Alex Zhavoronkov, Insilico

AI study led by In­sil­i­co's Alex Zha­voronkov bol­sters case for faster, cheap­er drug dis­cov­ery

Drug de­vel­op­ment is an ar­du­ous and ex­pen­sive busi­ness. The promise of ar­ti­fi­cial in­tel­li­gence is that ma­chines can wean man­u­fac­tur­ers away from the breadth of re­sources it takes to dis­cov­er a po­ten­tial­ly po­tent com­pound, and lever­age the med­ical con­di­tions it could be used to treat. Alex Zha­voronkov, whose Hong Kong-based AI-shop In­sil­i­co Med­i­cine has en­trenched it­self in the ten­ta­cles of bio­phar­ma R&D, now has some da­ta to bol­ster the fer­vor and in­vest­ment that AI has drummed up.

On Mon­day, Zha­voronkov and his fel­low AI scouts from Wuxi and sci­en­tists from the Uni­ver­si­ty of Toron­to, pub­lished a pa­per in Na­ture Biotech­nol­o­gy that sup­port­ed the fea­si­bil­i­ty of em­ploy­ing a ma­chine learn­ing ap­proach — gen­er­a­tive ad­ver­sar­i­al net­works (GANs) and re­in­force­ment learn­ing — for de no­vo drug de­sign. The re­searchers found that they were able to iden­ti­fy a myr­i­ad of com­pounds tar­get­ing a pro­tein called Dis­coidin do­main re­cep­tor 1 (DDR1) — which is ex­pressed in ep­ithe­lial cells and in­volved in fi­bro­sis — in a swift 21 days.

Af­ter that, six mol­e­cules were then se­lect­ed for syn­the­sis in the lab — those tests re­vealed four with po­ten­tial. Fur­ther test­ing whit­tled it down to one com­pound, which was test­ed in mice. That da­ta sug­gest­ed the mol­e­cule con­ferred a po­tent ef­fect against the pro­tein — al­though much like any oth­er com­pound, its safe­ty and ef­fi­ca­cy must be val­i­dat­ed in hu­man tri­als.

This study, which was con­duct­ed in re­sponse to a chal­lenge set by a part­ner com­pa­ny, is a taste of things to come, Zha­voronkov not­ed in an in­ter­view with End­points News. The plan was to take an al­go­rithm de­vel­oped a few years ago — an al­go­rithm that is open­ly avail­able, and use it as a proof-of-con­cept to demon­strate the po­ten­tial for AI in drug dis­cov­ery, he as­sert­ed.

“Imag­ine…Pfiz­er putting out all of their pre­clin­i­cal stuff for free for every­body to use, right? They won’t do it. Or J&J putting out all of their da­ta, even the old stuff? They just don’t do it. We de­cid­ed to kind of do a de­mo. So the skep­tics are… a lit­tle bit less skep­ti­cal.”

Re­cent es­ti­mates by the Tufts Cen­ter for the Study of Drug de­vel­op­ment main­tain that tak­ing a drug all the way from dis­cov­ery to ap­proval costs rough­ly $2.6 bil­lion (in 2013 dol­lars). Steven Paul, cur­rent Karuna chief,  pub­lished a study in 2010 in which he high­light­ed the mag­ni­tude of re­sources it cost his for­mer em­ploy­er Eli Lil­ly $LLY to dis­cov­er new com­pounds: the out-of-pock­et cost for lead op­ti­miza­tion came to a hefty $146 mil­lion.

“In this work, we de­signed, syn­the­sized, and ex­per­i­men­tal­ly val­i­dat­ed mol­e­cules tar­get­ing DDR1 ki­nase in less than 2 months and for a frac­tion of the cost as­so­ci­at­ed with a tra­di­tion­al drug dis­cov­ery ap­proach. This il­lus­trates the util­i­ty of our deep gen­er­a­tive mod­el for the suc­cess­ful, rapid de­sign of com­pounds that are syn­thet­i­cal­ly fea­si­ble, ac­tive against a tar­get of in­ter­est, and po­ten­tial­ly in­no­v­a­tive with re­spect to ex­ist­ing in­tel­lec­tu­al prop­er­ties,” Zha­voronkov et al wrote in their pa­per.

