Baylor's AI breakthrough slashes rare disease diagnosis times

Scientists from the Baylor College of Medicine in Houston have developed an innovative AI machine learning system named AI-MARRVEL (AIM). This system is tailored to expedite and simplify diagnosing diseases that originate from a single-gene mutation.

Diagnosing rare Mendelian disorders demands considerable effort, even from seasoned geneticists. The Baylor College of Medicine team is leveraging artificial intelligence to streamline this task. The AI-MARRVEL (AIM) machine learning system aids in pinpointing the most probable disorder variants.

Dr. Pengfei Liu, the Clinical Deputy Director at Baylor Genetics, has emphasized the low recognition rate of rare genetic diseases, which is about 30 percent. He has also highlighted the lengthy average of six years it takes to reach a diagnosis following the emergence of symptoms. He underscores the critical need for innovative approaches that enhance the speed and accuracy of diagnosis.

The AIM system is trained on a public database of known genetic variants and analyses known as Model Organism Aggregated Resources for Rare Variant Exploration (MARVEL), also developed by the Baylor team. This database encompasses over 3.5 million variants across thousands of diagnosed cases.

The system generates a hierarchy of the likeliest genetic causes for a rare disease by inputting data on gene sequences and patient symptoms into AIM. The performance of AIM has been benchmarked against other algorithms, utilizing three different data sets with confirmed diagnoses from Baylor Genetics, the Undiagnosed Diseases Network funded by the National Institutes of Health (UDN), and the Deciphering Developmental Disorders (DDD) project. AIM consistently outperformed all other methods, ranking the diagnosed genes as the top candidate in twice as many cases based on these real-world datasets.

Dr. Zhandong Liu, a contributing author of the Baylor study, mentioned that they tailored AIM to emulate human decision-making processes, allowing it to operate quicker, more efficiently, and more cost-effectively. This approach effectively improved the diagnostic accuracy rate twofold.

Furthermore, AIM brings renewed hope for resolving rare diseases that have remained unsolved for years by facilitating reanalysis with novel insights. "We can make the reanalysis process much more efficient by using AIM to identify a high-confidence set of potentially solvable cases and pushing those cases for manual review," Zhandong Liu stated. He anticipates that this tool could reveal many cases previously deemed undiagnosable.