Journal of Cheminformatics

Predicting blood-brain barrier (BBB) penetration remains a central challenge in CNS drug discovery. Existing computational models rely on physicochemical descriptors and are blind to active transport biology - the efflux pumps and carrier proteins that dominate drug exclusion at the BBB in vivo. We present BBB-Nuke, a modular prediction pipeline that integrates physicochemical scoring with explicit efflux transporter substrate modeling. The system computes ten molecular descriptors, predicts ionization state via a graph convolutional network, scores CNS-MPO desirability, and estimates substrate probability for seven efflux transporters (P-gp/MDR1, BCRP/ABCG2, MRP1, MRP2, MRP4, MATE1, OAT3) using Random Forest classifiers trained on curated ChEMBL bioactivity data. A gradient-boosted classifier trained on 67 features - ten physicochemical, seven efflux transporter probabilities, and fifty fingerprint-derived principal components - achieves an area under the receiver operating characteristic curve (AUROC) of 0.933 {+/-} 0.006 under five-fold cross-validation on 9,262 labeled compounds, and 0.810 on a fully held-out benchmark of 470 clinically validated compounds. In head-to-head comparisons, BBB-Nuke outperforms CNS-MPO, LightBBB, ADMETlab 2.0, and BBB-Score on both cross-validation and external test sets. We apply the pipeline to screen over one billion commercially available compounds from the Enamine REAL library and PubChem, identifying enriched regions of BBB-penetrant chemical space and characterizing the structural features that distinguish permeable from excluded molecules. BBB-Nuke is freely available as a Python package, REST API, and Model Context Protocol server.

Molecular representation learning has become a central component of modern computational drug discovery. Existing molecular foundation models mainly rely on SMILES strings, two-dimensional molecular graphs, or three-dimensional atomic coordinates. However, many molecular properties are ultimately governed by the molecular surface, where intermolecular recognition, solvation, electrostatic complementarity, and ligand-protein interactions occur. In this work, we propose MolMAE, a surface-guided multimodal masked autoencoder for molecular representation learning. MolMAE takes molecular surface point clouds, three-dimensional molecular graphs, and SMILES-derived fragment and functional-group tokens as complementary input modalities, and learns a unified multimodal molecular embedding through functional-group-aligned masked autoencoding. During pretraining, chemically corresponding local regions are jointly masked across surface, graph, fragment, and functional-group views, forcing the model to reconstruct missing geometric, physicochemical, structural, and semantic information from the remaining context. While molecular surface reconstruction serves as the primary pretraining objective, graph-, fragment-, and functional-group-level reconstruction tasks provide complementary supervision that encourages the model to capture molecular topology, bonding patterns, stereochemistry, local chemical environments, and substructure organization. In addition to reconstructing surface geometry, MolMAE reconstructs surface-associated physicochemical fields, including electrostatic potential and Fukui-related descriptors, enabling the model to learn chemically meaningful surface representations. Pretrained on approximately 261K lead-like bioactive molecules, MolMAE achieves strong performance on the ESOL benchmark under scaffold splitting and competitive performance across multiple molecular property prediction tasks. These results suggest that molecular surface-guided pretraining can complement conventional graph-, sequence-, and atom-coordinate-based molecular representations, especially for property prediction tasks influenced by exposed surface geometry and surface-associated physicochemical patterns.

Motivation: Human liver microsome (HLM)-based metabolic stability assays are fundamental in early drug discovery, shaping pharmacokinetic profiles and oral bioavailability. However, these experimental assays are labor-intensive and time-consuming, limiting their application in large-scale virtual screening. Computational models can prioritize compounds at scale, yet most are classification-based, leaving quantitative and interpretable prediction of HLM half-life limited. Results: In this study, we developed a quantitative machine learning model for the direct prediction of HLM half-life (T1/2) by integrating 11,790 compounds combining in-house and curated public data. Among various combinations of molecular features and learning algorithms, the XGBoost model with RDKit 2D descriptors achieved the best predictive performance, with an RMSE of 0.507 and an R2 of 0.431 on an independent test set. Shapley Additive Explanations (SHAP) analysis identified lipophilicity and known metabolic soft-spot features as the primary contributors to the predictions. These results suggest that this quantitative approach provides a practical framework for defining metabolic stability margins, thereby supporting rapid Go/No-go decisions in preclinical drug discovery. Availability: The source code, data, and trained model are available at https://github.com/joshua-416/PredHLM.

