Drug-target interaction prediction sits at the heart of modern drug repurposing and target identification. When a protein emerges as a promising therapeutic target, perhaps because of new genetic evidence linking it to disease, researchers urgently need to know which existing or candidate compounds might bind to it. Experimentally screening every possible compound against every possible target is prohibitively expensive, so computational models step in to prioritize the most likely interactions. The difficulty is that the most clinically interesting cases are precisely the ones where data is scarcest. A protein with no known ligands offers nothing for a model trained on interaction labels to latch onto, forcing the system to rely entirely on intrinsic properties: the three-dimensional arrangement of atoms in a drug molecule, the amino acid sequence of a protein, and the structural patterns that connect molecular architecture to binding behavior.

The central insight behind HierHGT-DTI is that biological entities naturally exist at multiple scales simultaneously, and that most existing models fail to exploit this hierarchy coherently. A drug molecule can be described at the level of individual atoms, at the level of functional substructures such as rings and side chains, and at the level of the whole compound. A protein can be described residue by residue, through sequence-derived communities of residues that capture intermediate-scale organization, and as a complete macromolecule. Rather than encoding each scale separately and hoping a downstream classifier stitches them together, the new framework represents each candidate drug-protein pair as a bilateral multiscale graph. This graph contains atom nodes, substructure nodes, whole-drug nodes, residue nodes, residue-community nodes, and whole-protein nodes, all connected within one typed structure that spans the fine details of chemistry and the broad sweep of protein organization.

On top of this multiscale scaffold sits a heterogeneous graph transformer, a neural architecture built to reason over graphs whose nodes come in different types and whose edges carry different meanings. The transformer coordinates several distinct kinds of relations within the graph: local relations that connect neighboring entities at the same scale, adjacent-scale relations that link atoms to substructures and residues to residue communities, direct fine-to-global relations that jump from individual atoms all the way to whole proteins, and drug-protein relations that bridge the two sides of the pair. This relation-aware attention mechanism allows the model to learn which pathways of evidence matter most for a given prediction, effectively letting an atom-level chemical detail inform a protein-level judgment through a chain of typed, learnable connections rather than through hand-engineered features.

To assess how well this design works, the researchers benchmarked HierHGT-DTI against six matched baseline models across three widely used drug-target interaction datasets: DrugBank, BioSNAP and BindingDB. Crucially, the evaluation was repeated across five random seeds, a practice that guards against the possibility that a single favorable data split produced flattering results. The evaluation settings included random splits, where training and test interactions are drawn from the same distribution, as well as cold-drug and cold-protein splits, where entire molecules or proteins are held out of training entirely. The cold-protein setting is widely regarded as the most demanding and the most clinically relevant, because it directly simulates the scenario of a newly implicated target arriving with no ligand history.

The results were striking. HierHGT-DTI ranked first in all six DrugBank and BioSNAP evaluation settings, covering both random and cold-start conditions on both datasets. The largest margins appeared exactly where the problem is hardest: under cold-protein evaluation, the model’s AUROC, a measure of how well the model separates true interactions from non-interactions across all classification thresholds, improved by 8.8 percentage points on DrugBank and 5.8 percentage points on BioSNAP compared with the strongest baseline. In a field where models often compete over fractions of a percentage point, gains of this magnitude under cold-start conditions represent a substantial advance, and they suggest that the multiscale, relation-aware architecture is genuinely capturing transferable structure rather than memorizing interaction patterns.

The picture on BindingDB was more nuanced, and this honesty about mixed results adds credibility to the overall claims. BindingDB differs from the other two benchmarks in that it focuses on quantitative binding affinities rather than simple binary interactions, and its chemical space is broader and more skewed. On this dataset, HierHGT-DTI achieved the highest AUPR, or area under the precision-recall curve, in the cold-protein setting, a metric that is particularly informative when positive interactions are rare and the cost of false leads is high. However, the baseline model HiGraphDTI led the random and cold-drug evaluations on BindingDB. This pattern indicates that the new framework’s advantage is concentrated precisely in the cold-protein regime, which the authors identify as the setting most relevant for prioritizing candidate interactions involving previously unseen protein targets.

