Modeling Selectivity with Low-Cost Binding-Affinity Methods

by Ari Wagen · Oct 28, 2026

I did the modeling for this post during our recent company-wide retreat in the Finger Lakes. We'll be sharing more from the retreat in the coming days.

There's always a lot of buzz in computational chemistry about binding-affinity prediction. For a target-based drug discovery program, on-target binding is a non-negotiable. Protein–ligand binding is also computationally tractable, so computational scientists have developed an abundance of methods to model it, ranging from simple QSAR models to compute-intensive approaches like free-energy perturbation (FEP). We've previously written about the range of binding-affinity predictors.

We've been paying particular attention to low-cost methods since launching our binding-affinity prediction workflow. We recently tested these methods on the Wang et al. 2015 JACS FEP benchmark, finding fair predictive accuracy.

I wanted to explore whether these low-cost methods could also predict trends in selectivity between two targets. Binding-affinity predictors are often evaluated on hit-finding or lead-optimization benchmarks. But if they capture the underlying physics accurately enough, they should also be able to predict differences in potency between targets in the same protein family.

Data

In a recent JMC paper, Yue et al. from Shenyang Pharmaceutical University and SunShine Innovation Co. in China reported a campaign to design selective, noncovalent JAK1 inhibitors. The paper reports IC50 values against both JAK1 and JAK2 for a series of 29 compounds, providing experimental potency and selectivity data to compare our predictions against. A few example compounds are shown below:

Representative JAK1 inhibitors from the JMC paper

Selectivity is expressed as the ratio of JAK2 IC50 to JAK1 IC50. Ratios greater than 1 indicate a preference for JAK1, with compound 37 showing particularly high selectivity.

Methods

I modeled these compounds using several approaches:

The paper was published in February 2026, well after the Boltz-2 training cutoff. However, Nesso-1 may have included these compounds in its training data.

Results

Both AutoDock Vina and Gnina were run with an exhaustiveness of 16, but Gnina produced more closely overlapping poses. I used the Gnina poses as starting structures for the UMA binding-affinity calculations.

Vina poses vs. Gnina poses

The table below shows Spearman correlations (ρ) across 29 compounds. The JAK1 and JAK2 columns compare predicted scores, oriented so higher values indicate stronger binding, with experimental pIC50 values. The selectivity column compares the difference in predicted scores between JAK1 and JAK2 with log10[IC50(JAK2)/IC50(JAK1)].

MethodJAK1 Spearman ρJAK2 Spearman ρSelectivity Spearman ρ
Vinardo−0.20+0.05+0.42
MM/GBSA on Vinardo+0.37+0.12+0.18
Gnina CNN affinity−0.14−0.21−0.19
MM/GBSA on Gnina+0.06+0.38+0.03
Boltz-2−0.01−0.36−0.06
Nesso-1+0.02−0.40−0.02
UMA SP+0.05+0.15−0.10
UMA opt → SP−0.02+0.14−0.06

No method consistently reproduces the experimental rankings across JAK1 potency, JAK2 potency, and selectivity. Vinardo gives the strongest selectivity correlation (ρ = 0.42), despite correlating poorly with potency against either target individually.

Overall, these correlations are much weaker than those we observed on the JACS FEP binding-affinity benchmark. There are several possible explanations worth investigating:

These are hypotheses, and the current results don't tell us which, if any, explains the difference.

Although these results are disappointing, we think it's important to be honest about where computational methods succeed and where they struggle. Finding those limits helps guide our development, and we hope this encourages others to evaluate selectivity alongside single-target affinity.

Banner background image

Ready to learn more?

Talk with a member of our team to learn how Rowan can accelerate your work. Security and compliance questions can go to security@rowansci.com, and our full security page covers current practices.

Contact us →

What to read next

Who will optimize the optimizers?

Who will optimize the optimizers?

reaching rock bottom; transformative work; minigames
Nov 12, 2026 · Jonathon Vandezande
Intro to Optimization - Mathified

Intro to Optimization - Mathified

Math and graphics on the Müller–Brown PES to introduce the optimization theory
Nov 12, 2026 · Jonathon Vandezande
Intro to Optimization - Gamified

Intro to Optimization - Gamified

Minigames on the Müller–Brown PES to introduce the difficulties experienced during optimization.
Nov 12, 2026 · Jonathon Vandezande
Returning from Retreat

Returning from Retreat

the fruits of our time in the Finger Lakes; a bountiful harvest of research miniblogs; manifold platform improvements
Nov 9, 2026 · Ari Wagen, Corin Wagen, Spencer Schneider, Zach Fried, Jonathon Vandezande, and Eli Mann
Do Boltz-2's Structure and Affinity Predictions Agree?

Do Boltz-2's Structure and Affinity Predictions Agree?

Comparing Boltz-2's initial and affinity-stage structures across 600 complexes, and testing how pose differences affect predicted affinity.
Nov 9, 2026 · Zachary Fried
Building a Frontier Lab for Chemistry: A Conversation With Hatem Helal

Building a Frontier Lab for Chemistry: A Conversation With Hatem Helal

A conversation with Hatem Helal, the founder of Potential Sciences.
Nov 6, 2026 · Corin Wagen
Modeling Carbonyl–Olefin Metathesis with Neural Network Potentials

Modeling Carbonyl–Olefin Metathesis with Neural Network Potentials

Reproducing a tricky mechanistic result from grad school with LLMs and NNPs.
Nov 2, 2026 · Corin Wagen
Do Co-folding Models Notice When You Change the Pocket?

Do Co-folding Models Notice When You Change the Pocket?

Running published pocket-mutation tests on six co-folding models in Rowan.
Nov 2, 2026 · Spencer Schneider
Making FEP Better

Making FEP Better

RESP charges, per-ligand custom forcefields, & cofactors in FEP; natural transition orbitals and transition densities; probing LLM capabilities in folding proteins; design for nighthawks
Oct 20, 2026 · Corin Wagen, Eli Mann, Jonathon Vandezande, Ari Wagen, and Ishaan Ganti
Persistent Hydration Sites and Water Exchange in Daraxonrasib Simulations

Persistent Hydration Sites and Water Exchange in Daraxonrasib Simulations

A site can stay occupied even as individual water molecules come and go.
Oct 19, 2026 · Ishaan Ganti