cs.LGOct 19, 2026

HierDTA: Hierarchical Structure-Aware Representation Learning for Drug--Target Affinity Prediction

Authors: Huangchao Xu, Baohua Zhang, Qian Liu, Zhong Jin, Yi Du

Organizations: Computer Network Information Center, Chinese Academy of Sciences, Beijing 100083, China · School of Computer Science and Technology, University of Chinese Academy of Sciences, Beijing 101408, China

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Drug--target affinity (DTA) prediction plays an important role in computer-aided drug discovery, where reliable computational models can help prioritize compound--protein pairs before experimental validation. A key challenge in DTA prediction is to effectively mine multi-level molecular information from drug structures while accurately modeling the binding-site environment of target proteins. To address these issues, we introduce HierDTA, a hierarchical structure-aware framework for DTA prediction. \textcolor{black}{HierDTA represents each drug as an atom--motif heterogeneous graph and introduces an augmentation module that propagates functional-motif semantics to the corresponding atoms. For proteins, kinase-domain structures are converted into residue interaction graphs. The resulting atom and residue representations are integrated using cross-attention to capture pair-specific relationships, while molecule fingerprints and pretrained ESM-2 embeddings provide complementary global information. Compared with state-of-the-art methods, HierDTA achieves superior performance on the Davis dataset across all metrics and attains the highest rm2r_m^2 of 0.791 on KIBA. It also exhibits strong generalization under cold-start settings, particularly on KIBA, where MSE is reduced by 21.7% in the unseen-protein scenario. Ablation study and interpretability analysis supported the contributions of the hierarchical augmentation and interaction components.} Case studies further showed that HierDTA provided early enrichment of candidate bioactive compounds. These results indicate that HierDTA is an effective framework for affinity prediction and target-specific virtual screening.

