Michael Abraham (chemist)

Transcription

>> Up next from the University of Toronto is Abraham Heifets. He's doing a PhD in computer science, supervised by Professor Igor Jurisica. His presentation title is, "How can we make better medicines? Computer tools for chemistry." [ Applause ] [ Noises ] >> Abraham Heifets: I want to tell you a story about how medicine is made, that begins in the Pacific Northwest. So there's a tree that grows there called the Pacific Yew tree, you can see it in the back of my slide, and has a chemical in its bark called Taxol -- up there in the upper left, that we use as a chemotherapy for breast and ovarian cancers. But there's problem, the Pacific Yew tree is endangered. The Pacific Yew tree takes 300 years to grow, so you really can't farm it. And there's so little of this Taxol in the bark that you have to strip off the bark of 10 trees for one patient, which kills the tree. Now if we were all philosophers or politicians we'd have a really interesting debate about the value of trees verses the value of people. But fortunately we also go to a chemist and say, chemist, please build me this Taxol. And the only the thing that you have to know about chemist is that they think the world is made out of Lego's [background laughter]. So they'll say, well, look I don't have any Taxol, but if I had these two pieces then I know a reaction that would stick them together. Or maybe if I torn it apart a different way than a different reaction would stick those pieces together. And so now we've progress. The question is, just how do you build those pieces? And we do the same thing. We tear them apart, and we tear them apart, getting simpler and simpler until the pieces are so simple we can just buy them. But all ready it turns out that here we need computer help, there are over 14 million different chemicals that you can buy today. And I don't know about you but there are some mornings that I can barely remember 10 million different chemicals [background laughter]. On the other hand you go to Amazon and you look through 14 million books without a problem, right? So clearly we need computer help here. And in fact that the whole problem of computer planning looks a lot like how computers play chess. So the way that computers play chess is you look at the current board, and you say, what is every move that I can make? And for each of those what's every move that you can make, and what's every move I can make in response. And so again we have this expanding tree of possibilities, but where before we were looking for paths from Taxol to commercially available starting materials. Now we're looking for paths from the current board to checkmate. And it turns out that the best chess players in the world are no longer people, not since 1997 when IBM's Deep Blue supercomputer beat the world's chess champion. Computers have far surpassed human chess ability. So in my research I take those algorithms that worked fantastically well for chess and applied them to the domain of chemistry. People have made this analogy before, but no one had actually used the actual algorithms. And it turns out that chemistry is such a bigger problem, such a more complicated domain that the really critical question is. How efficiently can we search these tress? And so what I did is I put together the largest publically accessible database of synthesis and abstracted statistical lessons that helped guide that search. So when we make a decision, do we go left, do we go right. My work is about giving chemist better tools to make better medicines. Thank you. [ Applause ]