Short answer: Dr. Chadi Nabhan says 13 AI agents parse a protocol of up to 200 pages in seconds, covering title, schedule of events, biomarkers and endpoints. AI also infers site capabilities from prior studies and finds the best contact person.
Watch this answer from Episode 34 (starts at 12:40) · Episode 34 notes and transcript
From Episode 34: AI Is Rebuilding Clinical Trials, starting at 12:40.
Jordan Metzner [12:40]: That's awesome. I guess, I think kind of following up on what Sam has said, I'm sure a lot of like kind of site matching has like always been a problem in this circumstance. How has AI specifically I mean, you started to talk about like the ability to leverage like, the DNA markers, etcetera. But how has AI kind of like pivoted the business in the sense of like giving it like kind of acceleration or ability to kind of make these twins, and how is it mean, obviously previously it was done with like, I guess like SQL type data, right? You would like run these kind of data queries and try to find, some type of data information, but I guess with a larger data set, obviously AI becomes incredibly more powerful, but yeah, maybe how has AI made it like a really big impact?
Dr. Chadi Nabhan [13:20]: Yeah. I mean, as you said, Jordan, I mean, the mass of data just you can't really do this, any of this manual at this point. But a couple of things are really important just to explain maybe a use case. Some of the you know, there are ability there are AI agents that we actually have built that allow us to find the right contact individual for each particular downstream after you select the sites. But before that, the AI is able to infer a lot of the capabilities of a specific site based on prior studies that they have conducted. An example, if site A has conducted a phase one trial in whether it's a basket trial, umbrella trial, or one tumor trial, just because they've conducted a phase one trial and the trial has opened, We know that they can conduct pharmacogenomic studies, pharmacokinetic studies. We know that they're able to do x, y, and z. So the AI infers the information from prior experience and gives us the capabilities of this particular site. The other thing where we leveraged AI is reading the protocol. So when we upload a particular study that could be 200 page study in any therapeutic area, we have 13 AI agents that dissect and parse out, read the entire protocol.
Dr. Chadi Nabhan [14:50]: We have an AI agent that reads the title, an AI agent for the schedule events, an AI agent for the biomarker, an AI agent for the endpoints, and so on. So by parsing the protocol in a matter of seconds, by the time you upload the protocol and click on the AI and so on, it just within seconds, the AI read all of the protocol and what it requires. What does it actually require to what does this trial require? It requires a site that has an MRI machine, they did a CT scanner, all of these things. And then it matches that against digital twins. So reading the protocol is an example, inference is an example, and the third example is the AI agent that brings you the best contact person for this specific site for this specific trial.
Jordan Metzner [15:38]: That's awesome. This is the second or third, I think, healthcare focused AI enabled guest we've had on our podcast and in both in wildly different areas, but just showing how AI becomes like a super enabler even for industries that are obviously like, more traditional like healthcare. Sam, what do you think? We go to the news.
