Learn more about how our pharmacometricians can help you de-risk your drug development program.
Transcript:
Hello, and welcome to Certara Talks. My name is doctor Suzanne Minton, and I’m a director of content strategy at Certara. Today, I’m talking to two of my colleagues, doctor Amy Cheung and doctor Mark Sale. Amy, would you like to introduce yourself?
Hi, everyone. I’m vice president of, Europe APAC QSS group lead, and, also, I’m leading for the Pediatric Maternal Health Center of Excellence.
Nice to meet you.
Nice to meet you. Welcome. Mark, could you introduce yourself, please?
Sure. Thanks, Suzanne. Mark Sale. Been with Certara a couple of years, and I’m now in the soft vice president in the software group, Formerly vice president in the, consulting group.
Excellent. Well, lovely to talk to both of you today. I heard that both of you went to the ASCPT annual meeting last month, so I’d love to chat with you both a little bit about your experiences.
Amy, I know that you presented a roundtable, a a lunch and learn meeting with one of our colleagues on cardiac safety support at Certara and how we can help with concentration QTC modeling for special populations. Could you tell us a little bit about how that went?
Yes. I would love to. So, indeed, Mark Sheldon and I have presented a really nice presentation, and it was very well subscribed in a lunchtime learn at ASCPT to allow us to share with everyone, what is our offering as a company. And, also, we have a bespoke as a tire cutting, safety assessment team that cover, many years of experience and supports a huge amount of submissions. And the teams covered include, clinical pharmacologists, toxicologists, cardiologists, and pharmacometricians as well.
With that sessions, we also managed to outline all the latest, regulatory landscape, the opportunity to help company to waive, thorough QT study using concentration QT analysis. And in terms of strengthened out, our idea, we presented a number of examples in oncology range from using the data from phase one to phase three. How can how can in those example that with a company collecting early data, including PK match, constant, PK match ECG data that are to support concentration klty and show that there is low QT prolongations, for the study drugs, and so that it provided evidence to waive the TQT. And those example also is in a public domain from our safety team as well. So it was really great.
Fantastic. Sounds like an like a great talk, Amy. And, Mark, I saw that you gave a talk, excellent title, called when AI meets drug development. Could you tell us about how that went?
Sure. It was a panel talk, and there was a number of topics. Our topic was around using AI machine learning for model selection for population PKPD models. And we have been working for a couple of years now under an FDA grant, and in collaboration with the FDA to develop a software package, an open source software package called PyDarwin that uses a number of machine learning algorithms to to search for for population PK models as an alternative to the usual approach of trying things and seeing if they work and then trying something else, much more efficient, much more, much faster, and we think more robust, more likely to find the best model. And, we we presented data comparing five machine learning algorithms, genetic algorithm, random forest, random tree with gradient boosting, particle swarm optimization, and Gaussian process.
And the central question was, do they find the best model? And we used a fairly large clinically relevant a real clinical database on a drug called DMAG.
And the gold standard that we compared all these two was actually running an exhaustive search, which given the number of models that we searched for, number of compartments and covariance and random effects and different absorption models for a total of, one point five million possible models. And we actually ran all of them in a with a collaborator at University at Buffalo.
And what we found was that all five algorithms, when combined with some other sort of traditional model selection automated, did in fact find the best model of all these one and a half million models. It found the best model after just just a couple of hundred models.
And very pleased to see that. Of those, the other issue that we looked at besides the robustness, whether it does find the best model, is whether it, is the efficiency. Which one found the the most quickly? And the answer to that was, it it depends, of course, for fairly simple models that have a fairly small search space, we call it number of models to look at. It looks like Gaussian process is the most efficient, but for most of ours sort of clinically relevant models where we’re looking at many, many different options is probably genetic algorithm.
And this package is available online. It’s called PyDarwin. If you search for PyDarwin, p y d a r w I n, you’ll find the the GitHub repo, and you can download it all open source, but Certara also has developed a number of companion applications. The the the standard interface to PyDAR one is is very complicated. It’s really difficult to use, more complicated even than none them itself. And so we’ve developed a couple of graphical user interfaces to to facilitate the use, and those are available in the the Piranha package, that that is available through Certara.
Fantastic, Mark. That sounds like a really, exciting use of how it AI can help pharmacometricians do their job more efficiently and and help find the best model.
Yeah. I think that both of those are important criteria. The best model certainly is important from a regulatory and a scientific perspective, but timelines are always important. And this is more efficient. It uses it basically replaces weeks of human effort with a lot of computer power, and computer power is cheap and readily available.
Yeah. That’s fantastic.
Amy, turning back to you.
I saw at the end of ASCPT, you gave a talk called the wheel of misfortune, global partnerships to tackle pediatric infectious diseases, and and that’s a fantastic name for a talk. Tell us how that went.
Yeah. It’s very interesting, title that we come up with, and then we try to also have a very interactive element built into it. So we built it as a game showed, with my cochair, who is, Doctor. Lanontan, who is from we’ve healthcare.
He’s the pediatric lead over there. And we have invited, various expert, in from industry in, HIV network and also from academia, and then divided them two teams. So we have Bessie McFarlane, Jennifer Kaiser, Melinda Watkins, and Cindy Brothers. So we divided the two teams, and then we created a wheel that resemble the wheel of the forge fortune of different questions that also created by, Certara with our great software engineer, James Craig, as well.
And each of the team will buzz to answer the question, then, give them five minutes to share with the audience what do they think. And the aim of the sessions, we really want to provide insights of how the r and d partnership’s helping with a particular population in pediatric age group, by some of studies for neonates, using PETID or impact, twenty twenty three network, and also including adolescent in phase three trials as well. And then another really important point we wanna outline is the Global Health Partnership, which, how can we accelerate novel medicine for future? And then one of our speaker have a lot of experience in Viv Chai genetic, for partnerships with company to of a particular products, and how can we accelerate that?
And the last one is how can we build the capability and knowledge in pediatric? For example, using different pediatric infectious network and then also incorporated innovative approach like, models informed drug development, to support the data. And then this also allowed us to give more thoughts and training, what information we need to provide and help to build infrastructure in low and middle income country to in to ensure that we can accelerate access as well. And all of this session is really aligned with our visions on how can we support, pediatrics, population and aligned really with our sense of excellence goals.
That’s fantastic. Well, pediatric drug development is such an important topic, and your passion for it is is so obvious. Well, Mark and Amy, it was wonderful talking to you both. Thank you. You’ve been watching Certara Talks. I’m Suzanne Minton. We’ll see you next time.