Harnessing artificial intelligence for genomic variant prediction: advances, challenges, and future directions

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Accurate genetic variant interpretation is essential for advancing disease research and developing targeted therapies.
• Traditional methods for variant interpretation face challenges with complex genomic data and the need for personalized medicine approaches.

Purpose of the Study:

• To review the evolution of artificial intelligence (AI) methodologies in genetic variant interpretation.
• To identify critical challenges and propose strategies for improving variant classification accuracy and clinical utility.

Main Methods:

• Review of traditional rule-based systems, statistical models, and contemporary AI approaches including machine learning, deep learning, and protein language models.
• Analysis of current Variant Impact Predictors (VIPs) and their limitations.
• Assessment of challenges such as data heterogeneity, interpretability, and Variants of Uncertain Significance (VUS).

Main Results:

• AI, including deep learning and protein language models, is increasingly integrated into variant interpretation.
• Significant challenges persist, including data heterogeneity, model interpretability, and the prevalence of VUS.
• Need for explainable AI and more inclusive genomic databases to enhance predictive accuracy across diverse populations.

Conclusions:

• AI offers powerful tools for enhancing genetic variant interpretation accuracy.
• Strategies for improved predictor selection, multi-omics data integration, and optimized workflows are proposed.
• Enhanced variant interpretation will accelerate personalized medicine advancements in research and clinical practice.