An artificial intelligence (AI)-based technology rapidly diagnoses rare disorders in critically ill children with high accuracy, according to a report by researchers from University of Utah (U Of U) Health, Fabric Genomics, and collaborators on a study led by Rady Children’s Hospital in San Diego.
The finding (“Artificial intelligence enables comprehensive genome interpretation and nomination of candidate diagnoses for rare genetic diseases“), published in Genomic Medicine, foreshadows the next phase of medicine, where technology helps clinicians quickly determine the root cause of disease so they can give patients the right treatment sooner, according to the scientists.
“This study is an exciting milestone demonstrating how rapid insights from AI-powered decision support technologies have the potential to significantly improve patient care,” says Mark Yandell, PhD, co-corresponding author on the paper. Yandell is a professor of human genetics and Edna Benning Presidential Endowed Chair at U of U Health, and a founding scientific advisor to Fabric.
“We benchmarked GEM in a retrospective cohort of 119 probands, mostly NICU [newborn intensive care unit] infants, diagnosed with rare genetic diseases, who received whole-genome or whole-exome sequencing (WGS, WES). We replicated our analyses in a separate cohort of 60 cases collected from five academic medical centers. For comparison, we also analyzed these cases with current state-of-the-art variant prioritization tools. Included in the comparisons were trio, duo, and singleton cases.
“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. Finally, 14 previously unsolved cases were reanalyzed.
“GEM ranked over 90% of the causal genes among the top or second candidate and prioritized for review a median of 3 candidate genes per case, using either manually curated or CNLP-derived phenotype descriptions. Ranking of trios and duos was unchanged when analyzed as singletons. 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.
“GEM showed similar performance in absence of parental genotypes. Analysis of 14 previously unsolved cases resulted in a novel finding for one case, candidates ultimately not advanced upon manual review for 3 cases, and no new findings for 10 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. In combination with deep phenotyping by CNLP, GEM enables substantial automation of genetic disease diagnosis, potentially decreasing cost and expediting case review.”
Worldwide, about seven million infants are born with serious genetic disorders each year. For these children, life usually begins in intensive care. A handful of NICUs in the U.S., including at U of U Health, are now searching for genetic causes of disease by reading, or sequencing, the three billion DNA letters that make up the human genome. While it takes hours to sequence the whole genome, it can take days or weeks of computational and manual analysis to diagnose the illness.
For some infants, that is not fast enough, Yandell says. Understanding the cause of the newborn’s illness is critical for effective treatment. Arriving at a diagnosis within the first 24 to 48 hours after birth gives these patients the best chance to improve their condition. Knowing that speed and accuracy are essential, Yandell’s group worked with Fabric to develop the new Fabric GEM algorithm, which incorporates AI to find DNA errors that lead to disease.
In this study, the scientists tested GEM by analyzing whole genomes from 179 previously diagnosed pediatric cases from Rady’s Children’s Hospital and five other medical centers from across the world. GEM identified the causative gene as one of its top two candidates 92% of the time. Doing so outperformed existing tools that accomplished the same task less than 60% of the time.
“Dr. Yandell and the Utah team are at the forefront of applying AI research in genomics,” says Martin Reese, PhD, CEO of Fabric Genomics and a co-author on the paper. “Our collaboration has helped Fabric achieve an unprecedented level of accuracy, opening the door for broad use of AI-powered whole genome sequencing in the NICU.”
GEM leverages AI to learn from a vast and ever-growing body of knowledge that has become challenging to keep up with for clinicians and scientists, explains Reese. GEM cross-references large databases of genomic sequences from diverse populations, clinical disease information, and other repositories of medical and scientific data, combining all this with the patient’s genome sequence and medical records, he continues, adding that to assist with the medical record search, GEM can be coupled with a natural language processing tool, Clinithink’s CLiX focus, which scans reams of doctors’ notes for the clinical presentations of the patient’s disease.
“Critically ill children rapidly accumulate many pages of clinical notes,” Yandell says. “The need for physicians to manually review and summarize note contents as part of the diagnostic process is a massive time sink. The ability of Clinithink’s tool to automatically convert the contents of these notes in seconds for consumption by GEM is critical for speed and scalability.”
Existing technologies mainly identify small genomic variants that include single DNA letter changes, or insertions or deletions of a small string of DNA letters. By contrast, GEM can also find “structural variants” as causes of disease, according to the researchers. These changes are larger and are often more complex. It’s estimated that structural variants are behind 10 to 20% of genetic disease.