Zha­voronkov com­pared the study to Al­pha­Go — the first com­put­er pro­gram de­vel­oped by Al­pha­bet’s AI com­pa­ny Deep­Mind and im­mor­tal­ized in a Net­flix doc­u­men­tary — to de­feat a pro­fes­sion­al hu­man Go play­er. Go, which orig­i­nat­ed in Chi­na over 3,000 years ago, is a de­cep­tive­ly sim­ple strate­gic think­ing board game that has an in­cred­i­ble 10 to the pow­er of 170 pos­si­ble board con­fig­u­ra­tions (which is more than the num­ber of atoms in the known uni­verse.)

“We kind of thought about this pa­per as a kind of mi­ni Al­pha­Go,” Zha­voronkov said. “I hope that the big ex­ec­u­tives will al­so kind of hear this and un­der­stand that…we ac­tu­al­ly put the en­tire dis­cov­ery process on dis­play.”

“Every sec­ond stu­dent in Chi­na wants to be an AI sci­en­tist…this (Al­pha­Go) mir­a­cle might not have im­pact­ed their lives sig­nif­i­cant­ly, but it re­al­ly changed the men­tal­i­ty for every­body,” he added. “So that is what we need in phar­ma. Peo­ple should fo­cus less on geopol­i­tics or…war­fare. This is the stuff to watch right on TV. This is the cool thing.”

A pla­toon of bio­phar­mas have linked up with the emerg­ing crop of AI spe­cial­ists itch­ing to cap­i­tal­ize on how large datasets can be har­nessed to dri­ve new ther­a­pies in­to the clin­ic. Zha­voronkov is well con­nect­ed — last year he raised funds at the be­hest of Shang­hai high-fly­er WuXi AppTec, Sin­ga­pore’s Temasek, Pe­ter Dia­man­dis and Ju­ve­nes­cence.

Iso­lat­ing com­pounds for de­vel­op­ment is one as­pect of the bal­loon­ing AI in­dus­tri­al com­plex — it al­so has ap­peal in an­oth­er is­sue drug de­vel­op­ers reg­u­lar­ly con­tend with — the low odds of suc­cess, even with com­pounds that show great po­ten­tial in ear­ly test­ing. Then there’s the gold­en ques­tion — even if AI can help make the process of drug de­vel­op­ment bet­ter, faster and cheap­er — will that trans­late to less ex­pen­sive treat­ments?

A fa­vorite in Alex­ion’s C-suite is leav­ing, and some mighty sur­prised an­a­lysts aren’t the least bit hap­py about it

Analysts hate to lose a biotech CFO they’ve come to trust and admire — especially if they’re being blindsided by a surprise exit.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 59,800+ biopharma pros reading Endpoints daily — and it's free.

While No­var­tis ban­ish­es Zol­gens­ma scan­dal scars — Bio­gen goes on a Spin­raza 'of­fen­sive'

While Novartis painstakingly works to mop up the stench of the data manipulation scandal associated with its expensive gene therapy for spinal muscular atrophy (SMA) Zolgensma— rival Biogen is attempting to expand the use of its SMA therapy, Spinraza. 

The US drugmaker $BIIB secured US approval for Spinraza for use in the often fatal genetic disease in 2016. The approval covered a broad range of patients with infantile-onset (most likely to develop Type 1) SMA. 

Jason Kelly. Mike Blake/Reuters via Adobe

Eye­ing big ther­a­peu­tic push, Gink­go bags $290M to build a cell pro­gram­ming em­pire

Ginkgo Bioworks is on a roll. Days after publicizing a plan to nurture new startups via partnerships with accelerators Y Combinator and Petri, the Boston biotech says it has raised another $290 million for its cell programming platform to reach further and wider.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 59,800+ biopharma pros reading Endpoints daily — and it's free.