ONTOSIGHT(R) ADMETron is a high-performance SaaS based AI platform designed for the rapid profiling and visualization of ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) properties. The platform integrates a highly interactive web interface with a robust predictive engine, enabling the batch processing of compounds for high-throughput virtual screening. The core engine employs an ensemble model that combines recurrent neural network (RNN)-derived embeddings from SMILES strings with physicochemical descriptors, which are fed into gradient boosting machines (GBMs). This architecture provides accurate predictions across 34 distinct ADMET endpoints, encompassing critical categories such as physicochemical properties, absorption, CYP450 inhibition, hERG toxicity, and mutagenicity. The platforms superior performance is quantitatively validated by its top-tier ranking on the Therapeutics Data Commons (TDC) ADMET Benchmark Group, demonstrating robustness and generalizability with notable results including 2nd place for Ames mutagenicity (AUROC 0.870) and 2nd place for LD50 (MAE 0.573). In addition to its predictive capabilities, ADMETron introduces a novel SAR analysis framework that enables real-time comparison of multiple compounds and approved drugs through an interactive radar graph visualization. Comparative evaluation against widely used online ADMET platforms demonstrated broader endpoint coverage, including pharmacokinetic, physicochemical, and medicinal chemistry assessments within a unified environment. The combination of benchmark-validated predictive performance, comprehensive ADMET profiling, and advanced visualization tools positions ADMETron as a next-generation platform for virtual screening, lead optimization, and data-driven decision-making in modern drug discovery (https://admetron.partex.ai/).

Targeted covalent inhibitors are a powerful, yet underexplored, class of therapeutics, and current computational covalent screeners are constrained in early drug discovery due to the need for prior knowledge of the target site and limited throughput. We present CovSite, a blind covalent screening tool that identifies candidate reactive residues across the entire protein surface, utilizing only the protein structure and electrophile SMILES. CovSite applies a pipeline of four orthogonal physicochemical filters (nucleophile identification, solvent accessibility, environment-dependent deprotonation prediction, and semi-quantum-mechanical reactivity ranking) to identify potential small-molecule candidate inhibitors. Validated against 2,062 diverse covalent protein-ligand complexes spanning six nucleophilic residue types, CovSite achieves a 98.5% blind target site hit on a held-out benchmark set of 207 cysteine-targeted complexes while reducing the search space by 97.8%. The target-site hit detection exceeds the 53-62% accuracy of popular covalent screening tools operating under non-blind conditions on the same benchmark set. By extending nucleophilic coverage beyond cysteine to include serine, threonine, lysine, histidine, and tyrosine, and completing a screening of a 200-residue protein in two to three minutes on standard hardware, CovSite serves as a platform technology with the potential to address critical gaps in throughput, generalizability, and accuracy in this field of covalent screening. We demonstrate this capability by using CovSite as a blind, ligand-specific approach that enables iterative, machine-learning-driven covalent inhibitor generation that is impractical with existing tools, establishing a foundation for computationally guided covalent drug discovery for novel and understudied targets. TOC Figure O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=100 SRC="FIGDIR/small/739288v1_ufig1.gif" ALT="Figure 1"> View larger version (27K): org.highwire.dtl.DTLVardef@17abeeforg.highwire.dtl.DTLVardef@18d6299org.highwire.dtl.DTLVardef@14437eaorg.highwire.dtl.DTLVardef@1b2ec81_HPS_FORMAT_FIGEXP M_FIG C_FIG

