Off-target protein binding is a major source of adverse effects for small-molecule drugs, yet most structure-based molecular design methods focus on generating selective compounds de novo rather than improving the selectivity of existing, well- characterized drugs. We introduce specificity optimization (SpecOpt), a molecular design task that seeks constrained structural modifications to an existing compound that increase its binding preference for an intended target over known off-targets while preserving its structural identity and drug-like properties. To enable systematic evaluation, we construct a ChEMBL-derived benchmark from compound-target interaction data, identifying intended targets through curated drug-mechanism annotations and off- targets through measured activities. We then develop an agentic framework that docks each compound against its intended target and off-targets, compares the resulting poses through residue-aware atom-protein contacts, and provides these differential interactions to a large language model to propose targeted structural modifications. Candidates are retained only if they satisfy molecular similarity, ADMET, and target-off-target docking selectivity criteria. On 915 compounds, the agent improves the target- off-target binding gap for 84.8% of compounds, shifting the mean gap from -0.72 to +0.47 kcal/mol while maintaining a mean Tanimoto similarity of 0.72 to the starting compounds. Ablation studies identify residue-specific contact information as the critical optimization signal: replacing residue identities with binary contact indicators eliminates improvement on all 29 ablation compounds. These results establish SpecOpt as a distinct molecular design problem and demonstrate residue-aware differential interactions as an effective signal for improving the specificity of existing compounds.
Figures & tables
Figure 1: Overview of SpecOpt. Given a compound, its intended target, and known off-targets, SpecOpt docks the compound against each protein, computes per-atom contact differences with residue identities, and uses an LLM agent to propose targeted molecular edits. Candidate molecules are evaluated using structural similarity, ADMET, and re-docking criteria, and accepted candidates are iteratively refined in subsequent rounds.
Figure 2: Differential interaction analysis and molecular editing in SpecOpt. Target- and off-target-bound poses provide atom-level contact differences and growth opportunities, which are aggregated into fragment-level editing suggestions to guide candidate molecule construction.
Baseline (original drug)
After optimization
Mean specificity gap (kcal/mol)
−0.718
+0.472
Median specificity gap (kcal/mol)
−0.600
+0.300
Compounds improved
—
776/915(84.8%)
Mean Tanimoto to original
1.000
0.720
Table 1: Main result over the full compound pool. Gap is defined in Eq. 1 ; positive means target-preferring. The paired test compares each compound’s own baseline against its own optimized result.
Figure 3: Main result across 915 compounds. (a) The specificity-gap distribution shifts from predominantly negative (off-target-preferring) to predominantly positive. (b) Per-compound improvement. (c) Structural conservation; the spike at 1.0 is compounds returned unchanged because no candidate passed the admission gates.
Figure 5: Selected SpecOpt case studies showing the original (left) and optimized (right) structures. Highlighted atoms indicate structural changes relative to the maximum common substructure of the original compound. Specificity gaps are reported in kcal/mol.
Compound
Target / off-target
Base
Opt.
Sim.
Edit
Edoxaban
factor Xa / prothrombin
−5.6
+1.5
0.76
F + CF 3 on cyclohexane
Ibipinabant
CB 1 / CB 2
−1.5
+5.1
0.73
F on arylsulfonyl ring
Brequinar
DHODH / DHFR
−4.9
+0.1
0.83
F repositioned on biaryl
Repotrectinib
TRKB / ROS1, JAK2, FAK1
−4.6
+0.2
0.73
nitrile on fluoroaryl
Berotralstat
kallikrein / plasminogen
−3.5
+1.9
0.78
NH →N -methyl
Table 2: Selected case studies from the main evaluation. “Edit” summarizes the net structural change relative to the original compound.
Figure 6: Ablation and signal analysis. (a) Mean change in specificity-gap improvement when each component is removed; positive values indicate that the component improves performance. Green indicates p<0.05 , orange p<0.1 , and grey no statistical significance. (b) Fragment-level interaction signals aggregated across the 915-compound evaluation. Exclusive target- or off-target-contact signals do not occur, while most informative signals arise from different residue identities across shared contacts and target-specific growth opportunities.
Relationship
n
Base
Opt.
Δ
% improved
Wilcoxon p
Isoform / subtype
149
−0.72
+0.48
+1.20
85.9%
9×10−23
Homologous (same class)
335
−0.43
+0.94
+1.38
83.3%
2×10−47
Different (no shared class)
431
−0.94
+0.10
+1.04
85.6%
3×10−62
Table 3: Performance grouped by the relationship between the intended target and the off-target defining the specificity gap. Performance is consistent across relationship types.
