diffusion model

SMFnet reveals principles of combinatorial transcription factor assembly

The paper introduces SMFNet and PRISM to analyze how transcription factor assemblies influence chromatin accessibility and identifies sequence-encoded principles of combinatorial TF assembly.

IDPForge: Deep Learning of Proteins with Global and Local Regions of Disorder

IDPForge generates all-atom ensembles of intrinsically disordered proteins and regions using a transformer diffusion model, showing good agreement with experimental data.

Predicting conformational changes induced by biomolecular interactions

This paper introduces K-Fold, a method that improves prediction of conformational changes in protein complexes by modeling structural transitions from unbound to bound states.

Accelerated discovery and optimization of a potent andselective MMP12 inhibitor with the machine learningplatform CAMOLE-1

This paper describes the development of CAMOLE-1, a machine learning platform for designing potent and selective MMP12 peptide inhibitors, achieving results in about one year with 270 synthesized peptides.

RiboDFM: Robust RNA sequence design using backbone-conditioned discrete flow matching

RiboDFM designs RNA sequences that fold into target 3D structures using a backbone-conditioned discrete flow-matching framework, achieving high sequence recovery and tertiary-structure consistency.

Complex enzyme active site scaffolding by iteratively detuned catalytic guidance

CaGE alternates guided structure prediction with sequence redesign, fading guidance each cycle, so sequences recover complex catalytic sites unaided; designed nonheme-iron and PLP-dependent enzymes were active.

Computational design of metalloproteases

This paper designs de novo metalloproteases using RoseTTAFold Diffusion 2, achieving high activity and specificity for cleaving various substrates.

Pocket-Conditioned Generation and Down-Selection of Non-Oxime Reactivators

This paper compares generative models for designing non-oxime reactivators of paraoxon-inhibited acetylcholinesterase, finding that down-selection methods outperform DrugDiff in practical drug-likeness and synthesis feasibility.

Large scale prospective evaluation of co-folding across hundreds of Mac1-ligand complexes and three virtual screens

On 489 evaluable new Mac1 complexes, AlphaFold3, Chai-1 and Boltz-2 placed over half of ligands within 2 Å, but on docking hit-lists AF3 and Boltz-2 added little over docking.

AlphaFold3 and Intrinsically Disordered Proteins: Reliable Monomer Prediction, Unpredictable Multimer Performance

AlphaFold3 pLDDT matches AlphaFold2 at calling disordered residues (Matthews correlation 0.693 vs 0.701); on disorder-containing complexes mean DockQ is similar, but interface features explain less variance (42% vs 58%).

RGI-Toolkit: Differentiable Restraints for Controllable Biomolecular Structure Prediction

RGI-Toolkit enables controllable biomolecular structure prediction by integrating experimental restraints into diffusion models without retraining.

Generative machine learning unlocks the first proteome-wide image of human cells

A conditional latent diffusion transformer predicts immunofluorescence images of 12,800 human proteins from nucleus, ER and microtubule stains, outperforming PUPS and CELL-Diff, including on an unseen drug-treated cell line.

siProGenA: Generative siRNA Candidate Construction via Position Proposal and Guide Generation

siProGenA generates siRNA candidates by proposing positions and generating guide sequences, achieving high Hit@1 and Hit@5 scores in controlled evaluations.

GDA-Pred: Generative AI-Driven Data Augmentation for Improved Prediction of IL-6 and IL-13 Inducing Peptides

GDA-Pred improves prediction of IL-6 and IL-13 inducing peptides by using generative AI to augment small datasets, enhancing classifier performance through data expansion.

Enhanced sampling of protein conformations in AlphaFold3 with repulsive bias in the diffusion generative model

This paper improves AlphaFold3's ability to predict protein conformations by introducing a repulsive bias in its diffusion model, enabling better capture of ligand-induced structural changes.

bulk2scDiff: A Pseudobulk-Conditioned Diffusion Model for Bulk-to-Single-Cell RNASeq Generation

bulk2scDiff generates single-cell RNA-seq data from bulk profiles using a diffusion model, showing good reconstruction of cell populations in cancer datasets.

The Synthetic Fidelity-Stability Framework (SFSF): A Systematic Multi-Dimensional Benchmark of Synthetic Clinical Laboratory Data Generators

The paper benchmarks 17 synthetic data generators for clinical lab data, finding no single model excels across all metrics, and highlights ForestDiffusion's best privacy-utility balance.

Brain2VLM: Hierarchical Alignment Between Cortical Representations and Vision-Language Latent Spaces

A residual MLP beats ridge by Δ≈0.47 correlation decoding CLIP embeddings from higher visual areas, but only Δ≈0.05 to 0.06 decoding Stable Diffusion latents from early visual cortex.

Biasing Conformational Sampling in AlphaFold 3 and Boltz-2 via Pair Representation Scaling

Multiplying the AlphaFold 3 or Boltz-2 pair representation by a single scalar (1+β) at the Pairformer input, without retraining, broadens sampled ensembles toward alternative states across 86 two-state targets.

Boltz-Perturb: Improving Diversity and Accuracy in Protein-Ligand Co-Folding through Training-Free Conditioning Perturbation

Boltz-Perturb improves ligand binding pose prediction by introducing inference-time perturbations, achieving higher success rates with less computation.

Seed-Guided De Novo Design Expands the Structural Diversity of Antitoxin Protein Binders

This paper introduces seed-guided diffusion-based design to improve structural diversity and target contact in antitoxin protein binders, achieving functional binders with enhanced selectivity.

Beyond Generic Signal Peptides: ApexSP Enables Cargo-Specific Secretion Design

ApexSP designs cargo-specific signal peptides for efficient secretion by integrating evolutionary information, compatibility modeling, and experimental validation across multiple expression systems.

Diffusion models learn underlying trends in actomyosin networks and predict behavior at unseen filament turnover

The paper trains a diffusion model on actomyosin system data to predict nonlinear trends in cortical flow at unseen filament turnover rates.

Targeting Ligand Shape While Preserving a Warhead through Guided Diffusion and Adaptive Staged Growth

This paper improves molecular growth models by using shape guidance and adaptive growth to connect fragments while preserving a warhead.

Programming protein shape as an explicit design layer via CAD blueprint-guided diffusion

The paper integrates CAD blueprints with diffusion models to explicitly program protein shapes, demonstrating emergent functions through molecular design.

ProteinSketch translates spatial intuition into protein design with bimanual interaction in VR

ProteinSketch enables interactive protein design in VR by translating spatial intuition into constraints for AI-driven protein structure generation.

A Glycan-Aware Diffusion Model for Carbohydrate and Glycoprotein Structure Prediction

This paper introduces SweetFold, a diffusion model for predicting glycan and glycoprotein structures, showing improved performance over existing models.

F.A.D.E. (Fully Agentic Drug Engine): A Conversational AI Platform for Drug Discovery

F.A.D.E. is a conversational AI platform that generates drug candidates from natural language queries, outperforming reference ligands in drug-likeness for two targets.

AFM-Fold: Rapid Reconstruction of Protein Conformations from AFM Images

AFM-Fold reconstructs 3D protein conformations from AFM images using a generative AI framework, outperforming rigid-body fitting and capturing conformational dynamics.

PertDiffBench: Benchmarking Diffusion Models for Single-Cell Perturbation Response Prediction

PertDiffBench benchmarks diffusion models for predicting transcriptional responses to perturbations in single-cell and bulk RNA-seq data, showing mixed performance across different tasks and settings.