The rise of antibiotic resistance necessitates the discovery of antibacterial compounds with novel mechanisms of action (MoAs). Recent machine learning approaches have shown promise in antibacterial compound discovery, but often identify derivatives of known antibiotic classes rather than mechanistically novel compounds. Previous approaches applied Tanimoto similarity filters at the end of screening pipelines, but this method has substantial drawbacks: Tanimoto similarity can be misleading in chemical space, and post-hoc filtering does not influence what activity models learn to prioritize. Here, we present a machine learning pipeline that addresses chemical novelty upfront by employing an XGBoost-based MoA classifier to explicitly prioritize compounds predicted to have mechanisms distinct from known antibiotic classes, combined with graph neural networks for antibacterial activity and toxicity prediction. Applied to the Zinc20 database, our approach successfully identified non-toxic antibacterial compounds structurally distinct from known antibiotics. Notably, the majority of these hits exhibited membrane-targeting activity with selectivity for bacterial cells over mammalian cells, suggesting potential for next-generation membrane-active antibiotics. However, we did not identify compounds with novel protein targets. Systematic analysis revealed that this limitation stems from mechanistic bias in training data rather than model architecture. Specifically, our activity model learned to preferentially score compounds similar to specific groups in the training data, thus overrepresenting certain MoA classes including membrane-active compounds. Even substantial model architecture and training data enhancements did not overcome this constraint. Our findings demonstrate that the primary bottleneck for discovering mechanistically novel antibiotics is the scarcity of diverse, mechanistically-annotated training data. This work provides both a methodological framework for mechanism-aware screening and critical insights into data requirements for genuinely novel antibiotic discovery.

Accurate prediction of protein-ligand binding affinity is a crucial goal in structure-based drug discovery, with the potential to significantly shorten development timelines. Recently, a new wave of machine learning models based on co-folding, such as Boltz-2 and IsoDDE, has demonstrated performance that matches or exceeds that of gold-standard physics-based methods like Free Energy Perturbation (FEP). This paper provides a critical assessment of these claims, revealing that current benchmarks are heavily influenced by data leakage, and proposes a new benchmark that explicitly controls for data leakage. We demonstrate that splitting by protein-sequence identity is inherently insufficient to prevent data leakage due to "target mirroring," in which homologous proteins with low overall sequence identity still exhibit highly correlated binding profiles. Our meta-analysis of documents in the ChEMBL 36 database identifies more than 6,000 such assay pairs and finds that leakage persists for sequence-identity thresholds as low as 0.2, well below the values commonly used in benchmarks today. Additionally, we show that a ligand-only baseline model, which lacks protein structural information, achieves surprisingly high performance on the FEP+ 4 and OpenFE benchmarks (r = 0.66 and r = 0.36, respectively). Our results indicate that current benchmarks tend to reward models for memorizing training data and exploiting localized leakage rather than truly learning biophysical principles. To address this issue, we propose the Novelty-Tiered Affinity Benchmark, in which the test data is partitioned into ligand novelty tiers. In the most challenging tier (Tanimoto similarity < 0.35), ligand-only models perform notably worse (r = 0.14), offering a clear baseline for evaluating genuine generalization. We argue that the field must move beyond sequence-based splits to ensure that AI-driven discovery translates into successful prospective laboratory research.

BackgroundAdverse drug-drug interactions (DDIs) cause preventable hospitalizations, but exhaustive experimental screening of all drug pairs is infeasible. Many computational predictors rely on SMILES or other molecular representations, limiting their direct applicability to biologics and other non-small-molecule therapeutics. We present a structure-agnostic framework that combines semantic representations derived from Medical Subject Headings (MeSH) with graph-derived topology from a Drug-Target-Protein knowledge graph constructed from DrugBank and UniProt. We further investigate how variation in MeSH annotation depth affects predictive performance. ResultsDrugs are grouped according to their deepest MeSH annotation level (Low, Mid, or Deep), and performance is evaluated across the resulting interaction categories in transductive and inductive settings. The Intermediate ontology scope (Low+Mid) provides the most stable performance, while adding Deep-level terms offers limited and inconsistent benefit. Lightweight topological descriptors are integrated with MeSH features through instance-wise, dimension-specific latent-space gating, using curated reliable-negative pairs for supervision. Fusion improves mean performance over the MeSH-only baseline across all six categories in the transductive setting. Under induction, the clearest gains occur for Low-Low interactions ({Delta}AUROC = 0.056;{Delta} F1 = 0.137) and Low-Mid interactions ({Delta}AUROC = 0.077;{Delta} F1 = 0.114). ConclusionsMeSH annotation depth is associated with systematic variation in DDI prediction performance that aggregate evaluation can obscure. Graph-derived topology is particularly beneficial when ontology annotations are shallow. The framework provides a common, structure-agnostic representation compatible with both small-molecule and biologic therapeutics and supports first-pass DDI prioritization for subsequent expert assessment.

