Out to ‘re­think the drug de­sign process,’ AI star Daphne Koller is cre­at­ing a new ma­chine learn­ing ven­ture at in­sitro with some heavy­weight back­ers

Daphne Koller has been busy.

Just two months since the high pro­file AI ex­pert ex­it­ed Google’s an­ti-ag­ing biotech Cal­i­co Labs, where she was chief com­put­ing of­fi­cer, Koller has gath­ered a group of mar­quee in­vestors to back a tech up­start — in­sitro — with plans to de­vel­op a ma­chine learn­ing plat­form equipped with cus­tom-built datasets to cre­ate a new and far more ef­fi­cient ap­proach to drug dis­cov­ery and de­vel­op­ment.

AI cir­cles have been buzzing with spec­u­la­tion and queries about what the for­mer Stan­ford pro­fes­sor planned to do af­ter ex­it­ing Cal­i­co, where she had been de­vel­op­ing com­pu­ta­tion­al meth­ods for an­a­lyz­ing bi­o­log­i­cal da­ta sets. She an­swered at least some of those ques­tions with a blog post late Tues­day that spelled out her com­mit­ment to “in­vest heav­i­ly” in cre­at­ing new datasets that can ac­cel­er­ate the use of ma­chine learn­ing in bio­phar­ma.

Koller has some high rollers back­ing the start­up, in­clud­ing some ven­ture groups well known for block­buster fundrais­ing and a yen for trans­for­ma­tion­al ven­tures. They are: Arch Ven­ture Part­ners, Fore­site Cap­i­tal, a16z, GV (for­mer­ly Google Ven­tures) and Third Rock.

In­sitro, she says, will “col­lect and use a range of very large da­ta sets to train ML mod­els that will help ad­dress key prob­lems in the drug dis­cov­ery and de­vel­op­ment process. To en­able the ma­chine learn­ing, we will use high-qual­i­ty da­ta that has al­ready been col­lect­ed, but we will al­so in­vest heav­i­ly in the cre­ation of our own datasets us­ing high through­put ex­per­i­men­tal ap­proach­es, datasets that are de­signed ex­plic­it­ly with ma­chine learn­ing in mind from the very start. The ML mod­els that are de­vel­oped will then help guide sub­se­quent ex­per­i­ments, pro­vid­ing a tight, closed loop in­te­gra­tion of in sil­i­co and in vit­ro meth­ods (an in­sitro par­a­digm).”

Ac­cord­ing to Koller, the low hang­ing fruit in drug dis­cov­ery has been picked. Reach­ing high­er, go­ing for much bet­ter drugs, will re­quire “a dif­fer­ent ap­proach to drug de­vel­op­ment.” Spend­ing bil­lions to de­vel­op new drugs — and then pass­ing the cost to pa­tients — is not sus­tain­able.

In launch­ing in­sitro with a group of hy­per con­nect­ed back­ers, Koller is in­stant­ly mak­ing her­self a top play­er in a field that has Big Phar­ma’s rapt at­ten­tion. Stream­lin­ing R&D and im­prov­ing the odds of suc­cess find­ing high im­pact ther­a­pies are con­sid­ered keys to longterm prof­itabil­i­ty. But now the fo­cus is on which out­fits can ac­tu­al­ly de­liv­er.

There will be plen­ty of peo­ple watch­ing to see if Koller and the team she’s now re­cruit­ing can per­form. And she knows it won’t be easy. She writes:

There is a lot of hype to­day around ma­chine learn­ing, with hy­per­bol­ic promis­es that it will mag­i­cal­ly solve all of hu­mankind’s prob­lems (and dire warn­ings that it will lead to the de­struc­tion of hu­mankind). We at in­sitro don’t ex­pect ML to be the so­lu­tion to all of the prob­lems in drug de­vel­op­ment, nor to be the mag­ic bul­let that helps find a treat­ment for every dis­ease. How­ev­er, we do be­lieve that the time is right to re­think the drug de­sign process us­ing a dif­fer­ent and more mod­ern toolk­it, in the hope that a new par­a­digm may help us cure more peo­ple, soon­er, and at a much low­er cost.

I’d queried Koller re­cent­ly af­ter hear­ing word about in­sitro. For now, she tells me in a mes­sage, her blog post goes about as deep in­to this as she wants to go right now. But more is com­ing.


Im­age: Daphne Koller. STAN­FORD

What Will it Take to Re­al­ize the Promise and Po­ten­tial of Im­mune Cell Ther­a­pies?

What does it take to get to the finish line with a new cancer therapy – fast? With approvals in place and hundreds of immune cell therapy candidates in the pipeline, the global industry is poised to create a fundamental shift in cancer treatments towards precision medicine. At the same time, unique challenges associated with cell and process complexity present manufacturing bottlenecks that delay speed to market and heighten cost of goods sold (COGS) — these hurdles must be overcome to make precision treatments an option for every cancer patient. This series of articles highlights some of the key manufacturing challenges associated with the production of cell-based cancer therapies as well as the solutions needed to transcend them. Automation, process knowledge, scalability, and assured supply of high-quality starting material and reagents are all critical to realizing the full potential of CAR-based therapies and sustaining the momentum achieved in recent years. The articles will highlight leading-edge technologies that incorporate these features to integrate across workflows, accelerate timelines and reduce COGS – along with how these approaches are enabling the biopharmaceutical industry to cross the finish line faster with new treatment options for patients in need.

The biggest ques­tions fac­ing gene ther­a­py, the XLMTM com­mu­ni­ty, and Astel­las af­ter fourth pa­tient death

After three patients died last year in an Astellas gene therapy trial, the company halted the study and began figuring out how to safely get the program back on track. They would, executives eventually explained, cut the dose by more than half and institute a battery of other measures to try to prevent the same thing from happening again.

