A comparison on predicting functional impact of genomic variants

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-nucleotide polymorphisms (SNPs) can alter gene function, influencing phenotype and disease risk.
• Accurate prediction of SNP functional impact is essential for identifying disease-causing variants within the human genome.
• Numerous computational tools exist for predicting variant effects, but their performance on large-scale genomic data remains unclear.

Purpose of the Study:

• To systematically evaluate the predictive performance of 14 computational methods for assessing the functional impact of genetic variants.
• To compare the effectiveness of different methods across various variant types.
• To guide the selection of appropriate tools for researchers and clinicians.

Main Methods:

• Systematic evaluation of 14 computational methods for predicting SNP functional impact.
• Assessment included methods for specific and multiple variant types.
• Performance was analyzed from several aspects, including AUC values on distinct datasets.

Main Results:

• No single method demonstrated excellent (AUC ≥ 0.9) performance across both evaluated datasets.
• CADD showed excellent performance for multiple variant types.
• REVEL achieved excellent performance specifically for missense variants.

Conclusions:

• Current computational methods show variable performance in predicting SNP functional impact.
• CADD and REVEL are highlighted as strong performers for specific use cases.
• Further development of predictive methods is needed for robust variant interpretation in genomics and clinical settings.