Machine learning (ML) models for molecular property prediction are increasingly deployed in drug discovery, yet their adoption in real-world scenarios requires an understanding of the conditions in which a model succeeds or fails. While standardized benchmarks are powerful instruments to measure and unlock progress in ML research, they should not be blindly treated as the end goal. Especially static and retrospective benchmarks, in which no true unknown test set is employed, limit our ability to robustly validate a model's performance. Building on the collective expertise of a cross-industry consortium, we present a model validation framework consisting of five recommendations that would enable the community to move beyond aggregate metrics toward understanding where and why molecular property prediction models fail. We connect evaluation choices to real-world applications and case studies encountered in pharmaceutical research. The framework proposes splitting strategies that mimic realistic distribution shifts and expose common failure modes. We apply the recommended framework to a recently released dataset of absorption, distribution, metabolism, and excretion (ADME) properties. Across two complementary model algorithms, our case studies reveal four distinct failure modes (extrapolation, interpolation, representation, and evaluation), showing that model errors arise not only from distribution shift but also from limitations in molecular representations. Our results show that commonly used evaluation protocols can significantly overestimate performance and may not detect important model failure modes. All software and data are released via https://github.com/srijitseal/polaris.

Protein-protein interactions (PPIs) are central to biological processes. Designing small molecules that modulate dysregulated PPIs holds strong promise for targeting undruggable proteins. However, existing structure-based drug design approaches focus on well-defined small-molecule binding pockets and struggle to generalize to large, shallow, and chemically complex PPI interfaces. Here, we introduce Pep2Mol, a diffusion-based generative model for 3D molecule design that targets orthosteric PPI sites by explicitly incorporating binding peptides or proteins as structural guidance, moving beyond conventional pocket-conditioned generation. To enable model development and benchmarking, we curate a large-scale, high-quality dataset of 10,956 experimentally resolved protein complex structure pairs, each pairing an orthosteric competitive ligand with a protein binder at overlapping receptor interfaces. Pep2Mol integrates two SE(3)-equivariant graph neural networks that encode protein-ligand and protein-peptide interactions respectively, and fuses these representations via attention-based conditioning to jointly guide the diffusion trajectory. Extensive evaluations demonstrate that Pep2Mol generates chemically valid ligands with state-of-the-art binding affinities, providing a strong foundation for small-molecule inhibitor design against challenging PPI interfaces.

Accurate ADMET prediction is essential for prioritizing compounds before costly experimental validation, yet ADMET tasks are highly heterogeneous. Properties such as solubility, permeability, protein binding, clearance, transporter activity and toxicity are governed by different molecular signals, ranging from local functional groups and physicochemical descriptors to bonded topology and three-dimensional geometry. Consequently, a single molecular representation or backbone is unlikely to be optimal across all ADMET tasks. We present Trimole-Hybrid, a task-wise multimodal framework that addresses ADMET heterogeneity by selecting or combining predictors built from complementary molecular representations. Trimole-Hybrid constructs a candidate pool of SMILES-, graph-, geometry-sensitive EPT/3D- and chemical descriptor-based predictors. For each task, Trimole-Hybrid selects the best-performing predictor to obtain the final prediction. On 22 Therapeutics Data Commons ADMET benchmarks, Trimole-Hybrid exceeded the public TDC top-1 methods on 10 tasks and ranked within the top 10 for 21 tasks. Ablation studies confirmed the contribution of both complementary multimodal molecular representations and task-specific ensemble strategies. In two small-molecule case studies, Trimole-Hybrid shows sensitivity to changes in essential functional motifs, suggesting its ability to capture ADMET-relevant molecular substructures.

High-quality molecular datasets are essential for reliable machine learning in cheminformatics and bioinformatics, yet dataset quality is rarely assessed systematically and its relationship with downstream model performance remains poorly understood. Here, we present MolJam, an open-source framework for quantitative assessment of molecular dataset quality across five dimensions-structural integrity, data quality, experimental information quality, chemical space coverage, and data distribution-using 12 standardized metrics. Application of MolJam to 11 MoleculeNet and eight ChEMBL-derived datasets revealed widespread and heterogeneous quality issues, including undefined stereochemistry in up to 70.72% of molecules, inconsistent molecular representations, and contradictory labels. We next asked whether improving these quality metrics necessarily improves machine learning performance. Refinement of the ESOL and Lipophilicity datasets increased their MolJam quality scores but produced mixed effects on predictive performance, suggesting a competing influence of reduced dataset size. Controlled ablation experiments further demonstrated that both dataset quality and data quantity contribute to model performance and, notably, that retaining molecules with incomplete stereochemical information can outperform their removal when the resulting gain in data quantity offsets the quality penalty. Thus, molecular dataset curation cannot be reduced to maximizing data cleanliness alone but requires balancing multiple dimensions of data quality against information loss. MolJam provides a standardized framework for diagnosing molecular dataset limitations, comparing benchmark quality, and quantitatively evaluating how data curation decisions influence downstream machine learning.

