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.”

The FDA’s Oncology Center of Excellence has been a bright spot within the agency in terms of speeding new treatments to patients. That flexibility was on full display this morning as FDA released new draft guidance spelling out exactly how oncology drug developers can fulfill both the accelerated and full approval’s requirements with just a single randomized controlled trial.

While Congress recently passed legislation that will allow FDA to require confirmatory trials to be recruiting and ongoing prior to granting an accelerated approval, the agency is now making clear that the initial trial used to win the AA, if designed appropriately, can also serve as the trial for converting the accelerated approval into a full approval.

US regulators cleared an ultra-rare drug from Pharming Group, by way of Novartis, on Friday afternoon.

The Dutch biotech said the FDA greenlit leniolisib for an immunodeficiency disease known as activated phosphoinositide 3-kinase delta (PI3Kδ) syndrome, or APDS. People 12 years and older can receive the oral drug, to be marketed as Joenja, beginning early next month, Pharming said, five days ahead of the decision deadline set by the FDA as part of a priority review.