Then tragically, Astellas announced this week that the first patient to receive the new regimen had died, just weeks after administration.

Endpoints Premium

Premium subscription required

Unlock this article along with other benefits by subscribing to one of our paid plans.

Lat­est news: It’s a no on uni­ver­sal boost­ers; Pa­tient death stuns gene ther­a­py field; In­side Tril­li­um’s $2.3B turn­around; and more

Welcome back to Endpoints Weekly, your review of the week’s top biopharma headlines. Want this in your inbox every Saturday morning? Current Endpoints readers can visit their reader profile to add Endpoints Weekly. New to Endpoints? Sign up here.

Next week is shaping up to be a busy one, as our editor-in-chief John Carroll and managing editor Kyle Blankenship lead back-to-back discussions with a great group of experts to discuss the weekend news and trends. John will be spending 30 minutes with Jake Van Naarden, the CEO of Lilly Oncology, and Kyle has a brilliant panel lined up: Harvard’s Cigall Kadoch, Susan Galbraith, the new head of cancer R&D at AstraZeneca, Roy Baynes at Merck, and James Christensen at Mirati. Don’t miss out on the action — sign up here.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 117,600+ biopharma pros reading Endpoints daily — and it's free.

Mi­rati's KRAS drug looks like the fa­vorite in colon can­cer with new da­ta, putting the pres­sure square on Am­gen

With Amgen already providing proof-of-concept for KRAS inhibitors with its sotorasib, Mirati Therapeutics is piecing together a follow-up effort in lung cancer with data it thinks are superior. But in colon cancer, where solo sotorasib has turned in a dud, Mirati may now have a strong case for superiority.

Mirati’s adagrasib, dosed solo or in combination with chemotherapy cetuximab, showed response rates grater than sotorasib solo  and as part of combination study in a similar patient population also revealed this week at #ESMO21. Mirati’s data were presented as part of a cohort update from the Phase II KRYSTAL-1 study testing adagrasib in a range of solid tumors harboring the KRAS-G12C mutation.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 117,600+ biopharma pros reading Endpoints daily — and it's free.

President Biden and Pfizer CEO Albert Bourla (Patrick Semansky/AP Images)

Chaot­ic ad­comm sees Pfiz­er/BioN­Tech boost­ers re­ject­ed for gen­er­al pop­u­la­tion, but rec­om­mend­ed for old­er and high-risk pop­u­la­tions

With just days before President Joe Biden’s Covid-19 booster rollout is set to go into effect, an FDA advisory committee appeared on the verge of not recommending boosters for anyone in the US before a last-minute change of wording laid the groundwork for older adults to have access to a third dose.

The FDA’s adcomm on Vaccines and Related Biological Products (VRBPAC) roundly rejected Pfizer/BioNTech booster shots for all individuals older than 16 by a 16-2 vote Friday afternoon. Soon after, however, the agency posed committee members a new question limiting booster use to the 65-and-older population and individuals at high risk of disease due to occupational exposure or comorbidities.

The best of the rest: High­lights from the be­low-the-fold pre­sen­ta­tions at #ES­MO21

This year’s ESMO Congress has had a major focus on Big Pharma drugs — most notably candidates from Merck and AstraZeneca — but there have also been updates from smaller biotechs with data looking to challenge the big-name drugmakers.

Today, we’re highlighting some of the data releases that flew under the radar at #ESMO21 — whether from early-stage drugs looking to make a mark or older stalwarts with interesting follow-up data.

As­traZeneca, Dai­ichi Sanky­o's ADC En­her­tu blows away Roche's Kad­cy­la in sec­ond-line ad­vanced breast can­cer

AstraZeneca and Japanese drugmaker Daiichi Sankyo think they’ve struck gold with their next-gen ADC drug Enhertu, which has shown some striking data in late-stage breast cancer trials and early solid tumor tests. Getting into earlier patients is now the goal, starting with Enhertu’s complete walkover of a Roche drug in second-line breast cancer revealed Saturday.

Enhertu cut the risk of disease progression or death by a whopping 72% (p=<0.0001) compared with Roche’s ADC Kadcyla in second-line unresectable and/or metastatic HER2-positive breast cancer patients who had previously undergone treatment with a Herceptin-chemo combo, according to interim data from the Phase III DESTINY-Breast03 head-to-head study presented at this weekend’s #ESMO21.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 117,600+ biopharma pros reading Endpoints daily — and it's free.

Merck Research Laboratories CMO Roy Baynes

Mer­ck­'s Keytru­da un­corks full da­ta on lat­est ad­ju­vant win — this time in melanoma — adding bricks to ear­ly can­cer wall

In recent months, the battle for PD-(L)1 dominance has spilled over into early cancer with Merck’s Keytruda and Bristol Myers Squibb’s Opdivo all alone on the front lines. Keytruda now has another shell in its bandolier, and it could spell a quick approval.

Keytruda cut the risk of relapse or death by 35% over placebo (p=0.00658) in high-risk, stage 2 melanoma patients who had previously undergone surgery to remove their tumors, according to full data from the Phase III KEYNOTE-716 presented Saturday at #ESMO21.

Mer­ck flesh­es out Keytru­da win in first-line cer­vi­cal can­cer, adding more fire­pow­er to its ear­ly can­cer push

Merck has worked hard to bring its I/O blockbuster Keytruda into earlier and earlier lines of therapy, and now the wonder drug appears poised to make a quick entry into early advanced cervical cancer.

A combination of Keytruda and chemotherapy with or without Roche’s Avastin cut the risk of death by 33% over chemo with or without Avastin (p=<0.001) in first-line patients with persistent, recurrent or metastatic cervical cancer, according to full data from the Phase III KEYNOTE-826 study presented Saturday at #ESMO21.