Riding on wave of clinical trial reforms, machine learning startup bags $50M to create ‘digital twins’ – Endpoints News

As drug de­vel­op­ers and reg­u­la­tors alike in­creas­ing­ly warm to cre­ative ideas for run­ning clin­i­cal tri­als, a ma­chine learn­ing plat­form cre­at­ed by three physi­cists is draw­ing in $50 mil­lion from tech VCs.

Un­learn bills it­self as the on­ly com­pa­ny that can gen­er­ate “dig­i­tal twins” of pa­tients for use in clin­i­cal tri­als, so that bio­phar­ma com­pa­nies can test their drugs with few­er re­al pa­tients than they would need to in tra­di­tion­al tri­als, but get sim­i­lar­ly, if not even more, re­li­able re­sults.

“Our prod­uct is not an AI mod­el — it’s a clin­i­cal tri­al,” CEO Charles Fish­er wrote in an email in­ter­view with TechCrunch.

He di­vulged that Ger­many’s Mer­ck KGaA is among three drug­mak­ers us­ing Un­learn’s plat­form in the de­sign of its clin­i­cal tri­als — al­though it’s not di­rect­ly for the dig­i­tal twin ser­vice but prog­nos­tic in­for­ma­tion from the dig­i­tal twins.

Where­as there have been ef­forts and guid­ance from the FDA to use re­al-world da­ta to sup­port reg­u­la­to­ry de­ci­sions, none have quite gone as far as in­cor­po­rat­ing con­struct­ed pa­tient pro­files.

Un­learn be­lieves its tech­nol­o­gy can ac­com­plish that through a com­bi­na­tion of ma­chine learn­ing and new sta­tis­ti­cal meth­ods. By tak­ing his­tor­i­cal con­trol da­ta from pre­vi­ous clin­i­cal tri­als and train­ing a ma­chine learn­ing mod­el on them, it promis­es to com­pute a dig­i­tal twin for every pa­tient en­rolled in a tri­al so that more pa­tients can be in the treat­ment arm and few­er in the con­trol arm.

Fish­er met his co-founders, Aaron Smith and Jon Walsh, while work­ing at a now-de­funct start­up that de­vel­oped mo­tion sen­sors, TechCrunch re­port­ed. They then band­ed to­geth­er to see if they could turn clin­i­cal tri­al da­ta sets to build dis­ease-spe­cif­ic ma­chine learn­ing mod­els.

“[Our] in­ten­tion wasn’t to speed clin­i­cal tri­als — it was pure re­search in­to ma­chine learn­ing. But [I] had a phar­ma­ceu­ti­cal in­dus­try back­ground and it soon be­came ap­par­ent that there had been no in­vest­ment in ma­chine learn­ing as a tech­nol­o­gy for phar­ma de­vel­op­ment,” said Fish­er, who was pre­vi­ous­ly a prin­ci­pal sci­en­tist at Pfiz­er. “[Un­learn] evolved through in­ter­ac­tion[s] with the phar­ma­ceu­ti­cal in­dus­try.”

The dig­i­tal twins would in­cor­po­rate da­ta from tri­als over time and cov­er a wide range of bi­o­log­i­cal in­for­ma­tion, the com­pa­ny added.

“By re­duc­ing the size of the con­trol arm, more pa­tients in a Twin­RCT have ac­cess to a po­ten­tial­ly ben­e­fi­cial ex­per­i­men­tal treat­ment in­stead of a place­bo,” Dy­lan Mor­ris, man­ag­ing di­rec­tor at In­sight Part­ners, said in a pre­pared state­ment. “Tri­als can be run faster and with the same re­sources so that pa­tients can re­ceive ac­cess to more ef­fec­tive treat­ments soon­er.”

Spread the love

Leave a Reply

Your email address will not be published.