Why should coordinating multiple scales matter so much for cold-start generalization? The likely explanation lies in how information flows when interaction labels are absent. A model that only sees whole-molecule representations must somehow compress all binding-relevant chemistry into a single vector, and a model that only sees atoms must rediscover high-level patterns from scratch for every new protein. By building explicit pathways from atoms through substructures to whole drugs, and from residues through residue communities to whole proteins, HierHGT-DTI gives the transformer a rich menu of abstraction levels to draw from. When a completely new protein appears, the model can still recognize familiar residue-community motifs, recurring substructural chemistry on the drug side, and cross-scale relations that survived training on other targets. In effect, the architecture provides many partial routes to a prediction, so the absence of interaction labels for one specific protein does not sever the entire chain of evidence.

The practical implications extend across several areas of pharmaceutical research. For drug repurposing, a model that performs well on cold proteins can scan established compound libraries against emerging targets from genetic studies, flagging candidates for experimental validation before costly assays begin. For structure-based drug design, the multiscale representation offers a way to connect atomic-level binding determinants with protein-level function. The authors also note that the framework is relevant to phenotypic drug screening and molecular target identification, both of which depend on reliably linking compounds to the proteins through which they act. Because the model transfers structural and sequence evidence rather than relying on interaction history, it is naturally suited to poorly characterized proteins, including those produced by understudied genes that the scientific community has only recently begun to annotate.

Transparency accompanies the methodological advance. The study was supported by the Hunan Provincial Natural Science Foundation Key Joint Project Between Province and City, and the authors declare no competing interests. Importantly for the research community, the source code, processed data splits and configuration files are publicly available, allowing other groups to reproduce the benchmarks, apply the framework to their own targets, and build on the multiscale graph design. Published open access in BMC Bioinformatics, the work arrives at a moment when machine learning is reshaping how the earliest stages of drug discovery are conducted. If the cold-start gains reported here hold up in real-world target discovery pipelines, the era in which a newly implicated protein sat beyond the reach of computational prediction may be drawing to a close, replaced by graph-based models that read biology at every scale at once.

Subject of Research: Cold-start drug-target interaction prediction using a multiscale relation-aware heterogeneous graph transformer

Article Title: HierHGT-DTI: a multiscale relation-aware heterogeneous graph transformer for cold-start drug-target interaction prediction

Article References: Chen, Z., & Wan, Y. (2026). HierHGT-DTI: a multiscale relation-aware heterogeneous graph transformer for cold-start drug-target interaction prediction. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06637-3

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06637-3

Keywords: drug-target interaction, cold-start prediction, heterogeneous graph transformer, multiscale representation, machine learning, drug discovery, DrugBank, BioSNAP, BindingDB, protein targets, graph neural networks, bioinformatics

The study, led by Hao Pang, Fiseha Berhanu Tesema, Tianxiang Cui, Yuan Cheng, and Yanwen Mao, introduces a model called BCAG-DTI, short for Bidirectional Cross-Attention and Global Aggregation drug-target interaction model. Rather than simply claiming superiority over the widely used MolTrans baseline, the researchers ran eight carefully controlled configurations that separately toggled bidirectional cross-attention, a global average-and-max pooling layer, the design of the classification head, and the optimisation strategy. By changing one factor at a time across fixed training, validation, and test partitions with five matched random seeds, they could attribute each gain to its true source, a level of experimental hygiene that remains rare in a literature crowded with simultaneous architectural and training tweaks.

The headline results are striking. On the BindingDB benchmark, the complete BCAG-DTI configuration lifted the mean area under the receiver operating characteristic curve, or AUROC, from 0.8815 for MolTrans to 0.9063. On BIOSNAP, the corresponding figure rose from 0.8631 to 0.8891, and on DAVIS from 0.8808 to 0.8955. Because AUROC alone can flatter models on imbalanced datasets, the team also measured the area under the precision-recall curve, AUPRC, which is often the more honest metric for interaction prediction. There the improvements were even more pronounced: gains of 0.0831 on BindingDB, 0.0346 on BIOSNAP, and 0.0619 on DAVIS, representing substantial advances in the model’s ability to rank true interactions above false ones.

The most provocative finding, however, lies in the ablation analysis that followed. When each component was tested in isolation, the enhanced classifier design emerged as the single most influential element, achieving the highest mean AUROC on BindingDB and BIOSNAP and the highest mean AUPRC and F1-score on all three datasets. The enhanced training strategy also delivered substantial improvements on its own. By contrast, bidirectional cross-attention alone actually reduced performance when paired with the baseline training configuration. In other words, the glamorous cross-modal attention machinery that gives the model its name contributes value only under the right optimisation conditions, and its benefit is dataset-dependent rather than universal. On DAVIS, it was the combination of cross-attention, global pooling, and enhanced training that achieved the highest mean AUROC, underscoring how sensitive these components are to context.

