Skip to main content
Premium Trial:

Request an Annual Quote

Emedgene Technology Shown to Automate Variant ID in 96 Percent of Cases


CHICAGO – At the American Society of Human Genetics annual meeting in mid-October, representatives from Baylor Genetics and the Baylor College of Medicine will present an abstract showing the efficacy of artificial intelligence in finding causative variants in whole-exome sequences.

The researchers will report that in a validation study of 200 cases of suspected rare Mendelian disorders, an automated interpretation platform from American-Israeli bioinformatics firm Emedgene helped geneticists identify the causative variant 96 percent of the time. This technology, according to both Baylor Genetics and Palo Alto, California-based Emedgene, helps increase diagnostic yield while also taking much of the manual workload off of overtaxed geneticists.

Lab customers run sequences through the Emedgene AI engine. The platform can be preloaded with the lab's standard operating procedures that users follow to identify unique variants in a patient's genome.

"We try to do things such as remove noise off of the bioinformatics pipeline and remove polymorphic variants within that population," said Emedgene CEO and Cofounder Einat Metzer. Rather than have geneticists manually sort through hundreds or thousands of variants, the system presents them with a list of, typically, no more than 10 most likely pathogenic variants.

In the Baylor study, the correct variant appeared on the short list in 192 of the 200 retrospective cases.

"It surfaces only those pieces of information that are most relevant to prove the case that a variant is potentially pathogenic," Metzer said of her company's platform.

"They now can start by looking at those particular potentially pathogenic variants that were identified by the AI," which can uncover those "hidden" in unstructured data or in new or obscure scientific literature, she said. Emedgene finds these causative variants and supporting evidence in genome sequences, exome sequences, and gene panels with the help of natural-language processing.

"With our NLP technology, we're constantly running and retrieving new pieces of information, new facts, and new evidence points to be able to present that to geneticists," Metzer said. This helps geneticists and clinicians diagnose rare diseases, screen healthy populations, and practice pharmacogenomics more accurately and efficiently.

"When we run our automated interpretation algorithms, [the algorithms] are examining dozens of additional gene-disease connections that aren't available in databases, and surfacing causative variants in those genes as well," added Orit Livnat-Levi, Emedgene's marketing director.

The technology makes geneticists more efficient, according to Fan Xia, senior director of clinical genomics at Baylor Genetics.

"It also helps us make the workflow more standardized," said Linyan Meng, a division director for clinical genomics interpretation at Baylor Genetics, a lab associated with Baylor College of Medicine in Houston. "The entire process is more uniform," compared to previous, manual review, Meng added.

Baylor conducts whole-exome sequencing for every suspected case of a rare Mendelian disease, and the lab has delivered some 14,000 exome test reports to patients since 2010, Xia said.

Meng said that Baylor started using Emedgene about a year ago, specifically working on the validation study and test development starting in November. The platform has been in production since April at Baylor for whole-exome sequencing analysis in the Baylor reference lab, not in the clinic or hospital.

"It has mainly improved our analysis efficiency. It takes us a shorter time to analyze cases and also it makes high-throughput analysis possible," Meng said.

The Baylor Genetics lab feeds sequencing data into the Emedgene platform in the cloud.

As a large lab, Baylor Genomics can give dozens or even hundreds of cases to Emedgene because there is adequate bandwidth, and most of the analysis happens overnight. "Every morning when we come in, we can just analyze a case," Xia said.

In the 4 percent of cases during the validation study that Emedgene could not produce a variant candidate for, Baylor had to rely on manual review, but the Baylor-specific rules that the vendor loaded onto the platform helped the lab hew more closely to its SOPs.

"We will be able to match all the previous pro formas with better consistency and a more efficient process," Xia said.

In February, Baylor announced that it was starting whole-genome sequencing for diagnosis of certain rare genetic disorders. Collaboration with Emedgene in this area is "on the roadmap," Xia said.


Livnat-Levi noted that Emedgene is capable of whole-genome interpretation, but that is a feature Baylor has not yet enabled.

Emedgene will announce new features around the time of ASHG. Metzer said this would include more use cases and improved workflow.

The three founders, Metzer, Chief Science Officer Shay Tzur, and Vice President for R&D Niv Mizrahi, started Emedgene in Israel in 2015. The company moved its headquarters and commercial operations to Silicon Valley in 2018, but still runs its R&D operations out of Tel Aviv.

Livnat-Levi said that half of the members of the R&D team are geneticists. "It's a very scientifically driven product," she said.

The technology platform has been in use for about two years. Emedgene entered the US market last year and landed a $6 million Series A round of venture capital in early 2019.

Emedgene now operates is in Europe, Latin America, and the US, though its primary focus is in the US. Its technology also is being used by the Israel Ministry of Health's Mosaic Project, an effort to sequence 100,000 human genomes by 2023.

"We are already crossing the boundaries of the lab and working with patient-facing organizations," Metzer said. "These are advanced organizations that are looking to adopt advanced genomic medicine. Utilizing this AI capability is a way for them to provide clinical decision-making at the point of care."

Future clients could be hospitals without genetic labs, she suggested.

"We are enabling them to take ownership of their patient data and to constantly interpret that instead of having that as a diagnostic task — it's enabling that as clinical decision support," Metzer said.