Computational Drug Discovery Methods

Ligand Binding Free Energy Landscapes at the Tubulin Colchicine Site from Coarse-Grained Metadynamics

This study evaluates coarse-grained funnel metadynamics for modeling ligand binding at the tubulin colchicine site, showing improved statistical convergence and reasonable accuracy compared to all-atom simulations.

Chemi-Proteome Language Attention Network Empowers Fragment-Based Ligand Interactome and Binding Sites Discovery with Evidence

C-PLANK improves ligand-protein interaction prediction by integrating cellular context and fragment-based chemoproteomics data, outperforming existing methods and identifying novel ligands.

HANSEN: An Integrated Structural and Functional Proteome Resource for Structure-Guided Drug Discovery in Mycobacterium leprae

Web resource offering predicted monomeric and oligomeric structures for 1,603 M. leprae proteins, with pockets, epitope scores, essentiality and a target priority ranking, plus a nanopore resistance-calling module.

Can SMILES be fragmented into a concatenable ordered sequence of retrosynthetically interesting string block ?

The paper tests if SMILES can be fragmented into ordered blocks for retrosynthetic analysis, finding that it improves physico-chemical property reproduction but degrades LLM performance.

A pretrained unified model enables cellular functional profile prediction and multi-objective virtual drug screening

InsilicoCell predicts cellular functional profiles and enables multi-objective virtual drug screening by integrating diverse molecular data and tasks.

A closed-loop reinforcement learning framework for rapid compound directed optimization

This paper introduces RCDO, a closed-loop reinforcement learning framework that accelerates compound optimization by integrating experimental feedback with generative models, achieving significant improvements in drug candidate properties.

NELLY enables patient-centric drug prioritization through interpretable drug-conditioned gene weighting

This paper introduces NELLY, a deep learning model that prioritizes patient-specific therapies by integrating transcriptomic and chemical data, outperforming existing methods in drug response prediction.

A Medicinal Chemistry-Centered Evaluation of AlphaFold 3 and Boltz-2 Across Diverse Binding Modalities

This study evaluates AlphaFold 3 and Boltz-2 for predicting biomolecular structures across diverse binding types, finding they accurately predict some complexes but struggle with others like allosteric and membrane-protein interactions.

EcoXAI: Autonomous Agentic Ecosystem for Explainable Artificial Intelligence and Biomedical Discovery

EcoXAI introduces a multi-agent system for biomedical discovery that integrates knowledge graphs and agentic AI to identify novel drug candidates for Alzheimer's Disease.

BindScreen: Protein-Centric Contrastive Learning for Sequence-Based Virtual Screening

BindScreen improves virtual screening by using a protein-centric contrastive learning approach with asymmetric loss and batch design, achieving better performance with fewer computational resources.

MolJam: A Multidimensional Framework for Assessing Molecular Dataset Quality and Its Impact on Machine Learning

MolJam assesses molecular dataset quality across five dimensions and shows that improving quality metrics does not always enhance machine learning performance due to trade-offs with data quantity.

A multimodal representation learning platform for accurate molecular ADMET prediction

Trimole-Hybrid improves ADMET prediction by combining multiple molecular representations and task-specific models, achieving top performance on 22 benchmarks.

Point-in-time evidence and cross-area clinical precedent anticipate clinical entry across 100 focal areas: retrospective validation of the Intangia triage layer

The paper validates a triage system that ranks potential drug combinations based on historical evidence and clinical precedent, showing it can predict clinical entry with high precision across 100 focal areas.

Model Validation Protocols for Machine Learning in Small Molecule Drug Discovery

The paper presents a model validation framework for molecular property prediction models in drug discovery, revealing four failure modes and showing that common evaluation protocols overestimate performance.

Prioritization of chemical scaffolds using the TDR Targets database: an integrative workflow for Trypanosoma cruzi drug discovery

This paper prioritizes chemical scaffolds for Trypanosoma cruzi drug discovery using the TDR Targets database, identifying 378 high-priority compounds and validating 7 with selective trypanocidal activity.

