
F-Prime backs a niche AI software startup hunting lofty goals in $22M Series B
For many of the AI companies sprouting on the biopharma field, validation — often meaning confirmation of whether the targets and drugs they identified or generated would actually work — won’t come in years, if at all. But for BenchSci, the drug hunting field is their home turf.
To be sure, the Toronto-based startup is doing something very different from the rest of the pack. Rather than staking claims about the results of drug discovery, it’s out to change the process by hunkering down on a specific problem: helping scientists select the right reagents to conduct their preclinical experiments.
Having started out with an antibody selection service 18 months ago, BenchSci is now ready to roll out a broader reagent selection platform and expand the clientele from academic institutions and Big Pharma to biotechs — thanks to a $22 million cash injection.
In contrast with the clinical or commercial realm, BenchSci found that its challenge wasn’t to compete with rival vendors but to convince investors that there’s a market, CEO Liran Belenzon told Endpoints News.
“There is no software company in preclinical,” he said, leaving scientists to work on software with interfaces that “look like they’re from the 1990s.”
But the experience of his co-founder and CSO Tom Leung, who saw first hand how an inappropriate antibody could cost him rare patient samples and lead to monthlong delays, and subsequent chats with other researchers convinced them there’s an opportunity here. Collating data from open access journals and inking deals with big name publishers like Springer Nature, Wiley and JAMA, BenchSci came up with a database of scientific literature that they then teach the computer to read.
“So the scientist basically asks our system the question: Out of those 5,000 reagents or antibodies that are out there, which one will work on BRCA1, in this tissue, in this specific cell line, in this model, with this specification,” Belenzon said, “and we really narrow down these 4 or 5,000 to 2 or 3 and then we say, hey, these are the antibodies most likely to work in your experiment, and here’s all the scientific data and the experimental results that this specific reagent has generated and scientists can actually see those results and validate it as well.”
That means condensing the whole process of selecting reagents — traditionally done by trial and error — from 12 weeks to 30 seconds, according to the company, reducing waste by 70% and saving millions of dollars in hard cost.
It’s not quite shaving years and tens of millions off the drug discovery process as others have promised (and many have doubted). Yet BenchSci still cites some big numbers: a $10.2 billion per year opportunity for savings on reagents, deduced from the estimate that $28 billion each year is wasted on irreproducible research, with reagents and reference materials accounting for 36.1%.
F-Prime Capital, which led the Series B, and other investors including Northleaf Capital Partners, Gradient Ventures, Inovia Capital, Golden Ventures and Real Ventures would love to see them get there. But will they?
“They would be a perfect acquisition target for Genescript, Qiagen, Thermo Fisher, or other antibody and reagent maker,” Alex Zhavoronkov, founder and CEO of AI drug discovery startup Insilico Medicine, wrote to Endpoints. “With this model they can quickly get to substantial revenue possibly in tens of millions but the market size is rather limited and it will be difficult to grow.”
As he prepares to double the size of his team to 140 to support the growth into recombinant proteins, RNAi, CRISPR, cell lines and more, Belenzon sees otherwise.
“Today in an age where you have AI companies generating more and more and more potentially great targets to study, there needs to be a company that helps to study those targets faster and better and cheaper,” he said. “That’s really what we are focusing on.”