Protein-ligand interaction diagrams are a routine part of structural and medicinal chemistry, but the tools that produce them tend to force a choice: comprehensive detection with tabular output, publication-quality figures behind a licence, or a scripting environment that assumes expertise. PandaMap (Protein AND ligAnd interaction MAPper) is an open-source Python package that produces a 2D interaction diagram, an interactive 3D viewer, a text report, a machine-readable CSV, and a four-panel graphical summary from a single command. It reads PDB, mmCIF and PDBQT files, detects 15 interaction classes using crystallographically validated distance thresholds, and depends only on NumPy, Matplotlib, BioPython and Requests; RDKit improves the 2D ligand layout when present but is not required. Hydrogen bonds are filtered on the true D-H{middle dot} {middle dot} {middle dot} A angle when the structure contains explicit hydrogens, matching PLIPs 100{whitebullet} criterion on the same evidence, and on distance alone otherwise, with the provenance of each measurement recorded. We benchmarked the package on three complexes chosen for different chemistry: enolase with a phosphonate transition-state analogue (PDB 1ELS), the EGFR kinase with erlotinib (1M17), and aldose reductase with IDD594 (1US0). PandaMap recovers the contacts these structures are known for, including the EGFR hinge hydrogen bond to MET769 and the IDD594 bromine{middle dot} {middle dot} {middle dot} THR113 halogen bond, both at distances identical to PLIPs. All detection thresholds, scoring weights and the exact commands used are given in the Supplementary Information, and the release carries a regression suite covering each interaction class. PandaMap 4.3.0 is available on PyPI under the MIT licence.

Agentic artificial intelligence (AI) systems increasingly claim to automate scientific research, yet independent evaluations report persistent gaps between those claims and demonstrated capability. We tested frontier agentic AI systems on three practical problems: prediction of treatment non-response in immune-mediated inflammatory diseases, optical chemical structure recognition for literature mining, and prediction of drug-design-related properties from small datasets. Each problem was first assigned to autonomous frameworks and then reattempted as human-led, AI-assisted work. Autonomous runs failed in most cases, while human-led work produced reusable resources and modest but defensible performance, including new evidence for possible mechanisms of treatment resistance and a more practical benchmark for mining chemical structures from scientific papers. Property prediction was the single task on which one autonomous AI framework matched the human expert. We conclude that current frameworks can carry out engineering and analysis once a human expert leads the project, but cannot yet engineer a novel solution without oversight. The use of AI on real-life scientific problems remains an art rather than a routine technology.

MotivationProteolysis-targeting chimeras (PROTACs) enable targeted protein degradation through ternary complex formation with E3 ubiquitin ligase. However, the rational design of PROTACs remains highly challenging due to limited structure-activity relationship data and the vast conformational diversity of linkers. Existing computational approaches can be broadly divided into structure-based ternary modelling methods and fragment-based linker generation models. Although these approaches have advanced PROTAC design, they typically neglect key physicochemical constraints and linker-length control during the generation process, causing the generated PROTACs to lack balanced structural properties required for effective ternary complex formation with drug-like characteristics. ResultsTo address these limitations, we propose DesignMaster, a diffusion-based generative framework that explicitly incorporates linker length and physicochemical properties as controllable conditioning signals. DesignMaster employs an E(3)-equivariant graph Transformer with a gated multi-condition fusion module to inject linker length and physicochemical constraints throughout the diffusion process, enabling fine-grained and constraint-aware molecular generation. Experiments on PROTAC-DB 2.0 and 3.0 demonstrate that DesignMaster outperforms state-of-the-art baselines, with a 3.2% improvement in validity and a 34.4% improvement in recovery. The Case study shows DesignMaster achieves a 51.78% reduction in RMSD when predicting the linker of PROTAC BCPyr targeting 6W7O, highlighting its potential for practical structure-guided PROTAC design. AvailabilityThe source code and datasets are available at https://github.com/ABILiLab/DesignMaster.