Appendix figures & tables7 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 7: Binding-score changes after optimization for FLT3-targeting compounds. Left: change in target binding, where negative values indicate stronger binding. Right: changes across five off-targets, where positive values (blue) indicate weaker binding. Five compounds weaken all five off-targets simultaneously, while the remaining compounds show mixed effects across off-targets.
Figure 8: Benchmark composition. (a) Number of off-targets per compound on a logarithmic axis, with a median of 3. (b) Clinical stage; approximately two thirds of compounds are approved drugs. (c) Target-class annotations. (d) Molecular weight versus QED for the starting compounds.
Protein
Endpoint
n
Spearman ρ
p
Muscarinic acetylcholine receptor M2
AC50
132
−0.574
6.3×10−13
Mu-type opioid receptor
AC50
119
−0.316
4.7×10−4
5-HT receptor 2B
Ki
99
−0.258
0.010
KCNH2 (hERG)
IC50
124
−0.237
0.008
5-HT receptor 2A
AC50
117
−0.229
0.013
Appendix
Table 4: The five proteins with the strongest docking–affinity agreement, after filtering to a single assay type and unit. Negative ρ is the expected direction (a more negative Vina score should accompany a higher p -activity). These are the strongest correlations among the 19 proteins tested; the median over all 19 is −0.075 .
Figure 9: Docking scores pooled into two sets over the whole benchmark: compound–protein pairs where the protein is the compound’s intended target ( n=1,670 ) and bindings where it is an off-target ( n=11,823 ). Density-normalized; dashed lines mark medians. The comparison shows no statistically significant difference and a small effect size: medians −8.20 vs. −8.30 kcal/mol, Mann–Whitney p=0.21 , Cohen’s d=0.02 , with quantiles agreeing to within 0.1 – 0.4 kcal/mol at the 5th, 25th, 75th and 95th percentiles.
Docking score
Experimental (pChEMBL)
Category
Vina / Kd
target
off-target
target
off-target
Strong
≤−9.0 ( ≲ 260 nM)
34.7%
34.5%
81.5%
27.7%
Moderate
−9.0 to −7.0 (260 nM–7.5 µM)
37.6%
40.2%
15.4%
64.1%
Weak
−7.0 to −5.0 (7.5–220 µM)
22.6%
20.4%
0 3.1%
0 8.2%
Negligible
>−5.0 ( ≳ 220 µM)
0 5.1%
0 4.9%
0 0.0%
0 0.0%
Strong or moderate
72.3%
74.7%
96.9%
91.8%
Appendix
Table 5: Binding-strength categories. Left: our docking scores. Right: the experimental affinities binned at the pChEMBL equivalents of the same cut points ( −9.0→6.59 , −7.0→5.13 , −5.0→3.66 ). Docking places intended targets and off-targets in nearly identical proportions; the experimental labels differ by approximately 54 percentage points in the strong category. The experimental “negligible” row is empty because off-targets enter the benchmark only at pChEMBL ≥5.0 .
Component removed
n
full
ablated
Δ
Wilcoxon p
Five-way contact diff (naive 3-way instead)
29
+0.903
+0.000
+0.903
<0.001
Within-compound memory (8-round budget)
29
+1.390
+0.786
+0.603
0.091
Combo mechanism
29
+0.903
+0.824
+0.079
0.082
Cross-compound memory
29
+0.903
+0.852
+0.052
0.648
Pose-consistency caution
17
+1.147
+1.141
+0.006
0.893
Within-compound memory (4-round budget)
28
+0.936
+1.046
−0.111
0.972
Appendix
Table 6: Paired ablation results. Δ is mean(full − ablated); positive means the component contributes.
Endpoint
Generated (top-20)
Real drugs (baseline)
Ours (optimized)
AMES (mutagenicity) ↓
0.539
0.288
0.178
hERG (cardiotoxicity) ↓
0.691
0.696
0.702
P-gp (efflux) ↓
0.349
0.315
0.290
DILI (liver injury) ↓
0.787
0.753
0.745
HIA (absorption) ↑
0.442
0.857
0.876
BBBP (BBB penetration)
0.387
0.680
0.619
Appendix
Table 7: ADMET comparison (mean predicted probability). Arrows indicate the preferred direction: ↓ lower is better; ↑ higher is better.
