Cel­lar­i­ty start­ed with a sim­ple — but far from easy — idea that Avak Kahve­jian and his team were float­ing around at Flag­ship Pi­o­neer­ing: to dig­i­tal­ly en­code a cell.

As he and his se­nior as­so­ciate Nick Plugis dug deep­er in­to the con­cept, they found that most of the mod­els oth­ers have de­vel­oped take a bot­tom-up ap­proach, where they as­sem­ble the mol­e­cules in­side cells and the con­nec­tions be­tween them from scratch. What if they opt for a top-down ap­proach, aid­ed by sin­gle-cell tran­scrip­tomics and ma­chine learn­ing, to gauge the be­hav­ior of the en­tire cel­lu­lar net­work?

“If you look at cell be­hav­ior from the per­spec­tive of a mol­e­c­u­lar net­work un­der­ly­ing it, then you free your­self from the tra­di­tion­al ap­proach of one-di­men­sion­al, two-di­men­sion­al, three-di­men­sion­al tar­get-based or phe­no­typ­ic-based drug dis­cov­ery ap­proach­es,” Kahve­jian, who took on the CEO role, told End­points News. “What it al­lows you to do is to use the net­work changes as your read­out.”

Flag­ship ded­i­cat­ed $50 mil­lion to get the biotech start­ed, which is how Cel­lar­i­ty has been fund­ing the build­out of its plat­form and an­i­mal ex­per­i­ments to ver­i­fy their ini­tial hy­pothe­ses in the past two years.

By in­ter­twin­ing wet labs and a dig­i­tal twin dubbed the Cel­lar­i­um, Kahve­jian be­lieves his biotech hasn’t just “re-ar­chi­tect­ed” ther­a­peu­tic dis­cov­ery, but al­so the or­ga­ni­za­tion of an AI up­start. Chad Nus­baum, founder of the Broad Tech­nol­o­gy Labs, leads the tech­ni­cal unit churn­ing out da­ta; while Milind Kamkolkar has joined as chief dig­i­tal & da­ta of­fi­cer af­ter pi­o­neer­ing the role at Sanofi.

“I want­ed to build stuff. I didn’t want to just keep sourc­ing stuff,” Kamkolkar said of his de­ci­sion to leave the phar­ma gi­ant, where ex­ter­nal part­ner­ship was the pro­to­col for gain­ing dig­i­tal com­pe­ten­cy.

It’s the com­plete op­po­site at Cel­lar­i­ty, as they are build­ing a new en­gine that can be bro­ken down in­to three lay­ers. He calls the first “da­ta in­ges­tion” — chan­nel­ing all the in­for­ma­tion gen­er­at­ed by Nus­baum’s team with mul­ti­ple method­olo­gies and species in­to a data­base where sci­en­tists can plot and cu­rate knowl­edge. Then they en­ter the ex­plo­ration lay­er, in­ter­ro­gat­ing the cell be­hav­iors while an­a­lyz­ing how well ex­ist­ing and new com­pounds can per­turb the cells. On the last lay­er, they vi­su­al­ize the find­ings by cre­at­ing a satel­lite im­age of sorts.

Right now Cel­lar­i­ty has about 250 of these dig­i­tal guides on dif­fer­ent dis­eases, which they call Cel­lar­i­ty Maps. And they can en­com­pass every step of the tra­di­tion­al drug dis­cov­ery process.

“The ma­chines are in­cred­i­bly ca­pa­ble of par­al­leliz­ing and col­laps­ing what typ­i­cal­ly used to be a lin­ear process to try to un­der­stand whether the im­pact of that drug ac­tu­al­ly does have tox­i­c­i­ty or side ef­fects,” Kamkolkar added.

With 40 staffers on board, Cel­lar­i­ty has gone broad with its tech plat­form, prob­ing any­thing from ep­ithe­lial bar­ri­er dis­or­ders and on­col­o­gy to hema­to­log­i­cal dis­or­ders and neu­rol­o­gy. The plat­form can ac­com­mo­date mul­ti­ple ther­a­peu­tic modal­i­ties, Kahve­jian said, and they’ve test­ed both small and large mol­e­cules. He isn’t dis­clos­ing a time­line for when they might steer their lead can­di­dates in­to the clin­ic, but he’s not shy about the am­bi­tion to tack­le “dozens of pro­grams” at a time, and part­ner­ing as he sees fit.

As of Sep­tem­ber, Cristi­na Rondi­none, the for­mer head of car­dio­vas­cu­lar, re­nal and meta­bol­ic dis­eases at As­traZeneca, has al­so joined as pres­i­dent to help grow the com­pa­ny and push it to the next stage, en­abling down­stream clin­i­cal de­vel­op­ment of leads.

The new hires will find them­selves in a hor­i­zon­tal or­ga­ni­za­tion where no one do­main su­per­sedes the oth­er, Kahve­jian said, and where bi­ol­o­gists, tech­nol­o­gists, and the com­pu­ta­tion­al folks work to­geth­er in an in­te­grat­ed and mul­ti­lin­gual en­vi­ron­ment where in­sights are gen­er­at­ed more quick­ly and are “ac­tion­able the minute they are gen­er­at­ed.”

Kamkolkar re­called the sur­prise of a ma­chine learn­ing sci­en­tist when he found out that he was to spend time in labs and see how the da­ta are gen­er­at­ed.

“Yeah, you’re gonna have to go in labs,” Kamkolkar ba­si­cal­ly said. “It’s quite unique.”

Steve Holtzman was awoken by a 1 a.m. call from a doctor at Duke University asking if he could put some pigs on a plane and fly them from Ohio to North Carolina that day. A motorcyclist had gotten into a horrific crash, the doctor explained. He believed the pigs’ livers, sutured onto the patient’s skin like an external filter, might be able to tide the young man over until a donor liver became available.