Skip to main content
recep.adiyaman
Vision About Experience Education Skills Publications Tools Newsletter Updates Podcasts Blog Contact CV Subscribe
All Issues
Older Newer
Daily Signal June 10, 2026 · 9 min read

Issue #127: BA-Pred and RMSD-Pred: Integrated Graph Neural Network Models for Accurate Protein-Ligand Binding Affinity and Binding Pose Prediction.

Protein Design Digest #127: BA-Pred and RMSD-Pred: Integrated Graph Neural Network Models for Accura…

Share X LinkedIn
Protein Design Daily

Building something in Protein Design?

I love collaborating on new challenges. Let's build together.

LinkedIn Email

Subscribe to Protein Design Digest

Daily curated signals from arXiv, PubMed, and BioRxiv.

Signal of the Day

BA-Pred and RMSD-Pred: Integrated Graph Neural Network Models for Accurate Protein-Ligand Binding Affinity and Binding Pose Prediction.

Accurate prediction of protein-ligand bound poses and their affinities is essential in structure-based drug discovery. Here, we present an integrated deep-learning framework that disentangles the two core tasks─affinity estimation and pose evaluation─within complementary graph neural network architectures. BA-Pred is for predicting binding affinity and RMSD-Pred is for binding pose assessment, predicting the root-mean-squared deviation of ligand poses from crystal structures. Both models employ a Gated Graph Convolutional Network with Learnable Structural Positional Encoding (GatedGCN-LSPE) architecture to capture spatial and chemical dependencies across protein-ligand graphs. BA-Pred achieved state-of-the-art scoring power on the CASF-2016 benchmark with a root-mean-squared error of 1.10 p K d , while RMSD-Pred exhibited strong docking power with a top-1 success rate of 96%, comparable to the best reported deep-learning scoring functions. The robust generalization capability of RMSD-Pred was further validated on the external Astex diverse set and PoseBusters benchmarks, where it significantly improved the pose selection success rates of AutoDock-GPU by up to 33.1%. The accuracy of our methodology was demonstrated on pharmaceutical targets in the 16th Critical Assessment of Structure Prediction, where our approach ranked second in the ligand binding affinity prediction category. By using our models, an integrated pipeline was developed for virtual screening, where pose selection was performed with RMSD-Pred and binding affinities were predicted with BA-Pred. This combined approach demonstrated robust screening performance, achieving an enrichment factor (EF) 1% of 21.1 on the CASF-2016 benchmark. Furthermore, on the LIT-PCBA benchmark, rescoring poses docked by AutoDock-GPU with our pipeline significantly improved the EF 1% from 2.18 to 3.19. These various benchmark results demonstrate that our graph-neural network models show good and balanced performance in diverse protein-ligand interaction prediction tasks. Thus, we expect that our models will serve as a promising framework to accelerate the drug discovery process.

Why this matters: Critical for improving fold accuracy and reducing structural uncertainty in de novo design.

Also Worth Reading

Computational Assessment of Phytochemical Inhibitors of Cytochrome P450 2E1 (CYP2E1) Implicated in Hepatotoxicity: Comparative Molecular Docking and Interaction Analysis Using Silymarin as a Benchmark

Abstract Liver diseases remain a major global health challenge, often driven by oxidative stress and reactive metabolites generated during xenobiotic metabolism. Cytochrome P450 2E1 (CYP2E1) plays a central role in hepatotoxicity through the bioactivation of toxic compounds and excessive production of reactive oxygen species. This study aimed to evaluate the inhibitory potential of selected phytochemicals (curcumin, quercetin, and kaempferol) against CYP2E1 using molecular docking, with silymarin as a benchmark compound. Molecular docking was performed using SwissDock, while UCSF Chimera was used for three-dimensional visualization and interaction confirmation. BIOVIA Discovery Studio Visualizer was employed for detailed 2D interaction profiling, including hydrogen bonding and hydrophobic interaction analysis. Binding affinities were evaluated based on estimated free energy values (ΔG), and interaction patterns were analyzed to assess ligand stability within the CYP2E1 active site. Results revealed that curcumin exhibited the strongest binding affinity (− 6.9889 kcal/mol), outperforming the benchmark ligand silymarin (− 6.5559 kcal/mol), followed by kaempferol (− 6.3145 kcal/mol) and quercetin (− 6.0048 kcal/mol). Curcumin also formed the highest number of hydrogen bonds and demonstrated favorable hydrophobic and short-range interactions within the active site. These combined interactions suggest strong ligand–enzyme stability and effective binding within the CYP2E1 catalytic pocket. On the whole curcumin demonstrated the most promising inhibitory potential against CYP2E1 among the evaluated phytochemicals, suggesting its possible role as a natural hepatoprotective agent. However, further molecular dynamics simulations, in vitro enzymatic studies, and in vivo validation are required to confirm its therapeutic efficacy.

Tocotrienol as a multi-target inhibitor of ICAM-1, VCAM-1, and E-selectin: Comparison using AutoDock and GNINA docking with molecular dynamics simulation.