Molecular optimization computationally designs or edits chemical structures to improve desired properties while respecting chemical and practical constraints.
Lead optimization in structure-based drug design aims to improve target binding while avoiding unintended interactions with off-target pockets. However, existing affinity-driven methods do not explicitly control specificity, whereas current specificity-aware approaches commonly require prior knowledge of off-target structures. We address off-target-agnostic specificity-aware lead optimization by analyzing the geometric mismatch between a ligand and the target pocket. We provide a conservative specificity lower bound for geometrically separated off-targets without requiring access to off-target structures. By metricizing pocket--ligand mismatch, the triangle inequality shows that reducing target--ligand mismatch improves a conservative lower bound on mismatch to a separated off-target class, which can be translated into a specificity lower bound through an empirical geometry--affinity calibration. Motivated by this analysis, we introduce SurfSpec, an off-target-agnostic lead optimization framework that iteratively grows ligands toward under-occupied regions of the target pocket surface. SurfSpec alternates between linker generation toward selected target-surface patches, which provides geometric pseudo-labels, and refinement under a pocket-conditioned ligand prior, which restores these pseudo-labels into valid ligands. On the CrossDocked2020 test set, SurfSpec reduces geometric mismatch and outperforms evaluated off-target-agnostic lead optimization baselines in empirical specificity, while maintaining competitive target-affinity improvement.
Minyeong Hwang, Yoorim Gang, Ziseok Lee +5
Kim Jaechul Graduate School of AI, KAIST · Interdisciplinary Program in Artificial Intelligence, Seoul National University · Department of Biomedical Sciences, Seoul National University +3
Structure-based drug design increasingly employs LLM agents to iteratively refine ligands against a target pocket, yet a viable ligand must satisfy two often-conflicting objectives---binding affinity and druggability---which single optimization steps rarely improve together. To quantify this difficulty, we introduce two diagnostic metrics: the first measures how often a single edit improves both objectives, and the second measures how often a gain on one objective comes with a loss on the other. Applying these diagnostics to current LLM-agent pipelines exposes a consistent failure mode: the agent performs molecular editing without knowing how the pocket-ligand complex responds to local modifications, thus rarely achieving joint improvement. Inspired by medicinal chemists, who probe the pocket-ligand complex with controlled analog edits before choosing an optimization direction, we propose PROBE, an optimization framework built around edit--response probing. PROBE first decomposes the ligand into editable sites and builds a pocket-specific site map that flags where joint gains are plausible, where the two objectives are likely in tension, and where liability substructures should be changed; it then performs controlled probe edits whose responses are distilled into an EditManual. Guided by the site map and EditManual, PROBE runs an iterative multi-agent optimization loop. On the CrossDocked2020 benchmark, PROBE achieves state-of-the-art performance and substantially mitigates the failure modes exposed by our diagnostics metrics. Codes are in https://github.com/yzf-code/PROBE/tree/main
Zaifei Yang, Weiyu Chen, Yaqing Wang +1
The Hong Kong University of Science and Technology · City University of Hong Kong · Beijing Institute of Mathematical Sciences and Applications
Designing small molecule ligands that bind with high affinity to specific protein pockets is a fundamental goal in drug discovery, as small molecules constitute a major fraction of approved therapeutics. Recent breakthroughs in structure prediction, such as AlphaFold-3 and Boltz-2, enable accurate biomolecular interaction prediction and show promise as foundation models for downstream tasks, including binding affinity prediction. We propose to leverage these models and introduce DBMol, a new structure predictor-guided framework for de novo small molecule design. DBMol formulates an alternating optimization and projection process. In the optimization stage, DBMol starts from an initial molecule and uses gradient-based optimization to improve pocket-specific interactions and predicted binding affinity using a structure prediction model. In the projection stage, a flow-matching model maps the optimized molecular graph to discrete and chemically valid molecules. Experiments show that DBMol effectively optimizes the Boltz-2 affinity proxy and generates molecules with strong predicted affinity and specificity under Boltz-2 evaluation. To reduce self-confirmation bias, we further evaluate generated molecules using held-out metrics, including AF3-based evaluation. DBMol substantially improves pocket coverage while maintaining molecular diversity over unconditional generation, and is competitive under held-out metrics despite the absence of reference-ligand supervision. These results support the promise of structure prediction models as effective optimization signals for de novo molecular design.
Yiming Qin, Kai Yi, Miruna Cretu +3
EPFL, Switzerland · MRC-LMB, UK · University of Cambridge, UK