Molecular size dominates α2-adrenergic subtype-selectivity benchmarks: five controls for reducing attrition in selective ligand design

The paper evaluates methods for predicting α2-adrenergic subtype selectivity, finding that molecular size is the primary factor, and introduces controls to improve benchmark reliability.

FlashBind: Towards Accurate and Efficient Structure-based Virtual Screening

FlashBind improves drug discovery by offering faster and accurate protein-ligand interaction predictions compared to Boltz-2, with real-world validation in antibiotic screening.

Causally-inspired meta-representation learning framework for predicting patient-specific clinical responses to drug combinations

CaMeRe predicts patient-specific clinical responses to drug combinations by integrating causal representation learning and meta-learning, showing robust performance across multiple datasets and cancer types.

A multi-agent molecular optimization framework leads to a rapid-recovery intravenous anesthetic candidate with an improved safety margin

The paper introduces MASCOT, a multi-agent framework for molecular optimization, which improves anesthetic candidates with better safety and recovery properties.

Structure-free, site-resolved contrastive learning extends small-molecule discovery beyond the reach of structure-based modeling

Ptarmigan-1 uses contrastive learning to predict small molecule-protein interactions without requiring 3D structure, outperforming traditional methods on challenging targets.

Mechanistic Dissection of Entropic Penalty upon Ligand Binding and Molecular Flexibility via Molecular Dynamics Simulations and Machine Learning

This study quantifies molecular entropy and its changes during ligand binding using MD simulations and machine learning, revealing key motions and their thermodynamic impacts.

Benchmarking Docking Protocols for GPCR Allosteric Modulators

This study benchmarks docking protocols for GPCR allosteric modulators using experimental structures and GaMD simulations, finding that ensemble re-ranking strategies improve early enrichment of known actives.

An AI System for Autonomous Algorithm Evolution in Drug Development

DrugEvolve autonomously evolves algorithms across drug development stages, improving performance on benchmark tests and showing generalizability across data modalities.

PandaDock: An Open-Source Molecular Docking Platform with Flexible-Ligand Search and Equivariant Neural Scoring

PandaDock is an open-source molecular docking platform that improves pose prediction and affinity scoring using a GNN and flexible-ligand search, achieving strong performance on benchmark datasets.

SILCS-Guided Feature Encoding Expands Ligand Recognition at an HBV Core Protein Interface

The paper identifies ligands that bind beyond a known hydrophobic pocket at the HBV core protein interface using molecular dynamics and fragment mapping.

OTRec: Deep learning recommender for prospective druggable disease-target associations

OTRec is a deep learning model that recommends druggable disease-target associations, outperforming existing methods in both retrospective and cross-validation settings.

Systematic Benchmarking of AI-Based Molecular Generation Models for Structure-Based Drug Design

This study benchmarks 12 AI-based molecular generation models for drug design across 176 protein-ligand systems, finding no single method optimizes all criteria and proposing a hybrid approach.

FlexAutoDock: A Flexible Platform for Automated Molecular Docking and Virtual Screening of Natural and Synthetic Compounds

FlexAutoDock is a cloud-based platform for automated molecular docking and virtual screening of natural and synthetic compounds, addressing limitations in existing tools by integrating curated repositories and streamlining workflows.

Beyond Chemical Similarity: Structure-Agnostic Drug-Drug Interaction Prediction with MeSH Semantics and a Drug-Target-Protein Knowledge Graph

The paper presents a structure-agnostic framework for predicting drug-drug interactions using MeSH semantics and a Drug-Target-Protein knowledge graph, showing improved performance with intermediate MeSH annotation depth.

LinkLlama: Enabling Large Language Model for Chemically Reasonable Linker Design

LinkLlama improves linker design by combining natural language prompts with 3D spatial awareness, achieving higher chemical validity than existing methods.