Year: 2026
Type: article
Abstract: Genetic testing may help diagnose diseases by finding genetic variants, with variant calling and variant interpretation as its main steps. Variant calling compares sequencing data with a reference genome and identifies possible variants. Variant interpretation then assesses whether these variants may cause disease by using different types of evidence, such as allele frequency, phenotype matching, review status, and familial evidence. However, sequencing noise, alignment problems, and complex disease mechanisms can still affect both steps and reduce the accuracy of genetic testing. Artificial Intelligence (AI) tools can find patterns across these processes, but they currently work mainly as decision-support tools rather than independent evaluators. They also face three main problems: poor performance across different datasets, unexplainable outputs, and frequent errors in training data. New multimodal models and foundation models may help combine more types of information. Still, researchers need to confirm their safety and performance in clinical settings. This review examines the value of AI tools in genetic testing and discusses the main barriers to their independent use. It also explains why human expertise remains essential for final clinical diagnosis.
Author: Ziyi Peng
Language: English
Topic: Genomics and Rare Diseases
Subfield: Genetics
Domain: Life Sciences
Open Access status: hybrid
OpenAlex ID: W7225142249