Insights: How Pfizer Is Using Data To Improve Medical Diagnoses
Dr. Morten Sogaard explains how Pfizer uses artificial intelligence to create better drugs. (May 2017)
CLIFTON LEAF: How does a company like Pfizer, or any drug maker, use artificial intelligence to help create better drugs, drugs that target the targets better? Just take it in the simplest terms if you could explain that. DR. MORTEN SOGAARD: Well, I could maybe give a couple of examples. One is kind of electronic biomarkers where particularly for neurological diseases, actually diagnosis one of the issues for us. Maybe you have 30% treatment response, 30% down response, and then you have 40% placebo response. So how do you design clinical trials? So having much more precise diagnosis, for example, having someone call in to a in a phone line, having the voice pitch recorded as a proxy of mood or depression. Yeah. So that's kind of one example. We have actually collaborated with IBM on Parkinson's electronic monitors where basically we characterize the phenotype much more precisely. I think another example I could give is deep learning. So that was I guess the technology pioneered by Google to identify cat faces as I'm sure you're all aware, right? So we use that for mining electronic health record and kind of associated lab data to understand disease trajectories for some of these chronic diseases like fatty liver disease or autoimmune disease to really understand what is likely to happen to two different patients. Of course, it's at the population level, but we also hope that it gives some productivity at that individual.