Computational Drug Discovery Methods

Bento: Benchmarking Classical and AI Docking on Drug Design-Relevant Data

This study benchmarks 11 protein-ligand interaction prediction tools, showing classical and deep learning methods perform similarly on drug-like ligands, while co-folding tools excel on complex ligands and all struggle with generalization to unseen pockets.

A Mechanism-Aware Dual Attention Deep Model for Molecular-Protein Binding Affinity Prediction with Enhanced Generalizability and Interpretability

This paper introduces BindMecNet, a deep learning model for molecular-protein binding affinity prediction that improves generalizability and interpretability through explicit mechanistic modeling.

Benchmarking Free Energy Computational Methods for Revealing the Interactions Driving PARP1 Selective Inhibition

This study benchmarks computational methods to predict PARP1 vs PARP2 inhibitor selectivity, finding FEP and PMF outperform MM/PBSA in accuracy.

PocketX: Preference Alignment for Protein Pockets Design through Group Relative Policy Optimization

PocketX uses reinforcement learning to align protein pocket design with biophysical properties, outperforming existing methods on binding energy and evolutionary plausibility.

In Silico Screening of Neem-Derived Phytochemicals Targeting the Avr1 (SIX4) Effector Protein of Fusarium oxysporum for Fusarium Wilt Management

This study identifies neem-derived phytochemicals as potential inhibitors of the Avr1 effector protein in Fusarium oxysporum using in silico methods, with nimbiol and gedunin showing favorable properties.

The Human Omnibus of Targetable Pockets

The paper introduces a dataset of predicted binding pockets across the human proteome and a machine learning model to score and filter these pockets, outperforming existing methods.

Explainable Active Learning Framework for Ligand BindingAffinity Prediction

This paper develops an explainable active learning framework for ligand binding affinity prediction, showing that method choices affect outcomes but dataset characteristics set performance limits.

Structure-Guided Discovery of CHI3L1 Inhibitors from Ultralarge Chemical Spaces for Glioblastoma Therapy

This study identifies a CHI3L1 inhibitor, compound 9e, through structure-based virtual screening of a large chemical space, showing efficacy in a 3D GBM model.

CoBRA: Compound Binding Site Prediction using RNA Language Model

CoBRA predicts RNA ligand-binding sites using sequence-based embeddings from RNA language models, outperforming structure-based methods without requiring structural data.

Prompt-to-Pill: Multi-Agent Drug Discovery and Clinical Simulation Pipeline

This paper introduces a multi-agent AI framework for drug discovery and clinical simulation, demonstrating its use case with DPP4 targeting.

Knowledge-guided machine-learning and reverse screening combined method to predict cancer cell line responses to cytotoxic molecules

The paper develops a method combining machine learning and reverse screening to predict cancer cell line responses to cytotoxic molecules, achieving 71% accuracy in identifying experimental targets.

FragmentScope - exploring the fragment space with learned surface representations

FragmentScope uses learned protein surface representations to guide fragment placement and small molecule design, achieving accurate placement and promising lead compounds.

eMOSAIC: Multi-modal Out-of-distribution Uncertainty Quantification Streamlines Large-scale Polypharmacology

The paper introduces eMOSAIC, a method for uncertainty quantification in multi-target ligand binding affinity predictions, showing it outperforms existing methods in out-of-distribution settings.

Benchmarking Chemical, Genetic, and Cell Line Encodings for Cancer Perturbation Response Prediction

The paper benchmarks encoding methods for chemical, genetic, and cell line perturbations to predict cancer cell viability and gene dependency.

ANUBI: A Platform for Affinity Optimization of Proteins and Peptides in Drug Design

ANUBI optimizes protein and peptide drug candidates by simulating sequence space and binding affinity calculations, improving target binding efficiency and reducing costs.

DeepDOX1: A Dual-Drive Framework Integrating Deep Learning and First-Principles Physics for Drug-Protein Affinity Prediction

DeepDOX1 integrates deep learning and quantum mechanics to predict drug-protein affinity, outperforming existing models in real-world and challenging drug design scenarios.

Utilising multi-modal data-driven network analysis to identify monotherapy and combinational therapy targets in SOX2-dependent squamous cell lung cancer

The paper develops a computational pipeline to identify drug targets for SOX2-dependent lung cancer using multi-modal data and validates them in vitro.

Deep Learning-Based Drug Repurposing Using Knowledge Graph Embeddings and GraphRAG

This paper develops a deep learning framework for drug repurposing using knowledge graph embeddings and LLMs to provide explainable predictions and drug-disease relationship explanations.

DRGSCROLL: Achieving Full Side-Chain Flexibility in Docking Simulations through Genetic Algorithm Framework

DRGSCROLL improves docking simulations by optimizing ligand pose and receptor side-chain flexibility using a genetic algorithm, showing better performance in energy and docking scores compared to existing methods.

Rethinking Molecular Beauty in the Deep Learning Era

This paper introduces OiiSTER-map, a 2D heatmap to evaluate molecular quality in generative chemistry by combining bioactivity and structural features, aiming to improve drug discovery assessment.

SigSpace: an LLM-based agent for drug responsesignature interpretation

SigSpace uses an LLM to interpret drug response signatures from cancer cell lines, generating summaries that outperform random baselines in matching to drug identity or mechanism of action.

A Unified Multi-modal LLM for Dynamic Protein-Ligand Interactions and Generative Molecular Design

BioDynaGen unifies protein sequences, ligand SMILES, and text into a single model for predicting binding and generating molecules.

SLAB: Simultaneous Labeling And Binding affinity prediction for protein-ligand structures

The paper introduces a joint regression and classification model to predict protein-ligand binding affinity and activity, improving accuracy using PDBbind and flavivirus protease datasets.

Structure-based discovery of Saponarin as a broad-spectrum allosteric inhibitor of banana viral coat proteins

This paper identifies Saponarin as a broad-spectrum antiviral compound targeting coat proteins of multiple banana viruses through in silico methods.

Integrative Transcriptomics and Phytochemical Screening Reveal Pratenol B, Eriodictyol, Losbanine, and Isookanin, as Potential EGFR and HRAS Inhibitors in Indian Oral Squamous Cell Carcinoma Patients

This study identifies four phytochemicals as potential EGFR and HRAS inhibitors in Indian OSCC patients through transcriptomics and phytochemical screening.

Novel Computational Pipeline to Identify Target Sites for Broad Spectrum Antiviral Drugs

This paper develops a computational pipeline to identify conserved antiviral target sites in alphaviruses using AI-driven methods, validating potential drug targets through mutagenesis.

Best practices to cluster large molecular libraries

This paper introduces strategies to optimize BitBIRCH clustering for large molecular libraries by adjusting similarity thresholds and branching factors.

Computational Repurposing and Phytochemical Screening of Inhibitors Against ClpK from Klebsiella pneumoniae.

This study computationally screens natural and synthetic compounds for their ability to inhibit ClpK from Klebsiella pneumoniae using ADMET screening, molecular docking, and dynamics simulations.

Adhirons are efficient tools to guide antiviral ligand discovery

This study develops a pipeline using Adhirons to guide small molecule discovery against viral targets, identifying hits for Crimean-Congo Haemorrhagic Fever Virus and Influenza A Virus.

HTS-Oracle: Experimentally validated AI-enabled prioritization for generalizable small molecule hit discovery

HTS-Oracle uses AI to prioritize small molecule hits for challenging immune targets, reducing screening needs and increasing hit rates significantly.