BackgroundCancer continues to pose a massive global health burden. While plant-derived phytochemicals offer promising therapeutic leads, existing natural product databases often lack cancer specificity, dataset downloadability, and integrated screening tools. MethodsWe developed Sanjeevani, an integrative web platform cataloguing 4,823 curated anticancer phytochemicals. Using a balanced dataset of 9,646 molecules, we trained Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbours classifiers using a hybrid feature representation of RDKit descriptors and 2048-bit ECFP4 fingerprints. The platform also integrates AutoDock Vina for web-based molecular docking for binding affinity, poses prediction and ADMET-AI for pharmacokinetics estimation. ResultsThe SVM model demonstrated the strongest predictive capability, achieving a top test accuracy of 0.966 and a ROC-AUC of 0.992. Benchmarking across five docking tools confirmed that AutoDock Vina successfully balanced computational automation with literature-consistent binding affinity replication. The final architecture provides rapid interactive 2D/3D visualizations integrated with downstream analysis tools. ConclusionSanjeevani provides an open-access, one-stop pipeline that bridges the gap between raw natural product data and actionable computational screening, accelerating natural product-based oncology drug discovery. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=80 SRC="FIGDIR/small/732344v1_ufig1.gif" ALT="Figure 1"> View larger version (17K): org.highwire.dtl.DTLVardef@77183borg.highwire.dtl.DTLVardef@d7e465org.highwire.dtl.DTLVardef@1d3dfd7org.highwire.dtl.DTLVardef@10cc94d_HPS_FORMAT_FIGEXP M_FIG C_FIG

Drug discovery (DD) is a complex, time-consuming, and resource-intensive process that involves the identification of therapeutic targets, selection of bioactive compounds, and extensive experimental validation. The discovery of promising therapeutic compounds from large libraries of phytochemicals and synthetic molecules remains a major challenge in modern drug development. Screening millions of compounds through conventional experimental approaches requires substantial time, cost, and computational resources. In recent years, in- silico molecular docking has emerged as an important computational approach for predicting interactions between small molecules and target proteins, thereby helping researchers prioritize promising compounds for further investigation. Several molecular docking webservers, including iScreen, SwissDock, CB-Dock2, DockThor, and MTiOpenScreen, have been developed to support virtual screening studies. However, many currently available platforms still face some important limitations. Most existing tools lack integrated repositories of medicinal plant-derived phytochemicals and organism-derived bioactive compounds, automated mapping between plants and their associated phytochemicals, and flexible ligand retrieval using chemical names, SMILES strings, PubChem CIDs, or drug names. In addition, many platforms require extensive manual protein and ligand preparation, provide limited support for AlphaFold-predicted protein structures, and lack efficient large-scale multi-target virtual screening. Most existing docking platforms offer limited support for interactive inspection of docked protein-ligand complexes, often requiring users to download the results and analyse them using external molecular visualization software. To address these limitations, we developed FlexAutoDock, an automated cloud-based molecular docking platform that provides a unified environment for protein-ligand docking and large-scale virtual screening. Unlike existing web servers, FlexAutoDock integrates curated repositories of medicinal plant- derived phytochemicals, organism-derived bioactive compounds, and synthetic compounds from the ZINC database while supporting flexible ligand acquisition through medicinal plant or organism selection, chemical names, SMILES strings, PubChem CIDs, and drug-name queries. The platform further streamlines the docking workflow through automated protein structure retrieval from the Protein Data Bank and AlphaFold databases, receptor and ligand preparation, chain-specific protein selection, blind and site-specific docking, interactive visualization of predicted protein-ligand complexes, and scalable multi-target virtual screening. The resulting platform enables rapid, flexible, and large-scale virtual screening while simplifying the molecular docking workflow, providing researchers with an accessible computational resource for accelerating early-stage drug discovery. FlexAutoDock offers a fast, reliable, and accessible computational platform for molecular docking and virtual screening, freely available to the scientific community at http://103.99.177.82:3000/.

