Computational Drug Discovery Methods

Heterogeneous graph neural networks with biological prior knowledge for interpretable drug repurposing in triple-negative breast cancer

PRECISION uses a heterogeneous graph neural network with biological knowledge to identify drug repurposing candidates in TNBC, showing associations between predicted drug sensitivity and survival in multiple cohorts.

How effective are contrastive learning-based approaches for activity-cliff prediction?

This study evaluates contrastive learning-based approaches for activity cliff prediction in drug discovery, finding they improve embedding separation but have dataset-dependent accuracy impacts.

Machine Learning-Guided Classification of Druggable Pockets and Phylogenetic Druggability Transfer Across the Human Kinome

The paper constructs a machine learning atlas of kinase inhibitor complexes to classify binding modes and demonstrates phylogenetic druggability transfer across the human kinome.

GEM-GPT Enables Personalized Cell Type-Resolved Therapeutic Design for Systems Pharmacology

GEM-GPT generates personalized therapeutic molecules to shift cell-type-specific disease states toward healthy phenotypes, outperforming baselines in chemical perturbation and CRISPR experiments.

Chemical Descriptors and Deep Learning Embeddings for Scoring de novo Peptide Designs

This paper compares chemical descriptors and deep learning embeddings for scoring de novo peptide designs, finding classical descriptors are effective and efficient for developability prediction.

A Computational Foundation Toward Targeting the ELMO1/DOCK2 Complex

The paper investigates the druggability of the ELMO1/DOCK2 complex and introduces tools for analyzing protein interaction and MD trajectories.

Closed-Loop Multi-Objective Optimization for Receptor-Selective Cell-Penetrating Peptide Design

The paper develops a closed-loop framework for designing receptor-selective cell-penetrating peptides using multi-objective optimization and evaluates its effectiveness through experimental validation.

PHAROS: turning single-cell perturbation models into target-directed drug-combination screens

PHAROS uses a pretrained single-cell perturbation model to predict and prioritize effective drug combinations without retraining, showing success in cell lines and patient-derived tumors.

Leveraging Neural Network Models for Drug Repurposing: A case study on Cardiac Hypertrophy

This paper uses a deep learning model to identify potential drugs for cardiac hypertrophy by analyzing gene expression data and linking disease signals to known drug targets.

Conformational Preference Classification of Integrin-Binding Ligands Using Free Energy Perturbation

This paper presents AB-FEP, a computational method to classify integrin-binding ligands by their conformational preference, showing high accuracy and insights into ligand mechanisms.

Early-Enrichment Hit Discovery via Reversible-Work c(t) Estimation in Metadynamics (CTMD)

This paper introduces CTMD, a physics-based method for early enrichment in virtual screening that outperforms approximate scoring functions and AI-based co-folding methods while being fast and transferable.

Predicted and experimental protein-ligand coordinates with distance labels and evaluation splits

The paper introduces PLI-Parallax, a resource that combines predicted and experimental protein-ligand coordinates with evaluation splits and accuracy estimates for 30,567 systems.

FUSED: A Functional Representation for Joint Structural and Elemental Analysis of Protein Ligand Binding Sites

The paper introduces FUSED, a functional representation for ligand binding sites that integrates structural and elemental data, achieving strong performance in classification and matching tasks.

PharmCast: rapid generation of three-dimensional pharmacophore fingerprints from two-dimensional structure without conformer generation

A feedforward network predicts 10,549-bit ensemble pharmacophore fingerprints from SMILES without conformers: 0.584 ms versus 5.71 s per pair comparison, with median MCC 0.860 to 0.914.

bioq: a unified, agent-native command-line interface to a fleet of AI drug-discovery methods

bioq provides a unified CLI to access a fleet of AI drug-discovery tools, enabling workflow automation without local setup or cloud configuration.

From Predicted Ki to Surrogate IC50: Similarity-Guided Empirical Calibration of Drug Target Affinity Predictions

The paper calibrates drug target affinity predictions to align with IC50 values using similarity-guided methods, reducing median fold error across kinase and non-kinase panels.

Integrated in silico identification of fungal-derived dual-target inhibitors for anti-schistosomal drug discovery

This paper identifies two fungal-derived compounds with dual inhibitory action against Schistosoma mansoni DHODH and cathepsin B1 using computational methods.

ReTIF: Granularity-Aware Multitask Interaction Routing for RNA-Compound Interaction Prediction and Binding-Site Localization

ReTIF improves RNA-compound interaction prediction and binding-site localization by using task-specific interaction representations, outperforming baselines in multiple evaluation scenarios.

OmniSyn unifies target-aware molecular generation and optimization within a synthesis-native LLM framework across the human proteome

OmniSyn generates and optimizes target-specific molecules with synthesis traces using a sequence-conditioned LLM, achieving state-of-the-art performance on drug discovery benchmarks.

When DL-Based Prescreening Meets Synthon-Based Docking: Target-Adapting PharmacoNet via MEL-Steered Correction

The paper adapts a general pharmacophore prescreening method to specific targets using fragment-docking data from the V-SYNTHES pipeline, improving its performance without additional data or retraining.

Thermodynamics and unbinding kinetics of A22 at multiple actin binding sites revealed by enhanced sampling simulations

This study uses enhanced sampling simulations to analyze A22-actin interactions, revealing binding affinities, residence times, and unbinding kinetics across multiple sites.

Open-Source DNA-Encoded Library Package for Design, Decoding, and Analysis: DELi

DELi is an open-source platform for DNA-encoded library design, decoding, and analysis, demonstrating its effectiveness through the identification of a BRD4 binder.

Drude SILCS-Nucleic: Harnessing Explicit Electronic Polarization in Targeting RNA and DNA for Drug Design

This paper extends the SILCS method with the Drude polarizable force field to improve ligand-nucleic acid interaction modeling and drug design.

Retrieval of binding sites across the AlphaFold human proteome using protein language model representations

PocketScope retrieves binding sites across the AlphaFold human proteome using protein language model representations and achieves top performance on a benchmark.

Comprehensive Evaluation of Protein Language Model Embeddings for Drug-Target Affinity Prediction

This study evaluates the impact of pre-trained protein language models on drug-target affinity prediction, showing that simple architectural changes can match large PLMs' performance.

Gaussian accelerated Molecular Dynamics - Thermodynamic Integration (GaMD-TI): Improved alchemical free energy calculations with enhanced sampling

This paper introduces GaMD-TI, a method combining Gaussian accelerated MD with TI to improve free energy calculations by enhancing conformational sampling in alchemical transformations.

CROWN: Curated Repository Of Well-resolved Noncovalent interactions

An automated pipeline turns the PDB into 178,263 repaired, protonated, restraint-minimized protein-ligand complexes; scoring functions trained on them beat HiQBind-trained ones on CASF-2016 ranking and docking power.

Predicting the toxicity of chemical compounds via Hyperdimensional Computing

This paper uses Hyperdimensional Computing to predict chemical compound toxicity efficiently, showing it can classify compounds as toxic or non-toxic with computational advantages over traditional methods.

The first OpenBind release: An open experimental structure-affinity dataset and benchmark for structure-based AI

A public crystallographic structure and affinity dataset for enteroviral 2A protease (925 binding events, 699 compounds), with docking, cofolding and screening baselines. Fragment fine-tuning of OpenFold3-p2 improves follow-on predictions.

Identifying a logical specification and a program for an LLM-based generator of lead molecules

The paper develops an LLM-based method for generating feasible lead molecules for DBH inhibitors, achieving comparable affinity to clinical trial drugs.