What are the key takeaways from this GEN Podcasts episode?

Key insights include: **Model Complexity Staging:** Start with a deliberately simple mechanistic model to generate a fast go/no-go developability decision, then layer in downstream biology incrementally. Adding complexity too early wastes resources; the initial model should answer one question — is this drug feasible to develop — before expanding to best-in-class design parameters.; **CAR-T Parameter Sweeps:** Run tens of thousands to hundreds of thousands of simulations varying receptor density, binding affinity, CAR-T cell half-life, doubling time, cytokine release rates, and Treg-to-effector-T-cell ratios simultaneously. This identifies which one or two parameters most dramatically shift predicted efficacy and safety outcomes, directly guiding which molecules to advance to clinical candidacy.; **Model Update Cadence:** Update the mechanistic model two to three times across preclinical development — after in vitro potency assays, after mouse studies, and after xenograft studies. Each update uses parameter estimation to constrain unknowns, progressively narrowing uncertainty in human dose predictions before IND submission rather than accumulating all learning at once.

What did Mechanistic Modeling discuss on GEN Podcasts?

John Burke, president and CEO of Applied Biomath, explains how mechanistic PKPD modeling advances drug development by incorporating mechanism-of-action biology into mathematical simulations. The episode covers model iteration across preclinical stages, competitive benchmarking, CAR-T cell therapy design, and FDA Project Optimus dose optimization for oncology IND submissions. Key topics include: **Model Complexity Staging:** Start with a deliberately simple mechanistic model to generate a fast go/no-go developability decision, then layer in downstream biology incrementally. Adding complexity too early wastes resources; the initial model should answer one question — is this drug feasible to develop — before expanding to best-in-class design parameters.; **CAR-T Parameter Sweeps:** Run tens of thousands to hundreds of thousands of simulations varying receptor density, binding affinity, CAR-T cell half-life, doubling time, cytokine release rates, and Treg-to-effector-T-cell ratios simultaneously. This identifies which one or two parameters most dramatically shift predicted efficacy and safety outcomes, directly guiding which molecules to advance to clinical candidacy..

How long is this episode of GEN Podcasts?

This episode is 25 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

GEN Podcasts

Moving the Drug Pipeline Forward with Mechanistic Modeling

December 1, 2022

25 min episode · 2 min read

Mechanistic Modeling

Episode

25 min

Read time

2 min

Topics

Investing, Startups, Fundraising & VC

AI-Generated Summary

Published Mar 15, 2026

Key Takeaways

✓Model Complexity Staging: Start with a deliberately simple mechanistic model to generate a fast go/no-go developability decision, then layer in downstream biology incrementally. Adding complexity too early wastes resources; the initial model should answer one question — is this drug feasible to develop — before expanding to best-in-class design parameters.
✓CAR-T Parameter Sweeps: Run tens of thousands to hundreds of thousands of simulations varying receptor density, binding affinity, CAR-T cell half-life, doubling time, cytokine release rates, and Treg-to-effector-T-cell ratios simultaneously. This identifies which one or two parameters most dramatically shift predicted efficacy and safety outcomes, directly guiding which molecules to advance to clinical candidacy.
✓Model Update Cadence: Update the mechanistic model two to three times across preclinical development — after in vitro potency assays, after mouse studies, and after xenograft studies. Each update uses parameter estimation to constrain unknowns, progressively narrowing uncertainty in human dose predictions before IND submission rather than accumulating all learning at once.
✓Competitive Benchmarking in Silico: Populate the model with publicly available PK/PD data from competitor molecules to run head-to-head simulations before entering the clinic. This allows teams to assess whether their candidate is best-in-class against current therapies and projected future therapies, informing the go/no-go investment decision with quantitative competitive evidence.
✓FDA Project Optimus Alignment: Mechanistic PKPD modeling supports the FDA's Project Optimus dose-optimization initiative by simulating patient population variability — including tenfold target overexpression scenarios and target-mediated drug disposition — to justify a higher, still-safe starting dose. This is especially critical for gene and cell therapies where traditional allometric scaling methods have no established framework.

What It Covers

John Burke, president and CEO of Applied Biomath, explains how mechanistic PKPD modeling advances drug development by incorporating mechanism-of-action biology into mathematical simulations. The episode covers model iteration across preclinical stages, competitive benchmarking, CAR-T cell therapy design, and FDA Project Optimus dose optimization for oncology IND submissions.

