News Release

A speedier pipeline to diagnosing genetic diseases in seriously ill infants

Peer-Reviewed Publication

American Association for the Advancement of Science (AAAS)

Building on previous research, scientists have made improvements to an artificial intelligence pipeline used to diagnose genetic diseases via blood samples obtained from gravely ill infants in a San Diego-based children's hospital. Their upgraded approach - assessed for 101 children with 105 genetic diseases - delivered provisional diagnoses in under 24 hours with minimal human intervention. The authors say that while this automated platform must be adapted for implementation in different hospital systems, such a tool could help clinicians diagnose genetic diseases more quickly and accurately, potentially hastening lifesaving changes to patient care. Genetic diseases are the leading cause of infant mortality in the U.S., particularly among about 15% of infants admitted to neonatal, pediatric, and cardiovascular intensive care units (ICUs). Rapid disease progression demands an equally fast diagnosis to help inform interventions that lessen suffering and mortality, yet routinely employed genome sequencing takes weeks to return results, which is too slow to guide patient management. In search of an urgently needed solution, Michelle Clark and colleagues analyzed electronic health record (EHR) and genomic sequencing data from both fresh and dried blood samples using a sequencing platform that offered a faster and less labor-intensive approach, as samples could be prepared in batches by an automated robot. The platform also included a written pipeline of computational scripts that automatically processed the childrens' EHR data and ranked the likelihood of specific disease-causing variants for each child. The researchers found that diagnoses matched expert interpretation in 95 children with 97 genetic diseases with 97% sensitivity and 99% precision. The platform also correctly diagnosed three of seven seriously ill ICU infants with 100% sensitivity and precision. In each case, the diagnoses affected treatment, Clark et al. report.


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.