The value of MBMA for determining best-in-class
Why Model-Based Meta-Analysis is one of the most useful tools for determining whether a drug pathway has hit its ceiling — and whether a best-in-class molecule is still achievable before you invest in optimization.
Before you optimize a molecule, ask: has this pathway hit its ceiling?
Model-Based Meta-Analysis (MBMA) is a modeling technique for running a head-to-head comparison among therapeutics with similar mechanisms of action in the same indication. It lets you assess the relationship between target engagement and clinical outcome — and determine what properties a molecule needs to hit your Target Product Profile.
I love doing MBMA. There's an efficiency and universality to it: multiple molecules' data get calibrated to a single modeling framework, so you're comparing apples to apples rather than squinting at separate publications. It's also genuinely challenging — sometimes the relationships aren't clean, and the modeling becomes part science, part art. Done well, it's one of the best tools for determining whether a pathway is saturated, and whether there's still room for a best-in-class molecule or optimized dosing — or whether that door is already closed.
Every company going after a clinically validated target needs to benchmark against the clinical competition. But a single head-to-head comparison only tells part of the story — an MBMA across several (or all) competitors gives you the full picture.
De-risk your drug development strategy through mechanistic modeling. Learn more at pharmath.io.
A couple of general references:
Chan et al. PMID: 35174432
Upreti & Venkatakrishnan. PMID: 30993679
Some of the MBMA work I've been involved in:
Johnson et al. https://doi.org/10.1002/psp4.70195.
Panday & Lang et al. https://doi.org/10.1002/cpt.3696
Kapitanov. doi: https://doi.org/10.1101/2021.03.07.21253086
Target occupancy — how you define it can make or break your drug
Why 100% target occupancy can be misleading for soluble targets — total target accumulation reflects binding, not the free-target reduction that actually drives therapeutic effect, and how the two can diverge based on complex vs. target half-life.
A drug can hit 100% “target occupancy” and still fail. Here's why.
For soluble targets at low concentrations (think cytokines), the usual PD assay is total target. If two ascending doses show no further accumulation, it's tempting to call the target saturated.
But when an antibody targets a cytokine, the antibody's half-life is usually longer than the target's — so binding "stabilizes" the target, and total target accumulates above baseline. That accumulation proves binding is happening. However, it doesn't prove free target has dropped enough for a therapeutic effect.
Target occupancy = bound target / total target. The saturated accumulation level (relative to baseline) is roughly set by the ratio of the drug-target complex half-life to the target's own half-life. If your complex's half-life is 100x the target's, occupancy approaches 100%, while telling you almost nothing about how much free target actually dropped. And free target reduction is the effect you're actually chasing.
So what does this mean for clinical strategy? It depends on whether a free-target assay is even feasible — post-dose, free target may be too low to measure reliably. Total target is still useful: with the right PK/PD model, it can reveal real information about the target. But it isn't sufficient on its own to determine the right dose or the right outcome.
Free target reduction is the outcome that matters.
Some references in the first post below
De-risk your drug development strategy through mechanistic modeling. Learn more at pharmath.io.
#PKPD #RO #DrugDevelopment #Biologics #TMDD
A non-exhaustive list of references for a deeper dive:
Stein, A. and Ramakrishna, R. https://doi.org/10.1002/psp4.12169
Dua P, Hawkins E, van der Graaf PH. doi: 10.1002/psp4.41
Kapitanov et al., doi: 10.3389/fbinf.2021.731340 — "Right Dose" section.
Agoram BM. doi: 10.1111/j.1365-2125.2008.03297
Lowe, P.J., et al. https://doi.org/10.1111/j.1742-7843.2009.00513.x
Chimalakonda AP et al. doi: 10.1208/s12248-013-9477-3.
Davda JP, Hansen RJ. doi: 10.4161/mabs.2.5.12833.
Mager DE, Jusko WJ. doi: 10.1023/a:1014414520282.
Kd ≠ IC50
Why Kd and IC50 aren't interchangeable — a real case where two molecules looked equivalent in one binding assay and 3x different in another, and how target concentration relative to Kd explains why.
A mistake I've seen more than once: Kd ≠ IC50
A well-known concept to most, but one that gets confused in practice — often leading to the wrong interpretation of an in vitro assay result.
