Media Releases

U of T team launches search for new Ebola treatments using artificial intelligence

November 5, 2014

TORONTO, ON — The Uni­ver­si­ty of Toron­to, Chema­tria and IBM are com­bin­ing forces in a quest to find new treat­ments for the Ebo­la virus.

Using a vir­tu­al research tech­nol­o­gy invent­ed by Chema­tria, a start­up housed at U of T’s Impact Cen­tre, the team will use soft­ware that learns and thinks like a human chemist to search for new med­i­cines. Run­ning on Canada’s most pow­er­ful super­com­put­er, the effort will sim­u­late and ana­lyze the effec­tive­ness of mil­lions of hypo­thet­i­cal drugs in just a mat­ter of weeks.

“What we are attempt­ing would have been con­sid­ered sci­ence fic­tion, until now,” says Abra­ham Heifets (PhD), a U of T grad­u­ate and the chief exec­u­tive offi­cer of Chema­tria. “We are going to explore the pos­si­ble effec­tive­ness of mil­lions of drugs, some­thing that used to take decades of phys­i­cal research and tens of mil­lions of dol­lars, in mere days with our tech­nol­o­gy.”

(Read the Mash­able sto­ry about Chema­tria.) (See the CTV news sto­ry.)

Chematria’s tech­nol­o­gy is a vir­tu­al drug dis­cov­ery plat­form based on the sci­ence of deep learn­ing neur­al net­works and has pre­vi­ous­ly been used for research on malar­ia, mul­ti­ple scle­ro­sis, C. dif­fi­cile, and leukemia.

Much like the soft­ware used to design air­planes and com­put­er chips in sim­u­la­tion, this new sys­tem can pre­dict the pos­si­ble effec­tive­ness of new med­i­cines, with­out cost­ly and time-con­sum­ing phys­i­cal syn­the­sis and test­ing. The sys­tem is dri­ven by a vir­tu­al brain that teach­es itself by “study­ing” mil­lions of dat­a­points about how drugs have worked in the past. With this vast knowl­edge, the soft­ware can apply the pat­terns it has learned to pre­dict the effec­tive­ness of hypo­thet­i­cal drugs, and sug­gest sur­pris­ing uses for exist­ing drugs, trans­form­ing the way med­i­cines are dis­cov­ered.

The World Health Orga­ni­za­tion has pro­ject­ed that new cas­es of Ebo­la could reach 10,000 each week by Decem­ber 2014, under­scor­ing the urgent need for research to address the cri­sis. The unprece­dent­ed speed and scale of this inves­ti­ga­tion is enabled by the unique strengths of the three part­ners: Chema­tria is offer­ing the core arti­fi­cial intel­li­gence tech­nol­o­gy that per­forms the drug research, U of T is con­tribut­ing bio­log­i­cal insights about Ebo­la that the sys­tem will use to search for new treat­ments and IBM is pro­vid­ing access to Canada’s fastest super­com­put­er, Blue Gene/Q.

“Our team is focus­ing on the mech­a­nism Ebo­la uses to latch on to the cells it infects,” said Dr. Jef­frey Lee of the Uni­ver­si­ty of Toron­to. “If we can inter­rupt that process with a new drug, it could pre­vent the virus from repli­cat­ing, and poten­tial­ly work against oth­er virus­es like Mar­burg and HIV that use the same mech­a­nism.”

While there are “broad spec­trum” antibi­otics that can treat mul­ti­ple kinds of bac­te­r­i­al infec­tions, most antivi­ral med­ica­tions are only effec­tive against a sin­gle kind of virus.

The ini­tia­tive may also demon­strate an alter­na­tive approach to high-speed med­ical research. While giv­ing drugs to patients will always require thor­ough clin­i­cal test­ing, zero­ing in on the best drug can­di­dates can take years using today’s most com­mon meth­ods. Crit­ics say this slow and pro­hib­i­tive­ly expen­sive process is one of the key rea­sons that find­ing treat­ments for rare and emerg­ing dis­eases is dif­fi­cult.

“If we can find promis­ing drug can­di­dates for Ebo­la using com­put­ers alone,” said Heifets, “it will be a mile­stone for how we devel­op cures.”

-30-

For more infor­ma­tion, please con­tact:

Michael Kennedy
Media Rela­tions
416–946-5025
m.kennedy@utoronto.ca