CHICAGO – PercayAI, spun out of the Genome Technology Access Center at the McDonnell Genome Institute at Washington University School of Medicine in St. Louis, emerged from stealth this month with a series of announcements, but the company has been leaving breadcrumbs hinting of its existence for a while now.
PercayAI officially launched Oct. 14 by introducing its first product, called CompBio. This offering applies artificial intelligence and contextual language processing to help researchers identify relationships within multi-omic datasets to inform drug discovery.
Days later, the startup unveiled its first partnership, with St. Louis-based Canopy Biosciences, which contracted with PercayAI to bring augmented intelligence to its multi-omic and gene expression testing services.
Last week, after displaying its technology at the Society for Neuroscience annual meeting, PercayAI announced that it has joined the Pistoia Alliance.
Pronounced "per-chy-ah," the name comes from an Indonesian word for "believe" or "believe in," according to Chief Commercial Officer Preston Keller. "You can think of PercayAI as believing in AI," he said.
In this case, the AI in the name stands for augmented intelligence rather than artificial intelligence. While PercayAI does use some of the same machine learning and deep learning techniques that might fall under artificial intelligence, Keller said that the company does not want the computer to make decisions for scientists.
"We aim to put the ability to form complex hypotheses or derived answers back in the hands of the human brain and researchers," he said. "Our technology extends and expands the capabilities of any individual human researcher beyond what they could do on their own."
CompBio helps researchers sort through the millions of publications indexed on PubMed and other data sources in search of relevant connections between the biological data they have and existing medical literature. The software performs what the company calls contextual language processing, which Keller said is a differentiator between CompBio and the multitude of other platforms that comb through the literature.
"We actually go through the extra step of being able to organize only the words that are relevant for the particular experiment, meaning there's been some sort of link between them within the corpus of literature that we're looking at," Keller said. For example, he explained, "occlusion" has a different meaning in neurodegeneration than it does in cardiovascular disease.
"CompBio, when you feed it a set of data, actually knows where you are and is able to figure it out and return the interconnections for that particular biology in that highly contextual manner that's relevant to the data at hand," Keller said. "It is not just an interconnection of similar words and synonyms throughout the entire corpus of literature."
The idea that became CompBio came from the laboratory of computational biologist Richard Head, director of the WUSTL Genome Technology Access Center. Local investment firm Kingdom Capital had sponsored the technology development for the last couple of years and has contributed an unspecified sum to PercayAI.
"We've been intimately involved with this technology from the beginning and we went through a validation test with it at WashU last year," said Keller, who also is vice president in charge of health and medical investments at the VC firm and a director of Canopy Biosciences.
Kingdom Capital invests in ideas hatched at research universities like WashU and, when the time is right, helps those institutions commercialize new products, particularly in life sciences.
Keller said that Head and colleagues grew tired of having to use spreadsheets to conduct and track literature searches. They also found it labor-intensive to use older tools for pathway enrichment.
"Generally you'd input 100 genes and you'd get 1,000 genes back across all types of different biology," Keller said. "They didn't find that useful."
He said that the bioinformaticians and computational biologists at PercayAI have been training the CompBio algorithm "across the entire body of biological research that happens at Washington University."
PercayAI initially is aiming CompBio at drug discovery and development. Keller said that the software is well suited for target identification, lead optimization, and drug repurposing activities.
CompBio can take the results of large, multi-omics experiments, perhaps 20 to as many as 1,500 entities, and show how genes, transcripts, proteins, metabolites, and other entities are interconnected across biology. "In particular, it's been extremely predictive" for translating animal-based models to human testing, Keller said.
PercayAI put CompBio through a 40-laboratory validation process last summer, using the software not only for retrospective analysis, but also for creating actionable, multidimensional hypotheses, according to Keller.
"Those hypotheses led to interesting new biology insights that either would not have been made before or are profoundly different from what was originally suspected by the researcher," Keller said.
The first article referencing the CompBio technology — but not mentioning the name — appeared in the Proceedings of the National Academies of Sciences in May. Keller said that this paper validates several hypotheses generated on CompBio.
Keller added that there are more than a dozen other papers on CompBio in prepublication stages.
Longer-term plans for the emerging company call for extending CompBio technology to clinical settings to connect patient genotypes and phenotypes with early phases of drug discovery to enable personalized medicine, Keller said.
While CompBio is PercayAI's first product, there are others in the works "The next series of products that we'll be releasing comprise what we like to call potentially the world's first complete digital representation of human biology," Keller said.
While a multitude of companies sells AI technologies that help scientists weed through all the literature and then match knowledge to genotypes and phenotypes, PercayAI is trying to stand out.
Canopy Biosciences, which, like PercayAI, is located in the Cortex Innovation District in St. Louis, chose CompBio because of its potential to improve the identification of pathways for drug discovery, according to Canopy CEO Edward Weinstein.
Canopy performs sequencing services and multi-omics analysis, including NanoString testing, RNA sequencing, and using the ChipCytometry microfluidic platform that the company acquired from German startup Zellkraftwerk. That acquisition is helping the company establish itself in protein analysis.
Regardless of the specific technology, researchers are having to search for changes in hundreds or thousands of transcripts or proteins. "The researcher has to take all of that information and make sense of it somehow," Weinstein said.
While Canopy Biosciences has built its own data-analysis technology, Weinstein expects CompBio to help that platform comb through reams of medical literature.
"[CompBio] takes the gene list and creates a signaling pathway based on its proprietary algorithm by being able to look through all the publications in PubMed and put together connections," Weinstein said. "It's not just mapping it onto the list of a couple hundred preexisting pathways."
Weinstein said that CompBio has shown itself to have "more rigor" than other literature search engines.
"We knew what signaling pathways should have been modulated, but we blinded the data and came out with the relevant pathways being drawn up from PubMed and from the algorithm that searches both those PubMed-related genes and pathways. It was amazing how well it identifies the pathways," he said.