In the realm of diagnosing rare Mendelian disorders, where the task is often laborious and time-consuming, researchers at Baylor College of Medicine are pioneering a more efficient approach through the application of artificial intelligence (AI). Their breakthrough machine learning system, dubbed AI-MARRVEL (AIM), is poised to streamline the identification of potentially causative variants for Mendelian disorders, as revealed in a groundbreaking study published in NEJM AI.
Addressing the pressing need for enhanced diagnostic speed and accuracy, Dr. Pengfei Liu, co-corresponding author and associate professor of molecular and human genetics at Baylor Genetics, emphasizes, "The diagnostic rate for rare genetic disorders is only about 30%, and on average, it is six years from the time of symptom onset to diagnosis."
AIM leverages the extensive Model organism Aggregated Resources for Rare Variant ExpLoration (MARRVEL) database, curated by Baylor researchers, housing over 3.5 million variants from thousands of diagnosed cases. By analyzing patients' exome sequence data and symptoms, AIM generates a prioritized list of potential gene candidates underlying the rare disease.
Comparative analyses against existing algorithms underscore AIM's superiority in real-world scenarios. Testing against data cohorts from Baylor Genetics, the Undiagnosed Diseases Network (UDN), and the Deciphering Developmental Disorders (DDD) project, AIM consistently outperformed other methods, doubling the rate of accurate diagnosis.
Dr. Zhandong Liu, co-corresponding author and associate professor of pediatrics—neurology at Baylor, notes AIM's transformative potential in addressing long-standing unsolved cases. By efficiently identifying high-confidence solvable cases for manual review, AIM offers new hope in resolving previously elusive diagnoses.
Moreover, AIM demonstrates promise in uncovering novel disease genes. Successfully predicting two newly reported disease genes in UDN cases, AIM represents a significant advancement in diagnostic capabilities for rare diseases.
Dr. Hugo Bellen, co-corresponding author and Distinguished Service Professor in molecular and human genetics at Baylor, underscores AIM's pivotal role in guiding the discovery of previously unknown disorders.
Supported by real-world training data and Baylor Genetics' expertise, AIM stands at the forefront of diagnostic intelligence, heralding a new era in clinical practice.
More: https://medicalxpress.com/news/2024-04-ai-diagnosis-rare-genetic-disorders.html
