Computational Drug Discovery Methods

Cirrina: LLM-driven pharmacological reasoning agent enables preclinical CNS drug evaluation

Cirrina, an LLM-driven pharmacological reasoning agent, improves preclinical CNS drug evaluation by integrating mechanistic tools and providing better decision accuracy than deterministic pipelines.

High-throughput PBK modelling for dermal exposure: a pragmatic approach to predict systemic pharmacokinetics

This study uses high-throughput PBK modelling to predict systemic pharmacokinetics from dermal exposure using in silico data, achieving reasonable accuracy for key metrics.

Evolutionary exploration of drug-like chemical space utilizing generative AI and virtual screening

This paper presents an evolutionary algorithm framework combining generative AI and virtual screening to efficiently explore large chemical spaces and identify novel drug candidates.

Pan-Pharmacological Drug-Target Interaction Prediction with 3D-Informed Protein Encoding at Scale

OmniBind predicts drug-target interactions across multiple endpoints using structural protein encoding, outperforming existing methods and recovering known drug targets.

A shape-constrained regression and wild bootstrap framework for reproducible drug synergy testing

SIR improves drug synergy testing by using shape-constrained regression and bootstrap methods to provide reliable statistical inference and error control.

Cross-chemical and cross-species toxicity prediction: benchmarkingand a novel 3D-structure-based deep learning model

The paper benchmarks toxicity prediction models and introduces a 3D-structure-based deep learning model that improves classification accuracy.

Elucidating the potential carcinogenic molecular mechanisms of parabens in head and neck squamous cell carcinoma through network toxicology and molecular docking

This study identifies parabens' potential carcinogenic mechanisms in HNSCC through network toxicology and molecular docking, revealing key hub genes and immune correlations.

Towards automatic derivation of geometry-based descriptors as surrogates for complex computational approaches in enzyme-substrate prediction

The paper introduces geometry-based descriptors to predict enzyme-substrate interactions with high accuracy and interpretability, outperforming existing machine learning methods with minimal training data.

In silico evaluation of the effects of temperature on the affinity of the SV2C ligand UCB-1A to SV2 isoforms

This study evaluates how temperature affects UCB-1A binding to SV2A and SV2C using experimental assays and molecular dynamics simulations, finding that binding to SV2C remains stable while SV2A shows temperature sensitivity.

Discovery of novel covalent ligands with AlphaFold3

This study demonstrates that AlphaFold3 can effectively discover novel covalent inhibitors for kinases through virtual screening, outperforming existing methods and validating predictions with co-crystallography.

Predicting Toxicity and Bioactivity of the Chemical Exposome: A Case Study for the Blood Exposome Database

The paper predicts toxicity and bioactivity of chemicals in the Blood Exposome Database using machine learning models trained on Tox21 data, enabling scalable prioritization of chemical exposures.

MetaReact: A Reaction-Aware Transformer for End-to-End Prediction of Drug Metabolism

MetaReact predicts drug metabolites and enzyme selectivity using a reaction-aware Transformer, outperforming existing methods in accuracy and generalizability.

OmniBind: Proteome-Wide Promiscuity Predictions for Early-Stage Drug Screening

OmniBind predicts small-molecule promiscuity across the human proteome using a neural network, enabling faster early-stage drug screening with improved specificity over affinity alone.

Integrated Artificial Intelligence and Quantum Chemistry Approach for the Rational Design of Novel Antibacterial Agents against Ralstonia solanacearum.

This paper designs a novel antibacterial compound, Solres, using computational methods targeting Ralstonia solanacearum virulence proteins and validates its properties through docking and quantum chemistry.

Single-Pass Discrete Diffusion Predicts High-Affinity Peptide Binders at >1,000 Sequences per Second across 150 Receptor Targets

LigandForge generates high-affinity peptide binders rapidly across multiple receptor targets using a discrete diffusion model, outperforming existing methods in speed and binding quality.

AetherCell: A generative engine for virtual cell perturbation and in vivo drug discovery

AetherCell is a generative model that unifies clinical and experimental data to predict drug responses and repurpose existing drugs in human contexts.

Graph Neural Networks Model Based on Atomic Hybridization for Predicting Drug Targets

This paper develops a hybrid deep learning model combining graph neural networks and molecular descriptors to predict drug targets with high accuracy and interpretability.

Dynamic pharmacophores unveil binding mode ensembles for classical partial agonists at the M2 receptor

This study uses dynamic pharmacophores to analyze binding mode ensembles of M2 receptor partial agonists and shows correlations between ligand efficacy and spatial distribution of lipophilic contacts.

GCN-Mamba: Graph Convolutional Network with Mamba for Antibacterial Synergy Prediction

GCN-Mamba predicts antibacterial synergies by combining graph convolutional networks and Mamba models, outperforming classical methods and validating novel combinations in vitro.

Discovery of Semicarbazone and Thiosemicarbazone Analogs as Competitive SARS-CoV-2 Virus Main Protease (Mpro) Inhibitors

This study identifies semicarbazone and thiosemicarbazone analogs as competitive inhibitors of SARS-CoV-2 Mpro through virtual screening and biochemical validation.

From General-Purpose to Disease-Specific Features: Aligning LLM Embeddings on a Disease-Specific Biomedical Knowledge Graph for Drug Repurposing

The paper introduces CLEAR, a framework that aligns LLM embeddings with disease-specific knowledge graphs to improve drug repurposing, achieving state-of-the-art results on benchmark datasets.

Digital Twin Approaches for Interpretable Side Effect Prediction in Drug Discovery

This paper proposes a digital twin-based model for interpretable side effect prediction in drug discovery using off-target proteins and biophysical mechanisms, offering actionable insights.

Multi-Target In Silico Investigation of Withaferin A as a Potential Antiviral Inhibitor Against Key Marburg Virus Proteins

This study evaluates Withaferin A's antiviral potential against Marburg virus proteins using computational methods, showing favorable binding and ADMET properties.

TwinCell: Large Causal Cell Model for Reliable and Interpretable Therapeutic Target Prioritisation

TwinCell identifies therapeutic targets by analyzing cell state transitions using a causal model trained on in vitro data, outperforming existing methods in both virtual and clinical settings.

In silico drug repurposing and in vitro validation of cestode fatty acid binding proteins

The paper identifies potential drug candidates against echinococcosis by combining in silico screening and in vitro validation of FABP binders.

On improving experimental binding affinity predictions with synthetic data

The paper improves binding affinity predictions by using synthetic data, showing that combining synthetic and experimental data enhances model performance.

Unified Protein-Small Molecule Graph Neural Networks for Binding Site Prediction

YuelPocket predicts small molecule binding sites on proteins using a graph neural network with a global graph design, achieving high AUC-ROC scores on benchmark datasets.

Phenotypic Bioactivity Prediction as Open-set Biological Assay Querying

This paper introduces OpenPheno, a multimodal model that predicts bioactivity using phenotypic profiles and natural language descriptions, achieving strong zero-shot performance on unseen assays.

ToxiVerse: A Public Platform for Chemical Toxicity Data Sharing and Customizable Predictive Modeling

ToxiVerse provides a public platform for chemical toxicity data sharing and predictive modeling using curated datasets and machine learning tools.

BiGAT-Fusion: Node-Wise Gated Bidirectional Graph Attention for Drug Repurposing

BiGAT-Fusion improves drug-disease association prediction by addressing class imbalance, directional asymmetry, and static feature fusion through a novel graph neural network model.