This pattern carries a lesson that extends well beyond one model family. Transformer-style cross-attention between molecular and protein representations has become a fashionable design choice in bioinformatics, frequently credited with allowing models to learn which parts of a ligand speak to which residues of a binding pocket. The Nottingham results suggest that such credit may sometimes be misplaced, or at least overstated, because classifier head design and optimisation schedules can account for a large share of the reported improvement. For a field racing to publish ever-larger architectures, the message is that rigorous component-wise evaluation is not optional bookkeeping but the difference between genuine scientific progress and an artefact of training procedure.

Robustness testing added another layer of nuance. In experiments on BIOSNAP designed to simulate real-world sparsity, BCAG-DTI outperformed MolTrans when predicting interactions for unseen drugs, for unseen proteins, and in settings where 70 to 90 percent of interaction labels were missing. These scenarios matter enormously in practice, because the vast majority of possible drug-protein pairs have never been measured, and any screening tool must generalise beyond the sparse matrix of known interactions. The one blemish appeared at 95 percent missing data, where BCAG-DTI’s AUROC and F1-score dipped slightly below the baseline, a reminder that even improved models have limits when supervision becomes extremely thin.

To situate their results against the broader literature, the authors also evaluated an external reference model, CPI-GGS, on the same fixed BIOSNAP partitions. CPI-GGS achieved 0.8619 plus or minus 0.0023 AUROC, 0.8645 plus or minus 0.0042 AUPRC, and 0.7938 plus or minus 0.0028 F1-score, compared with 0.8891 plus or minus 0.0078, 0.8992 plus or minus 0.0067, and 0.8178 plus or minus 0.0094 for BCAG-DTI. Crucially, the team interpreted this comparison with appropriate caution, noting that the two systems use different input preprocessing pipelines and differ substantially in model capacity. Rather than declaring outright victory, they framed the numbers as a same-split reference point, a refreshing contrast to the practice of comparing scores across incompatible datasets and splits that still plagues the field.

Perhaps the most intellectually honest portion of the paper concerns attention interpretability. Cross-attention weights are often visualised and marketed as evidence that a model has learned genuine binding contacts between a ligand and a protein pocket. The authors’ case analysis pushes back on this convention, indicating that cross-attention weights should be treated as model-internal allocation patterns rather than validated binding contacts. This distinction matters for downstream users: a researcher who trusts an attention heatmap as a structural hypothesis could waste laboratory resources chasing a pattern that reflects the model’s internal bookkeeping rather than physical chemistry. It is a caution that resonates across the growing literature on attention-based models in the life sciences.

The study’s methodology deserves emphasis as a model for the field. Fixing the data partitions across all principal experiments and running five matched random seeds allowed the authors to report means with a meaningful sense of variance, guarding against the lucky-seed effect that can inflate single-run results. The three benchmarks themselves span different regimes: BindingDB offers a large, chemically diverse collection of measured affinities, BIOSNAP provides a network-derived interaction dataset well suited to cold-start and missing-data experiments, and DAVIS presents a smaller, kinase-focused panel. Showing consistent gains across all three, while honestly reporting where components underperform, lends the conclusions a credibility that single-benchmark papers rarely achieve.

For drug discovery pipelines, the practical implications are immediate. Teams building interaction predictors should scrutinise their classification heads and optimisation strategies before investing in architectural complexity, since the cheapest improvements may lie in the least glamorous parts of the stack. Cross-attention and pooling modules should be adopted with the understanding that their value is conditional, emerging only in combination with appropriate training regimes and on certain data distributions. And anyone publishing comparisons against established baselines should take a page from this study’s playbook: hold the data fixed, vary one thing at a time, and report what actually moved the needle. As machine learning continues to reshape how candidate drugs are prioritised for synthesis and testing, evaluations of this rigour will be essential to ensure that the field’s progress is real rather than an artefact of experimental design.