The story: in an SPR assay, molecule A was 3x more potent than molecule B. The engineering team decided to affinity-optimize molecule B for use as the clinical candidate. But in a concentration-based binding assay (ELISA), molecules A and B looked equivalent. The team was confused — why keep optimizing B if they're the same? It turned out the target concentration in the assay was so high that affinity no longer mattered — both molecules simply hit IC50 at half the target concentration. After adjusting the assay's target concentration for the next run, the team moved forward with further affinity maturation on molecule B.
Binding assays need to reflect the affinity of the molecule — which means the target concentration needs to be in range of the KD. If target concentration >> KD, you fall into a concentration-dependent regime that no longer reflects true affinity.
Of course, there are plenty of caveats: mono- vs. multivalent binding, whether the recombinant SPR reagent reflects the natural target (which can vary by company and batch), avidity effects, and more. But whenever I have both SPR data and data from another assay (often on-cell binding), I do at least a back-of-envelope check to see if the results align — and whether we're in a concentration-dependent regime for the binding assay.
This matters in the clinic as well — the inherent target concentration and how it relates to the therapeutic's affinity can both determine the "danger" of TMDD as well as whether affinity maturation is necessary.
What am I missing? Do you have a preferred binding assay for in vivo translation?
De-risk your drug development strategy through mechanistic modeling. Learn more at pharmath.io.
#PKPD #affinitymaturation #KD #IC50 #DrugDiscovery #DrugDevelopment #TMDD
Translating biologics PK from preclinical species to humans
A practical breakdown of the approaches for translating biologics PK from preclinical species to humans — from full PBPK to minimal and site-of-action models — and how to pick the right one for your program.
Clients often need to translate biologics PK from preclinical species to humans. The right method depends on the modeling needs of the program.
If full PBPK is required, for mAbs the standard in my practice remains the Shah & Betts PBPK [1] or the Jones et al. paper [2]. More broadly, anything from Shah's lab over the past 15+ years is gold for understanding biologics distribution. For translation, you can take the rates and volumes from the original papers, calibrate against available data, and apply allometric scaling to translate to human.
When a full PBPK model is overkill, several minimal PBPK models can be applied depending on the situation — [3], [4], [5] (and many others!). These models offer the advantage of physiological detail, but the disadvantage is that they don't easily convert to a typical two-compartment framework — which can limit intuition when discussing with a classically trained pharmacometrician.
Sometimes you only care about what happens in a specific organ. In that case, an ultra-minimal PBPK model focused on the site of action may be preferred. I was fortunate to be involved in publishing two such models — SoA [6] and piPK [7]. These still describe the compartment of interest physiologically, while using parameters from classical two-compartment PK models.
For the classical two-compartment model itself, standard allometric scaling applies — Betts et al. is a useful reference here: [8].
The right approach depends heavily on your program's specific needs.
What am I missing? Do you have a go-to method or publication for this?
References in the first comment below 👇
#QSP #Pharmacometrics #PBPK #DrugDevelopment #Biologics
De-risk your drug development strategy through mechanistic modeling. Learn more at pharmath.io.
References:
[1] Shah DK, Betts AM. J Pharmacokinet Pharmacodyn. 2012. doi: 10.1007/s10928-011-9232-2
[2] Jones HM et al. CPT Pharmacometrics Syst Pharmacol. 2019. doi: 10.1002/psp4.12461
[3] Cao et al. J Pharmacokinet Pharmacodyn. 2013. doi: 10.1007/s10928-013-9332-2
[4] Bloomingdale P et al. J Pharmacokinet Pharmacodyn. 2021. doi: 10.1007/s10928-021-09776-7
[5] Spinosa P et al. CPT Pharmacometrics Syst Pharmacol. 2026. doi: 10.1002/psp4.70167
[6] Kapitanov GI et al. Front Bioinform. 2021. doi: 10.3389/fbinf.2021.731340
[7] Kapitanov GI et al. CPT Pharmacometrics Syst Pharmacol. 2026. doi: 10.1002/psp4.70160
[8] Betts A et al. MAbs. 2018. doi: 10.1080/19420862.2018.1462429.