Speak­er Nan­cy Pelosi to un­veil bill for fed­er­al­ly ne­go­ti­at­ed drug prices

After months of buzz from both sides of the aisle, Speaker Nancy Pelosi will today introduce her plan to allow the federal government to negotiate prices for 250 prescription drugs, setting up a showdown with a pharmaceutical industry working overtime to prevent it.

The need to limit drug prices is a rare point of agreement between President Trump and Democrats, although the president has yet to comment on the proposal and will likely face pressure to back a more conservative option or no bill at all. Republican Senator Chuck Grassley is reportedly lobbying his fellow party members on a more modest proposal he negotiated with Democratic Senator Ron Wyden in July.

David Grainger [file photo]

'Dis­con­nect the bas­tard­s' — one biotech's plan to break can­cer cell­s' uni­fied de­fens­es

Chemotherapy and radiotherapy are the current gladiators of cancer treatment, but they come with well-known limitations and side-effects. The emergence of immunotherapy — a ferocious new titan in oncologist’s toolbox — takes the brakes off the immune system to kill cancer cells with remarkable success in some cases, but the approach is not always effective. What makes certain forms of cancer so resilient? Scientists may have finally pieced together a tantalizing piece of the puzzle, and a new biotech is banking on a new approach to fill the gap.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 59,800+ biopharma pros reading Endpoints daily — and it's free.

Jeff Kindler's Cen­trex­ion re­news bid to make pub­lic de­but

Jeffrey Kindler’s plan to take his biotech — which is developing a slate of non-opioid painkillers — public, is back on.

The Boston based company, led by former Pfizer $PFE chief Kindler, originally contemplated a $70 million to $80 million IPO last year— but eventually postponed that strategy. On Wednesday, the company revived its bid to make a public debut in a filing with the SEC — although no pricing details were disclosed.

Zachary Hornby. Boundless

'A fourth rev­o­lu­tion in can­cer ther­a­pies': ARCH-backed Bound­less Bio flash­es big check, makes big­ger promis­es in de­but

It was the cellular equivalent of opening your car door and finding an active, roaring engine in the driver seat.

Scientists learned strands of DNA could occasionally appear outside of its traditional home in the nucleus in the 1970s, when they appeared as little, innocuous circles on microscopes; inexplicable but apparently innate. But not until UC San Diego’s Paul Mischel published his first study in Science in 2014 did researchers realize these circles were not only active but potentially overactive and driving some cancer tumors’ superhuman growth.

It’s fi­nal­ly over: Bio­gen, Ei­sai scrap big Alzheimer’s PhI­I­Is af­ter a pre­dictable BACE cat­a­stro­phe rais­es safe­ty fears

Months after analysts and investors called on Biogen and Eisai to scrap their BACE drug for Alzheimer’s and move on in the wake of a string of late-stage failures and rising safety fears, the partners have called it quits. And they said they were dropping the drug — elenbecestat — after the independent monitoring board raised concerns about…safety.

We don’t know exactly what researchers found in this latest catastrophe, but the companies noted in their release that investigators had determined that the drug was flunking the risk/benefit analysis.

Endpoints News

Keep reading Endpoints with a free subscription

Unlock this story instantly and join 59,800+ biopharma pros reading Endpoints daily — and it's free.

Mer­ck helps bankroll new part­ner Themis' game plan to fin­ish the chikun­gun­ya race and be­gin on­colyt­ic virus quest

As Themis gears up for a Phase III trial of its chikungunya vaccine, the Vienna-based biotech has closed out €40 million ($44 million) to foot the clinical and manufacturing bills.

Its heavyweight partners at Merck — which signed a pact around a mysterious “blockbuster indication” last month — jumped into the Series D, led by new investors Farallon Capital and Hadean Ventures. Adjuvant Capital also joined, as did current investors Global Health Investment Fund, aws Gruenderfonds, Omnes Capital, Ventech and Wellington Partners Life Sciences.