U.S. genomic intelligence company Genomenon has combined computer-based indexing of genomic publications with a scientific literature review to offer rapid insights into how genes can cause disease.
In a study that took less than six months, the company indexed 9.5 million full-text genomic publications with a systematic literature review conducted by genetic scientists.
Through this combination of natural language processing and parallel manual curation, it was able to identify thousands of gene-disease relationships (GDRs) and resolve variants of uncertain significance (VUS).
The findings were presented this week at the ACMG Annual Clinical Genetics Meeting in Toronto, Canada.
They follow last month’s release of 275 million previously unreported genetic variants as part of the National Institutes of Health (NIH) landmark All of Us program. Of those, nearly four million were in areas that may be tied to disease risk.
“With an onslaught of new sequencing data, it is becoming increasingly urgent to rapidly and accurately curate and characterize VUS and GDRs across all genes associated with the clinical exome,” said Genomenon’s chief scientific officer Mark Kiel, PhD.
He maintained that the results he presented demonstrated the power of integrating computational indexing with expert curation of scientific evidence.
“This approach allowed us to increase the speed and accuracy of defining variant pathogenicity, which is essential to keep pace with the publication of new variants and improve the precision of genetic diagnoses,” he explained.
Genomenon aims to provide genomic intelligence for pharma and clinical diagnostic labs, converting complex genetic data into actionable insights.
Specifically, it aims to uncover the genomic drivers of genetic disease and cancer to aid diagnosis and the development of precision medicine.
The company hopes to become the first to curate the entire human genome and understand the pathogenicity of any genetic variant.
It offers software, data, and curation solutions as well as a consulting service.
In the current study, Genomenon identified 10,745 germline GDRs and 5,973 germline GDRs with positive associations between a disease and gene.
This included 2225 genes, totaling 3075 curated GDRs, that were unique to the study.
Each GDR was accompanied by well-documented scientific evidence curated by Genomenon’s team of genetic scientists.
Results substantially agreed with ClinGen, the NIH-funded resource to define the the clinical relevance of genes and variants, at 79.8%.
Most discrepancies resulted from new evidence being published after the last ClinGen curation, reflecting the pace at which novel gene-disease relationships are being published.
Comparing the results with 12 separate groups submitting to the Gene Curation Coalition, a global effort to harmonize gene-level resources, revealed levels of disagreement of 14%. This was consistent with internal disagreement between individual submitters.
In his presentation, Kiel stated: “This study reflects an ability to curate the germline exome for gene-disease relationships using a combination of natural-language processing and massively parallelized manual curation.”
