Using AI Created Digital Twins to Accelerate Clinical Trials

The Bio Report
|

Automatic TRANSCRIPT

Nature scientific reports peer reviewed Publication that now. It must have come out outlive well over a year ago and but we peer review takes very long time so that was written two years ago right even though it only got one year ago. Well basically what we do there is. We will take eighty percent of the train data. We have will train. The model will leave twenty percent of the data how to make predictions about these other. Twenty percents at the patients in making sure that all of the predictions that we make are really good. So that's one that's very early We've done much much more along those directions We've presented about our model presented data to fda on the office of neurosciences. We have a. We have done a number of retrospective studies where we can go back and look at previously completed clinical trials in reanalysed them. And make sure that when we're getting a better results out of those files Than than how they were originally run And then we also are working in ongoing prospective trials now In working with different pharmaceutical partners. Both in ways that are where our customers these collaborators. I getting value from the use of these models. But also that provide more validation off for our platform in the discussions with the fda. What kind of validation have been seeking with the fda. It's this is a really interesting area to to dig into I think that what we what we basically did with the fda's we showed them some data looking at we would take patients who were in cbo control arms of trials and then we would create digital twins of those patients that So now you have. The real patients receiving placebo in have the model predicting what would happen if they received placebo. So now you have a direct way to measure how well is the model doing it. Actually capturing for cbo behavior or these alzheimer's In so those are the kinds of data that we presented to the fda at a meeting in march of this year looking again at at leaving because we see so many things about these patients We have to really comprehensive Evaluation protocols evaluating all of the different things that were predicting But then the other thing of course is when you actually go to use these digital twins in clinical trials. You know that actually gets into another aspect of just the context of use because there are different ways that you can take these digital twins in incorporate them into the final analysis of the treatments affected in each one of those really discussed with regulators on a case by case basis because again adapting the use of the digital twins to the particular problem that the pharmaceutical facing the try year focusing on complex neurological diseases in particular alzheimer's disease. Why complex neurological diseases in general alzheimer's disease specifically why. I think the first thing comes down to an unmet need these are areas where clinical trials are very long in the very expensive. They included enormous numbers of patient. Volunteers it. We're not really having any success in developing new treatments so anything that we can do to make those trials more efficient to make them more ethical and better for the patients volunteer and to speed up drug development in those areas so that we can finally get affected therapies. The patients something that we really need. So that's that's the first thing. Is that the this large unmet. Need the second thing is there's availability of data as a machine learning company. We really relied. Not only on there being a lot of data but on those data being very high quality and because there's this long history of many many companies trying to develop drugs for these areas in many of those drugs failing there's an enormous amount of data that we can draw on to learn about how the disease progresses But we are are eventually looking to expand across disease area so even though our initial focus has been in these more complex longitudinal

Coming up next