An artificial intelligence system has been shown to identify the causal genes in rare genetic pediatric diseases in 92% of cases.
Rapid diagnosis of diseases allows doctors to start providing the correct treatment at a time when it is most likely to be effective, resulting in shorter hospital stays, fewer complications, and improved patient outcomes; however, obtaining an accurate diagnosis, especially in critically ill children suffering from rare diseases, can be a difficult and time-consuming process.
The clinical interpretation of genetic variants in the context of a patient’s phenotype is the costliest component of the genome-based diagnosis of rare genetic diseases in terms of cost and time. It may now only take hours to sequence the entire genome, but the subsequent computational and manual analyses can take days or weeks before an illness is diagnosed. In some cases, by the time a disease is diagnosed it is too late for treatment to be effective.
Artificial intelligence systems have the potential to greatly reduce time to diagnosis. These systems are not intended to replace clinicians, instead, they can provide valuable insights into the likely cause of illness to support clinical decisions and can potentially shorten the time to diagnosis from weeks to 1 or 2 days.
One such AI-based system, Fabric GEM, was evaluated in a study conducted at Rady Children’s Hospital in San Diego by researchers at University of Utah Health and Fabric Genomics. The system integrates predictive models with the rapidly growing knowledge of genetic diseases and was shown to diagnose rare diseases with a high degree of accuracy.
The system was benchmarked using a retrospective cohort of 119 probands, mostly newborn intensive care unit infants who had been diagnosed with rare genetic diseases. Those individuals had undergone whole-genome or whole-exome sequencing (WGS, WES).
The researchers replicated their analyses in a separate cohort of 60 cases collected from five academic medical centers, and also analyzed the cases with current state-of-the-art variant prioritization tools.
“Variants underpinning diagnoses spanned diverse modes of inheritance and types, including structural variants (SVs). Patient phenotypes were extracted from clinical notes by two means: manually and using an automated clinical natural language processing (CNLP) tool,” explained the researchers. They also reanalyzed 14 previously unsolved cases.
Currently used analysis tools are effective at identifying small genomic variants such as a single letter DNA change, but it has been estimated that around 15%-20% of genetic diseases involve much more extensive changes – termed structural variants – which are more difficult to identify. The GEM system was also capable of identifying these large-scale changes.
In the study of 179 previously diagnosed pediatric cases, the GEM system ranked 92% of causal genes among the top or second candidate, and prioritized for review a median of 3 candidate genes per case. “In 17 of 20 cases with diagnostic SVs, GEM identified the causal SVs as the top candidate and in 19/20 within the top five, irrespective of whether SV calls were provided or inferred ab initio by GEM using its own internal SV detection algorithm,” explained the researchers.
The system was shown to outperform currently used tools and identify potential causal genetic changes in 60% of the time. The performance of the system was similar even without the parental genotype, and one novel finding was identified in 1 of the 14 unsolved cases.
“GEM enabled diagnostic interpretation inclusive of all variant types through automated nomination of a very short list of candidate genes and disorders for final review and reporting,” explained the researchers. “In combination with deep phenotyping by CNLP, GEM enables substantial automation of genetic disease diagnosis, potentially decreasing cost and expediting case review.”
You can read more about the research in the paper – Artificial intelligence enables comprehensive genome interpretation and nomination of candidate diagnoses for rare genetic diseases – which was recently published in Genomic Medicine. DOI: 10.1186/s13073-021-00965-0