Figures & tables

Figure 1: Architecture of the proposed framework.(A) Encoders. Drug features are encoded from motif and atom graphs, with motif-derived context augmenting atom representations, while protein residue features are extracted from the kinase-domain structure. (B) Interaction. Atom and residue tokens are integrated through multi-head attention to capture fine-grained drug–protein interactions. (C) Global Representation. ECFP fingerprints and ESM-2 embeddings are projected and fused to obtain a global representation. (D) Prediction. Local and global interaction features are concatenated and fed into an MLP to predict binding affinity.
NameSize
Atom node features37
atom type (B, Br, C, Ca, Cl, F, H, I, N, Na, O, P, S)13
atomic number1
explicit valence (0–6)7
total hydrogen count (0–4)5
atom hybridization (SP, SP 2 , SP 3 , SP 3 D, SP 3 D 2 )5
Table 1: Feature definition of the hierarchical molecular graph for drug representation.
NameSize
Residue node features41
solvent-accessible surface area (normalized)1
phi angle ( ϕ\phi )1
psi angle ( ψ\psi )1
one-hot encoded secondary structure6
AAPHY7 physicochemical descriptors7
Table 2: Feature definition of the residue interaction graph for protein representation.
Figure 2: Architecture of the motif-to-atom augmentation module. Motif representations are propagated to their constituent atoms through atom–motif membership edges and aggregated to obtain motif-derived semantic messages. For each atom, TLA operation uses the incoming atom representation as the query and attends to the motif context and atom message. The resulting augmentation message is incorporated into the atom representation through a residual connection before subsequent atom-level message passing.
ModelDavisKIBA
MSE ↓\downarrowCI ↑\uparrowrm2↑r_{m}^{2}\uparrowMSE ↓\downarrowCI ↑\uparrowrm2↑r_{m}^{2}\uparrow
DeepDTA0.284±0.0080.284\pm 0.0080.869±0.0020.869\pm 0.0020.624±0.0150.624\pm 0.0150.201±0.0030.201\pm 0.0030.850±0.0020.850\pm 0.0020.671±0.0170.671\pm 0.017
GraphDTA0.242±0.0100.242\pm 0.0100.880±0.0010.880\pm 0.0010.683±0.0100.683\pm 0.0100.192±0.0010.192\pm 0.0010.848±0.0020.848\pm 0.0020.718±0.0040.718\pm 0.004
FusionDTA0.226±0.0050.226\pm 0.0050.891±0.0010.891\pm 0.0010.686±0.0200.686\pm 0.0200.155±0.0020.155\pm 0.0020.882±0.0010.882\pm 0.0010.750±0.0150.750\pm 0.015
MGraphDTA0.228±0.0040.228\pm 0.0040.883±0.0020.883\pm 0.0020.679±0.0220.679\pm 0.0220.152±0.0030.152\pm 0.0030.884±0.0020.884\pm 0.0020.766±0.0160.766\pm 0.016
MSGNN-DTA0.234±0.0050.234\pm 0.0050.890±0.0020.890\pm 0.0020.686±0.0140.686\pm 0.0140.149±0.0030.149\pm 0.0030.885±0.0020.885\pm 0.0020.763±0.0090.763\pm 0.009
Table 3: Performance comparison under the warm setting on the Davis and KIBA datasets. The best result for each metric is shown in bold, and the second-best result is underlined.
Figure 3: Scatter plots of predicted vs true binding affinities on the Davis and KIBA test sets under the warm setting. Redder points indicate larger prediction errors, while blackr points indicate smaller errors.
SettingModelMSE ↓\downarrowCI ↑\uparrowrm2↑\boldsymbol{r_{m}^{2}}\uparrow
Unseen drugGraphDTA0.920±0.0290.920\pm 0.0290.678±0.0360.678\pm 0.0360.160±0.0190.160\pm 0.019
FusionDTA0.581±0.0940.581\pm 0.0940.737±0.0120.737\pm 0.0120.187±0.0340.187\pm 0.034
MGraphDTA0.563±0.0650.563\pm 0.0650.729±0.0220.729\pm 0.0220.192±0.0210.192\pm 0.021
MSGNN-DTA0.560±0.0750.560\pm 0.0750.731±0.0360.731\pm 0.0360.186±0.0580.186\pm 0.058
GRA-DTA0.578±0.0050.578\pm 0.0050.725±0.0150.725\pm 0.0150.121±0.0240.121\pm 0.024
DMFF-DTA0.548±0.063\mathbf{0.548\pm 0.063}0.742±0.037\mathbf{0.742\pm 0.037}0.228±0.0730.228\pm 0.073
Table 4: Performance comparison under cold-start settings on the Davis dataset. The best results for each metric are shown in bold.
SettingModelMSE ↓\downarrowCI ↑\uparrowrm2↑\boldsymbol{r_{m}^{2}}\uparrow
Unseen drugGraphDTA0.471±0.0470.471\pm 0.0470.713±0.0020.713\pm 0.0020.342±0.0070.342\pm 0.007
FusionDTA0.429±0.0310.429\pm 0.0310.748±0.0050.748\pm 0.0050.364±0.0120.364\pm 0.012
MGraphDTA0.425±0.0470.425\pm 0.0470.746±0.0020.746\pm 0.0020.366±0.0160.366\pm 0.016
MSGNN-DTA0.426±0.0430.426\pm 0.0430.747±0.0030.747\pm 0.0030.358±0.0220.358\pm 0.022
GRA-DTA0.337±0.0070.337\pm 0.0070.765±0.0120.765\pm 0.0120.466±0.0240.466\pm 0.024
DMFF-DTA0.408±0.0360.408\pm 0.0360.753±0.0090.753\pm 0.0090.397±0.0200.397\pm 0.020
Table 5: Performance comparison under cold-start settings on the KIBA dataset. The best results for each metric are shown in bold.
ConfigurationMSE ↓\downarrowCI ↑\uparrowrm2↑\boldsymbol{r_{m}^{2}}\uparrow
(A) Drug graph encoder
GAT0.174±0.016\mathbf{0.174\pm 0.016}0.905±0.004\mathbf{0.905\pm 0.004}0.757±0.026\mathbf{0.757\pm 0.026}
GCN0.183±0.0290.183\pm 0.0290.899±0.0050.899\pm 0.0050.744±0.0400.744\pm 0.040
GIN0.184±0.0220.184\pm 0.0220.903±0.0050.903\pm 0.0050.756±0.0270.756\pm 0.027
(B) Component
HierDTA0.174±0.016\mathbf{0.174\pm 0.016}0.905±0.004\mathbf{0.905\pm 0.004}0.757±0.026\mathbf{0.757\pm 0.026}
Table 6: Ablation results on the Davis dataset. The best result within each panel is shown in bold.
Figure 4: Comparison of the TLA and No TLA operation on the Davis dataset across the four evaluation settings: (a) MSE, (b) CI, and (c) rm2r_{m}^{2} . Lower MSE and higher CI and rm2r_{m}^{2} values are better.
Figure 5: Interpretability analysis of drug-protein complexes. (a) 1KV2 complex: The left panel shows the experimentally resolved 3D binding conformation, with key residues assigned high attention weights by HierDTA labeled. The right panel shows the atom-level attention heatmap over the ligand (darker colors indicate higher weights). (b) 1UWH complex: All subpanels have the same meanings as those in (a).
Figure 6: Motif attention weights from the first and second TLA operations for 1KV2 ligand. Values closer to one indicate greater reliance on the motif-derived context, whereas values closer to zero indicate greater reliance on the atom structural message.
RankCompoundPredicted KIBAVina Score (kcal/mol)
1Dacomitinib13.912−10.2-10.2
2Canertinib13.801−9.2-9.2
3JAK3-inhibitor-I13.792−7.9-7.9
4Afatinib13.786−8.9-8.9
5Poziotinib13.755−9.5-9.5
6Gefitinib13.696−8.6-8.6
Table 7: Top-10 unique DrugBank compounds ranked by HierDTA for EGFR (PDB ID: 3POZ). Boldface denotes established EGFR inhibitors.
Figure 7: Top- kk active compound hit rates of HierDTA and MSGNN-DTA on the external ChEMBL evaluation sets for (a) EGFR and (b) VEGFR2.
Figure 8: Table of Contents
More in this topic

Protein-Ligand Binding Affinity Prediction

Protein-ligand binding affinity prediction estimates how strongly a small molecule binds a protein to help prioritize compounds for drug discovery.