Atherosclerosis is a chronic inflammatory disease characterized by endothelial dysfunction and leukocyte adhesion, mediated by cell adhesion molecules such as E-selectin, intercellular adhesion molecule-1 (ICAM-1), and vascular cell adhesion molecule-1 (VCAM-1). Tocotrienols, a subgroup of vitamin E, exhibit potent antioxidant and anti-inflammatory properties, suggesting their potential role in attenuating atherosclerosis. This study comparatively evaluated the binding affinities and molecular interaction profiles of α-, β-, γ-, and δ tocotrienol isomers towards E-selectin, ICAM-1, and VCAM-1 using molecular docking approaches, followed by molecular dynamic simulation to assess the stability of the top-ranked protein-ligand complexes. The docking experiment was conducted using MolModa, an automated molecular docking platform based on AutoDock Vina and convolutional neuronal network (CNN)-based AI-assisted GNINA. Overall, the conventional molecular docking tool AutoDock Vina results showed that all tocotrienol isomers exhibited the strongest average binding affinities to VCAM-1. Among the isomers, α-tocotrienol displayed the highest binding affinity towards E-selectin (-6.69 ± 0.00 kcal/mol) and ICAM-1 (-6.79 ± 0.00 kcal/mol), whereas β-tocotrienol exhibited the strongest affinity toward VCAM-1 (-7.59 ± 0.00 kcal/mol) in the molecular docking analysis using conventional molecular docking tool AutoDock Vina. In contrast, the AI-assisted molecular docking tool GNINA leveraging deep learning, demonstrated a more accurate and consistent affinity profile by consistently identified β-tocotrienol as the most favorable binder toward E-selectin (-6.91 ± 0.01 kcal/mol) and ICAM-1 (-7.08 ± 0.90 kcal/mol), characterized by hydrogen bonding, hydrophobic interactions, and extensive van der Waals forces, that are crucial for the lipid-soluble ligand. The AI-assisted molecular docking tool GNINA docking for VCAM-1 was not generated due to structural limitations of the receptor model. Molecular dynamics (MD) simulations over 200 ns demonstrate a significant stabilizing interaction with GLU87, whereas the hydrogen bonding at ASP178 was found to be intermittent and contributory throughout the trajectory. This study provides the first comprehensive computational evidence differentiating the multi-target potency of tocotrienol isomers in targeting inflammatory and vascular-related pathways. Further experimental validation is warranted to confirm these in silico predictions and explore their biological significance.

Discriminator-Guided Inverse Folding for Multi-Property Protein Design.

Designing proteins for real-world applications requires the simultaneous satisfaction of multiple physicochemical properties. Structure-based de novo protein design has become the prominent design paradigm, successfully creating numerous proteins. Property optimization is commonly introduced during the sequence generation stage of protein design, i.e., inverse folding. Existing methods primarily rely on fine-tuning inverse folding models to design sequences with desired characteristics. However, multi-property optimization through fine-tuning demands datasets annotated with multiple properties-resources that remain extremely limited. Consequently, structure-based protein design has not yet achieved joint optimization of multiple properties. Here, we present Discriminator-Guided Inverse Folding (DGIF), a framework that guides the inverse folding model by adjusting its internal history states through an auxiliary discriminator module. The discriminator integrates multiple property predictors, each trained independently on a single-property dataset, thereby enabling multi-property optimization in the absence of datasets annotated with multiple properties. In addition to substantial improvements in key traits like thermostability and solubility, DGIF can generate protein sequences optimized for both properties simultaneously, with the designed proteins shifting markedly toward the Pareto front that represents optimal trade-offs. Experimental results validate the effectiveness of DGIF for multi-property protein design.

Research & AI Updates

From the Industry

Quick Reads

Integrating Network Pharmacology and Molecular Docking to Investigate the Action Mechanism of Poria Almond and Liquorice Decoction in Chronic Heart Failure: A Computational Prediction Study.

Chronic heart failure (CHF) represents the end-stage progression of cardiac diseases, and its prognosis remains suboptimal. Read more →

Molecular Docking and Dynamics Simulations of Natural Limonoids Interacting with the Nipah Virus Attachment Glycoprotein.

Introduction The Nipah virus (NiV) belongs to the Paramyxoviridae family and is a zoonotic pathogen associated with severe respiratory disease and encephalitis. Read more →

Electrochemical Investigation of Cu(II) Complexation with a Novel Pyridine-Based Schiff Base Ligand: Cyclic Voltammetry, Molecular Docking, and Antioxidant Evaluation.

This study investigates the electrochemical interaction between CuCl 2 and a pyridine-based Schiff base ligand in aqueous KCl using cyclic voltammetry at 302.15 K. Read more →

Exploring Ligand Flexibility in Nucleic Acid Scaffolds Using Graph Neural Networks.

Interactions between nucleic acids and ligands are vital for gene regulation and therapy development, yet accurate modeling is hindered by shallow, flexible pockets and conformational changes. Read more →

Study on the mechanism of action of Chaihu Guizhi Ganjiang Decoction for the treatment of slow transit constipation combined with depression based on network pharmacology and molecular docking.

Based on network pharmacology and molecular docking technology, this study investigated the mechanism of action of Chaihu Guizhi Ganjiang Decoction (CGGD) in the treatment of slow transit constipation (STC) combined with depression. Read more →

Nonparametric Riemannian Empirical Bayes, and Denoising Measurements on Manifolds

We initiate the study of nonparametric empirical Bayes denoising methods in the setting where both the latent variables and their measurements lie on a compact Riemannian manifold, and where the likelihood is a Riemannian Gaussian distribution. Read more →

Screening of putative inhibitors for cytotoxin VacA from Helicobacter pylori.

Cytotoxin VacA, a major virulence factor of Helicobacter pylori-the principal etiological agent of gastric cancer-represents a compelling therapeutic target. Read more →

Corrigendum to: Potential Anti-inflammatory Properties of Corydalis rhizoma for Co-Treatment of Neuropathic Pain and Major Depressive Disorder: Network Pharmacology with Molecular Docking and Molecular Dynamics Simulation Approaches.

In the published version of the article [1], an error occurred in the numbering of the section headings, which has now been corrected. Read more →

Pipeline Tip

Use Snakemake for reproducible end-to-end protein design workflows.

Resources & Tools

Deep learning is not a magic wand, but a powerful lens for structural biology. — Recep Adiyaman

BS HF DK