With the growing use of machine-learning-assisted pipelines for designing, characterizing, and optimizing biomolecules, the reliability of structure prediction models is increasingly important. PolyFold is a benchmarking framework developed to evaluate open-use structure prediction models, Boltz-2 and OpenFold 3, as commercially accessible alternatives to AlphaFold 3. We outline an end-to-end workflow automation tool to streamline input file creation, batch automation, and comprehensive analysis of model outputs for leading open-use structure prediction models. We curated an evaluation dataset of several thousand high-quality Protein Data Bank structures, homology-filtering against the training sets of both models to ensure a fair analysis. We then implemented an evaluation pipeline incorporating structural metrics (RMSD, TM-score, lDDT, etc.), interface metrics (DockQ, ilDDT, iRMSD, etc.), and physicochemical realism checks (based on bond lengths, angles, molecular internal energies, etc.). We identify key performance disparities, observing that Boltz-2 is generally superior to OpenFold 3, though the differential is partially attributable to residual homology leakage not accounted for by prevailing test set curation practices. We thus recommend a new method for homology-reducing when building a test set using length-weighted average fractional identity cutoffs rather than lowest chain fractional identity cutoffs. Even in eliminating residual leakage, Boltz-2 still performs better on full-set comparisons and a variety of important partitions (nucleic acids, protein-ligands, Ab-Ags, etc.). Both models are strong at folding monomeric structures, though struggle with homomultimer placement and small molecule physical realism, demonstrating enduring limitations of machine learning methods. This work is the first end-to-end, open-use, and reproducible platform for systematically assessing state-of-the-art structure prediction models. PolyFold enables practitioners to determine how models compare in performance on specific inference tasks and supports the broader adoption of accessible computational tools to facilitate biomolecular science.

Structure-based scoring functions leveraging machine learning have recently demonstrated superior performance over classical scoring functions, particularly on virtual screening benchmarks. However, due to the fundamental differences between their underlying model principles and architectures, it remains unclear to what extent performance stems from an understanding of molecular binding or from exploitation of systemic biases. Thus, disentangling the factors underlying benchmark performance is essential for determining whether a scoring function will generalize to novel chemical space and succeed in prospective drug discovery. To address this need, we present a case study investigating the nature and impact of systemic biases on benchmark comparisons between different scoring function paradigms. By systematically analyzing the evaluation workflows of prominent models, we reveal pocket bias, a form of spatial coordinate frame leakage arising from static binding pocket extraction, which artificially inflates benchmark performance. To progressively eliminate these sources of bias, we benchmarked two selected graph neural network scoring functions against two minmalist machine learning models and a classical scoring function under four increasingly stringent evaluation levels, successively removing pocket bias, reducing structural data leakage, and finally evaluating on out-of-distribution (OOD) protein targets. Upon removal of pocket bias and structural data leakage, the performance of all machine learning models dropped substantially. When evaluated on out-of-distribution protein families, the classical baseline AutoDock Vina outperformed the machine learning models in five of seven virtual screening tasks and dominated the docking power evaluation. Our findings indicate that benchmark performance can be heavily shaped by evaluation design and dataset artifacts, potentially overshadowing algorithmic improvements. While the tested machine learning models remain heavily dependent on encountering familiar data distributions to achieve competitive results, AutoDock Vina demonstrated superior generalization capacity on OOD targets. This work underscores the critical need for rigorous, artifact-free benchmarking protocols to guide the development of truly prospective machine learning models for virtual screening.

Genome mining is a powerful technique in natural product discovery, where biosynthetic gene clusters that are likely to produce novel or desirable natural products are identified through bioinformatic analysis. There are many more predicted biosynthetic gene clusters than can easily be experimentally characterized. Additional computational methods to prioritize biosynthetic gene clusters by the bioactivity, structural properties, or novelty of the product would make genome mining more efficient. Multiple machine learning/artificial intelligence models have been developed to predict product properties from biosynthetic gene cluster sequence, but they are limited by small quantities of training data. Model pretraining with unlabeled data is a powerful technique to develop models that can learn on a limited amount of labeled training data. Biosynthetic gene clusters are well suited to this strategy because there are many predicted clusters with only a small percentage being characterized. This paper reports BGC-MLM, a foundation model that is pretrained with a masked language task on predicted biosynthetic gene clusters and then fine-tuned for downstream applications including prediction of product structural class, bioactivity, chemical properties, counts of functional groups, and chemical fingerprint. Comparison to a model trained without pretraining shows that pretraining generally improves performance. BGC-MLM shows better or similar performance to existing specialized methods for these tasks, demonstrating its utility as a foundation model for natural product genome mining.