Subject of Research: Controlled ablation evaluation of architectural, classifier, and training refinements in deep learning-based drug-target interaction prediction

Article Title: Controlled evaluation of architectural, classifier, and training refinements in MolTrans-based drug-target interaction prediction

Article References: Controlled evaluation of architectural, classifier, and training refinements in MolTrans-based drug-target interaction prediction. (n.d.). https://doi.org/10.1186/s12859-026-06634-6

Image Credits: AI Generated

DOI: 10.1186/s12859-026-06634-6

Keywords: drug-target interaction, MolTrans, cross-attention, ablation study, deep learning, BindingDB, BIOSNAP, DAVIS, AUROC, classifier design, optimisation, bioinformatics

Small molecules face a formidable structural challenge, however. The PD-1/PD-L1 interface is a broad, shallow protein-protein contact surface, long considered a difficult target for drug-like compounds. Progress has come from chemotypes such as the biphenyl-based inhibitors developed by Bristol-Myers Squibb, including BMS-202, which bind PD-L1 and induce dimerization that blocks PD-1 recognition. Compounds such as INCB086550 and the oral modulator CA-170 have entered early clinical evaluation, proving that small-molecule checkpoint inhibition is feasible. Still, only a small fraction of explored scaffolds has advanced to the clinic, and screening the relevant chemical space experimentally is costly and slow. Computational triage has therefore become essential for deciding which candidates deserve synthesis and bioassay follow-up.

A team led by Monsin Sangsawat and Pornchai Rojsitthisak at Chulalongkorn University, with collaborators including Worathat Thitikornpong, Cong Feng, and Ming Chen, has now published an integrated solution in the journal Results in Chemistry. Their study reports an open-source, end-to-end cheminformatics framework for predicting the potency of small-molecule PD-L1 inhibitors, built entirely in the KNIME Analytics Platform. Every processing step, from data retrieval to model deployment, is encoded as an explicit node sequence, and the executable workflow file, curated datasets, and fixed split identifiers have been deposited in a public GitHub repository. The design directly confronts a persistent problem in the field: many published machine-learning models for PD-L1 are difficult to reproduce because training data, preprocessing details, or workflow code are inaccessible, and few are evaluated across independent data sources.

The foundation of the framework is a rigorously curated dataset. The researchers queried ChEMBL, targeting the human PD-L1 entry CHEMBL4523993, and retrieved more than 15,000 bioactivity records. To reduce label noise, they applied assay-level filters restricting the modeling set to homogeneous Alpha and HTRF assay formats, the two readouts with the most consistent annotation and the broadest activity coverage. After removing ambiguous qualifiers, duplicates, and non-numeric annotations, and after converting IC50 values to logarithmic pIC50 units, 2413 unique compounds remained. These were stratified into three potency tiers: active compounds with IC50 below 10 nanomolar, moderate inhibitors between 10 and 100 nanomolar, and weak or inactive compounds above 100 nanomolar. The curated set spans 953 unique Bemis-Murcko scaffolds, preserving the chemical diversity reported for PD-L1 inhibitors, from biphenyls and combretastatin analogs to cyclic peptides.

Molecular representation was treated as a central experimental variable. Each compound was encoded with 119 RDKit physicochemical descriptors, which were filtered for zero variance, pruned for high inter-correlation, and refined using recursive feature elimination with cross-validation, ultimately yielding 23 descriptors covering lipophilicity, ring content, van der Waals surface areas, and molecular quantum numbers. In parallel, two fingerprint families were generated: MACCS structural keys and Morgan circular fingerprints with a radius of two, together with a hybrid concatenation of the two. Four regression algorithms were benchmarked across the three representations, producing twelve configurations: Random Forest, a Keras deep neural network, Gradient Boosting, and XGBoost. Each configuration was trained under four fixed random seeds, and results were reported as means with standard deviations to capture stochastic sensitivity.

The Random Forest consistently delivered the most reliable balance between fit and generalization. Using the hybrid MACCS-Morgan representation, it achieved a coefficient of determination of 0.871 on the held-out test set and 0.863 on a strictly external validation set of 403 compounds that were never touched during model selection. Cross-validated Q2 values ranged from 0.888 to 0.902 across descriptor sets, and external Qext2 values reached as high as 0.919 with MACCS fingerprints alone. The deep neural network, although it attained higher training performance for some representations, showed a larger decline on held-out data, suggesting a greater tendency toward overfitting at the current sample size. Importantly, the team verified that no test or external compound shared an identical canonical SMILES with any training compound, and performance was unchanged after excluding high-similarity nearest neighbors, ruling out information leakage through duplicates.

Generalization was then stress-tested with a scaffold-disjoint partition, in which entire Bemis-Murcko scaffold families were assigned exclusively to training or held-out sets. Under this stricter regime, the hybrid Random Forest model achieved a mean test R2 of 0.858 and external validation R2 of 0.851, closely matching the random-split results. This indicates that the reported accuracy was not inflated by scaffold sharing and that the model genuinely extends to structural families absent from training. Prediction reliability was strongest in the moderate-to-potent activity range most relevant to virtual screening, while weak inhibitors with pIC50 below 7 showed larger errors and a modest tendency toward over-prediction, a pattern attributed to the relative scarcity of weak compounds in the curated data.