See all 25 papers

Explore similar work

CardsList
Sep 2, 2026

Chemical and geometric representation fidelity improves drug--target affinity prediction

Predicting drug--target binding affinity (DTA) requires models to distinguish subtle chemical and structural determinants underlying molecular recognition. Although recent approaches increasingly incorporate richer drug and protein information, such information may be compressed, homogenized or discretized during representation construction, causing affinity-relevant distinctions to be lost before interaction modelling. We hypothesized that this representation-stage information loss constitutes an upstream bottleneck that cannot be reliably overcome by increasingly complex interaction predictors. To test this hypothesis, we developed ReGeoDTA, a representation-preserving framework that maintains affinity-relevant chemical heterogeneity in molecular representations and continuous geometric relationships in protein structures. Across three benchmark datasets, ReGeoDTA consistently improved affinity prediction, and the proposed representation-preserving strategies retained their benefits across diverse DTA architectures. Controlled representation degradation progressively reduced predictive performance, whereas increasing downstream predictor complexity failed to recover information lost during representation construction. These findings identify representation fidelity as an upstream design principle for accurate and generalizable drug--target affinity prediction, with potential implications for computational compound prioritization.

Yixiao Li, Yining Qian, Yefan Chen +5

College of Information Science and Engineering, Northeastern University, Shenyang 110819, China · School of Computer and Communication Engineering, Northeastern University at Qinhuangdao, Qinhuangdao 066004, China · School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China +1

Privacy-Preserving Representation LearningProtein-Ligand Binding Affinity PredictionRepresentation LearningMolecular Representation LearningGeometric Representation LearningProtein Representation Learning
AskPDF
Mar 16, 2026

LaPro-DTA: Latent Dual-View Drug Representations and Salient Protein Feature Extraction for Generalizable Drug--Target Affinity Prediction

Drug--target affinity prediction is pivotal for accelerating drug discovery, yet existing methods suffer from significant performance degradation in realistic cold-start scenarios (unseen drugs/targets/pairs), primarily driven by overfitting to training instances and information loss from irrelevant target sequences. In this paper, we propose LaPro-DTA, a framework designed to achieve robust and generalizable DTA prediction. To tackle overfitting, we devise a latent dual-view drug representation mechanism. It synergizes an instance-level view to capture fine-grained substructures with stochastic perturbation and a distribution-level view to distill generalized chemical scaffolds via semantic remapping, thereby enforcing the model to learn transferable structural rules rather than memorizing specific samples. To mitigate information loss, we introduce a salient protein feature extraction strategy using pattern-aware top-kk pooling, which effectively filters background noise and isolates high-response bioactive regions. Furthermore, a cross-view multi-head attention mechanism fuses these purified features to model comprehensive interactions. Extensive experiments on benchmark datasets demonstrate that LaPro-DTA significantly outperforms state-of-the-art methods, achieving an 8% MSE reduction on the Davis dataset in the challenging unseen-drug setting, while offering interpretable insights into binding mechanisms.

Zihan Dun, Liuyi Xu, An-Yang Lu +2

College of Information Science and Engineering, Northeastern University, Shenyang 110819, China · School of Artificial Intelligence, Beihang University, Beijing 100000, China · School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China

Protein-Ligand Binding Affinity PredictionMolecular Representation LearningProtein Representation Learning
AskPDF
Mar 11, 2026

Co-Diffusion: An Affinity-Aware Two-Stage Latent Diffusion Framework for Generalizable Drug-Target Affinity Prediction

Predicting drug-target affinity is fundamental to virtual screening and lead optimization. However, existing deep models often suffer from representation collapse in stringent cold-start regimes, where the scarcity of labels and domain shifts prevent the learning of transferable pharmacophores and binding motifs. In this paper, we propose Co-Diffusion, a novel affinity-aware framework that redefines DTA prediction as a constrained latent denoising process to enhance generalization. Co-Diffusion employs a two-stage paradigm: Stage I establishes an affinity-steered latent manifold by aligning drug and target embeddings under an explicit supervised objective, ensuring that the latent space reflects the intrinsic binding landscape. Stage II introduces modality-specific latent diffusion as a stochastic perturb-and-denoise regularizer, forcing the model to recover consistent affinity semantics from noisy structural representations. This approach effectively mitigates the reconstruction-regression conflict common in generative DTA models. Theoretically, we show that Co-Diffusion maximizes a variational lower bound on the joint likelihood of drug structures, protein sequences, and binding strength. Extensive experiments across multiple benchmarks demonstrate that Co-Diffusion significantly outperforms state-of-the-art baselines, particularly yielding superior zero-shot generalization on unseen molecular scaffolds and novel protein families-paving a robust path for in silico drug prioritization in unexplored chemical spaces.

Yining Qian, Pengjie Wang, Yixiao Li +4

School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China · College of Information Science and Engineering, Northeastern University, Shenyang 110819, China · Westlake University, Hangzhou 310000, China +2

Drug DiscoveryLatent Diffusion Models
AskPDF