial for elucidating disease mechanisms.

Purpose of the Study:

• To develop a robust computational model for predicting the functional effects of SAVs.
• To improve the understanding of disease-associated genetic variations.

Main Methods:

• Constructed a high-quality dataset of 2,048 SAVs from 679 protein structures, categorized as disease-associated or neutral.
• Developed a two-stage random forest (RF) model, FunSAV, integrating sequence, structure, and residue-contact network features.
• Implemented a two-step feature selection process to identify key predictive features.

Main Results:

• FunSAV demonstrated strong predictive performance with an Area Under the Curve (AUC) of 0.882 in cross-validation.
• The model's performance is competitive with and surpasses several existing SAV prediction tools.
• Identified novel features contributing to the prediction of SAV disease association.

Conclusions:

• FunSAV offers an effective approach for predicting the functional impact of SAVs.
• The model aids in understanding the pathogenicity of genetic variations.
• The developed tool and datasets are publicly available for research use.