Key Questions Answered

•Model Complexity Staging: Start with a deliberately simple mechanistic model to generate a fast go/no-go developability decision, then layer in downstream biology incrementally. Adding complexity too early wastes resources; the initial model should answer one question — is this drug feasible to develop — before expanding to best-in-class design parameters.
•CAR-T Parameter Sweeps: Run tens of thousands to hundreds of thousands of simulations varying receptor density, binding affinity, CAR-T cell half-life, doubling time, cytokine release rates, and Treg-to-effector-T-cell ratios simultaneously. This identifies which one or two parameters most dramatically shift predicted efficacy and safety outcomes, directly guiding which molecules to advance to clinical candidacy.
•Model Update Cadence: Update the mechanistic model two to three times across preclinical development — after in vitro potency assays, after mouse studies, and after xenograft studies. Each update uses parameter estimation to constrain unknowns, progressively narrowing uncertainty in human dose predictions before IND submission rather than accumulating all learning at once.
•Competitive Benchmarking in Silico: Populate the model with publicly available PK/PD data from competitor molecules to run head-to-head simulations before entering the clinic. This allows teams to assess whether their candidate is best-in-class against current therapies and projected future therapies, informing the go/no-go investment decision with quantitative competitive evidence.
•FDA Project Optimus Alignment: Mechanistic PKPD modeling supports the FDA's Project Optimus dose-optimization initiative by simulating patient population variability — including tenfold target overexpression scenarios and target-mediated drug disposition — to justify a higher, still-safe starting dose. This is especially critical for gene and cell therapies where traditional allometric scaling methods have no established framework.

Notable Moment

Burke argues that for gene and cell therapies, traditional allometric scaling is essentially unusable for human dose prediction, and that mechanistic modeling is the only scientifically defensible alternative — without it, he suggests, dose selection amounts to guessing rather than evidence-based prediction.

Know someone who'd find this useful?

Episode Transcript

Welcome to Gencast, a sponsored podcast series brought to you by Genetic Engineering and Biotechnology News. I'm your host, Jeff Bogaliskas. In part one of this three part Gencast series, I met up with doctor John Burke, who's the president, CEO, and cofounder of Applied Biomath, a company that is trying to help pharma and biotech organizations reduce their late stage attrition by applying advanced mathematical and systems biology techniques at critical decision points in the drug discovery process. In that initial podcast, John introduced us to the world of mechanistic modeling as it applies to drug discovery and development. In this latest gen cast, John and I continue the discussion we started in the first podcast and tackle the critical questions around mechanistic modeling and get into more of the nitty gritty of how the process works and how applied biomed helps their partners. Let's listen in. Hello, everyone, and welcome to this new episode of Gencast. As always, I'm your host, Jeff Boguliskas, technical editor for genetic engineering and biotechnology news, which has been at the vanguard of the life science industry for over forty years now, bringing you news and insights on the latest tools and technologies. And it's my pleasure today to be joined once again by the cofounder, president, and CEO of Applied Biomath, doctor John Burke. And we're gonna continue our discussion and deep dive into the world of mechanistic modeling. Welcome, John. Welcome. How are you doing today? I'm doing great. Thanks for joining us again. We really appreciate it. So, John, to kick things off, maybe as a refresher for those who have not yet listened to the first podcast, and if you haven't listened to it yet, go back and check it out. It's really good. Can you give sort of a high level overview of the mechanistic modeling, what it is and why it's important for drug r and d? A way to think about mechanistic modeling or mechanistic PKPD modeling or systems pharmacology is to think about PKPD maybe at the next level, where in traditional PKPD modeling, you might have very well defined curves that are fit to individual data, and then you may allometrically scale, from one species to another or look at concentrations. At at the next level with systems modeling, of course, we look at the the the real concentration data. We look at PD. But we incorporate mechanism of action of the drug and mechanism of action of the biology or target biology. So if you are, example, thinking about, a gene therapy that's delivered by a lipid nanoparticle, you would include, for example, in the mathematical model dosing in of the LNP. There might be, payload within the LNP. Maybe the LNP is targeted or not. And we would include the dynamics of dosing, distribution of the LNP to different tissues, binding to the tissue that it's it's intended to go to or maybe not intended to go to. And then, the payload, …

Get the full transcript (4,097 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all GEN Podcasts transcripts →