A distinctive strength of the study is its emphasis on knowing when to trust the model. Two applicability-domain diagnostics were implemented: a distance-based criterion derived from the mean and standard deviation of minimum training-set distances, and a leverage-based Williams plot with the classical threshold of three times the number of descriptors divided by the number of training compounds. Nearly all test and external-validation compounds fell inside the domain, and out-of-domain compounds in the independent BindingDB application set showed markedly larger errors than in-domain ones, with mean absolute errors of 1.884 versus 0.809 under the distance criterion. The researchers applied the model to 124 unique BindingDB compounds after removing 444 overlapping entries. Compounds sharing biphenyl scaffolds or cyclic peptide chemotypes with the training distribution showed small residuals, while structurally distinct pyrazolone derivatives and arylindanyloxypyridine frameworks showed substantially larger errors, pinpointing precisely where future data expansion would pay off.

The framework also translates statistical features into medicinal chemistry insight. SHAP analysis of the Random Forest identified the number of aromatic heterocycles, molar refractivity, and lipophilicity as the most influential descriptors, with threshold-like dependencies. The strongest potency-associated MACCS key simply records the presence of chlorine, but the corresponding Morgan features resolve it more sharply as a chlorinated aryl ring system and a chloro-substituted biaryl, discriminating potency with correlations up to 0.75. Conversely, a potency-decreasing CH2-O linkage characteristic of dibenzyl ether motifs is captured by MACCS but not the leading Morgan features, explaining why the two fingerprint blocks carry non-redundant, oppositely signed information. Scaffold analysis showed simple dibenzyl-ether and stilbene-like cores averaging pIC50 values of 6.3 to 7.3, while phenylpyridine and phenylpyrazine cores bearing basic amines averaged 8.4 to 10.2. These trends map directly onto the predominantly hydrophobic PD-L1 binding cleft, defined by residues such as Tyr56, Met115, and Tyr123, where halogenated aromatic extension and nitrogen heterocycles favor productive binding, and outlier compounds flagged by the model were examined through GNINA docking at this cleft against the reference inhibitor BMS-202.

The authors are candid about limits. Assay-format annotations come from ChEMBL descriptions rather than independently verified conditions, which may bias the set toward chemotypes preferentially tested in Alpha and HTRF formats, and a comparison against the published PDL1-inhi.predictor showed that an external model captured the BindingDB chemical space somewhat better. Yet the study’s contribution extends beyond any single metric: it delivers a transparent, inspectable, and adaptable platform that couples potency prediction with uncertainty quantification, supports pre-docking prioritization, and can be extended with uncertainty-guided active learning. The team suggests that high-ranking compounds within the applicability domain, especially those representing novel chemotypes, could be selected for prospective testing in the same Alpha or HTRF assays, followed by model retraining. For a field where reproducibility is often the missing ingredient, an open workflow anyone can audit and rerun on an ordinary laptop may prove as valuable as the model itself.

Subject of Research: Machine learning-guided discovery and prioritization of small-molecule PD-L1 inhibitors using an open-source KNIME workflow with curated bioactivity data and applicability-domain assessment

Article Title: Machine-learning-guided discovery and prioritization of PD-L1 inhibitors: An open-source KNIME workflow with curated bioactivity data, molecular fingerprints, and applicability-domain assessment

Article References: Sangsawat, M., Watpuang, B., Nathapinthu, T., Srisiriroj, P., Nalinratana, N., Taechawattananant, P., Thitikornpong, W., Feng, C., Chen, M., & Rojsitthisak, P. (2026). Machine-learning-guided discovery and prioritization of PD-L1 inhibitors: An open-source KNIME workflow with curated bioactivity data, molecular fingerprints, and applicability-domain assessment. Results in Chemistry, 30, Article 103877. https://doi.org/10.1016/j.rechem.2026.103877

Image Credits: AI Generated

DOI: 10.1016/j.rechem.2026.103877

Keywords: PD-L1, PD-1/PD-L1 immune checkpoint, machine learning, KNIME workflow, Random Forest, molecular fingerprints, MACCS, Morgan fingerprints, applicability domain, ChEMBL, BindingDB, cancer immunotherapy