Bren­dan Frey set out, when he be­gan the project to pick out a lead pro­gram for Deep Ge­nomics, to prove that the ar­ti­fi­cial in­tel­li­gence sys­tems his lab has de­signed can iden­ti­fy new drug tar­gets and find a win­ning can­di­date much faster than tra­di­tion­al meth­ods. Now that they have ze­roed in on an an­ti­sense oligonu­cleotide ex­on-skip­ping ther­a­py for a sub­type of Wil­son dis­ease — se­lect­ed out of 2,400 ail­ments and 120,000 un­der­ly­ing ge­net­ic mu­ta­tions — as their face case, the Toron­to-based biotech is ready to delve in­to new fron­tiers with their AI tech.

Deep Ge­nomics plans to spend rough­ly half of its new $40 mil­lion on clin­i­cal de­vel­op­ment of DG12P1, whose first-in-hu­man tri­al is slat­ed for 2021; and the oth­er half on strength­en­ing var­i­ous com­po­nents of its ma­chine learn­ing sys­tems, from ro­bot­ics to as­says. Fu­ture Ven­tures, an in­vestor in Tes­la and SpaceX, led the Se­ries B round while Am­pli­tude Ven­tures, Mag­net­ic Ven­tures, Khosla Ven­tures and True Ven­tures joined.

While they took a splice mod­u­lat­ing ap­proach for their first AI-dis­cov­ered com­pound akin to what Bio­gen has with Spin­raza, Deep Ge­nomics is next fo­cused on hap­loin­suf­fi­cien­cies — dis­or­ders where in­creased ex­pres­sion lev­els of one func­tion­al gene would be valu­able. In oth­er words, where­as the pre­vi­ous chal­lenge lied in find­ing the right mu­ta­tion to tin­ker, the new puz­zle is how best to boost a known tar­get.

It’s a rel­a­tive­ly new ap­pli­ca­tion of oligonu­cleotides with no ap­proved drugs, which is al­so be­ing ex­plored at Stoke Ther­a­peu­tics.

“There are many dif­fer­ent mech­a­nisms that could be rel­e­vant, and so when you look at all these dif­fer­ent mech­a­nisms and all these dif­fer­ent re­gions of the gene that you could de­sign the oli­go for, there are tens of thou­sands, or even hun­dreds of thou­sands of pos­si­ble com­pounds,” says Frey, a star re­searcher that in­sil­i­co’s Alex Zha­voronkov con­sid­ers “with­out doubt in top 10 sci­en­tists in this field in the world.”

Frey elab­o­rates:

That could be in the 5’ UTR, it could be in the in­tron, it could be in an ex­on, in­tron­ic se­quences are very large, could be in a 3’ UTR, dif­fer­ent kinds of mech­a­nisms may be in­volved. It could be a mat­ter of al­ter­ing the up­stream open read­ing frame, or it could be a mat­ter of an in­tron or ten­sion bot­tle neck, it could be a mat­ter of chang­ing the polyadeny­la­tion site, it could be a com­pound that al­ters sec­ondary struc­tures.

New tools will be re­quired to test all these paths they can po­ten­tial­ly trav­el down, and Deep Ge­nomics’ team of 40-plus is al­ready per­fect­ing one mod­el de­signed to pre­dict polyadeny­la­tion pat­terns.

“Mul­ti­lin­gual­ism is an im­por­tant core val­ue for us,” he pre­vi­ous­ly said. “Every­one at the com­pa­ny has had AI train­ing, and every­one at the com­pa­ny has done wet lab work. In fact I ac­tu­al­ly se­quenced the ge­nom­ic DNA for the Wil­son tar­get to val­i­date it once we edit­ed the cell line to put the pa­tient mu­ta­tion in­to the cells.”

In ad­di­tion to deep­en­ing the clin­i­cal bench, Frey is al­so re­cruit­ing for the busi­ness de­vel­op­ment unit.

“Our pipeline is over­flow­ing, and so we’re fo­cused on sign­ing some busi­ness deals,” he tells End­points News in the lead­up to the an­nu­al JP Mor­gan con­fab in San Fran­cis­co.

That in­cludes two ad­di­tion­al pro­grams for Wil­son’s dis­ease that, to­geth­er with the lead pro­gram, would ad­dress up to a quar­ter of the 230,000 pa­tients world­wide. There’s one oth­er undis­closed meta­bol­ic can­di­date for which Frey plans to sub­mit an IND this year, some com­pounds for retinopa­thy, and a dozen hap­loin­suf­fi­cient projects in de­vel­op­ment, he says.

The co-owner of a Florida-based clinical research site pleaded guilty to lying to an FDA investigator during a 2017 inspection, revealing that she falsely portrayed part of a GlaxoSmithKline pediatric asthma study as legitimate, when in fact she knew that certain data had been falsified, the Department of Justice said Wednesday.

Three other employees — Yvelice Villaman Bencosme, Lisett Raventos and Maytee Lledo — previously pleaded guilty and were sentenced in connection with falsifying data associated with the trial at the CRO Unlimited Medical Research.

New research shows how generic drug companies can successfully market a limited number of approved indications for a brand name drug, prior to coming to market for all of the indications. But several recent court decisions have created a layer of uncertainty around these so-called “skinny” labels.

While courts have generally allowed generic manufacturers to use their statutorily permitted skinny-label approvals, last summer, a federal circuit court found that Teva Pharmaceuticals was liable for inducing prescribers and patients to infringe GlaxoSmithKline’s patents through advertising and marketing practices that suggested Teva’s generic, with its skinny label, could be employed for the patented uses.