From conservation scores to language models: a feature engineering perspective on machine learning for genetic variant interpretation.

Advances in machine learning have transformed genetic variant effect prediction from hand-crafted features to foundation models that learn directly from biological sequences.

• Why it matters: Accurate interpretation of genetic variants is crucial for precision medicine, yet current methods face challenges like data bias and limited reliability, hindering clinical application.
• What they did: The review traces the evolution of methods, from traditional classifiers to deep-learning architectures and foundation models, classifying them by variant type and feature categories, and discusses their biological basis and limitations.
• The result: Understanding these developments and challenges enables improved, more reliable variant interpretation tools, advancing their integration into clinical decision-making.