TW

Novelty: Transforms controlled 3D molecule generation into a data-efficient latent space optimization problem; this model is more data-efficient than the standard 3D diffusion model.

3D equivariant diffusion decoder with a semantic encoder to control molecule generation. It reformulates property guidance as latent space optimization prior to 3D decoding via a denoising process #GenerativeAI #DrugDesign nature.com/articles/s4146…

Novelty: An edge-fusing architecture conditions generation on directed graphs encoding local stereochemistry, enabling the targeted prediction of specific diastereomeric transition states Limitation: GFN2-xTB for TS :)

Flow-matching model, trained on 2 million GFN2-xTB reaction pathways for predicting 3D transition state geometries directly from 2D reactant-product connectivity graphs. chemrxiv.org/doi/full/10.26… #CompChem #MachineLearning #Chemistry

Stochastic interpolants for fragment-based molecular design, explicitly training on conditional fragment masking. Illustrated to perform better than repurposing unconditional models for structure-based generation. #Cheminformatics #DrugDesign pubs.rsc.org/en/content/art…

1 Clone the repo. 2. Build the doc locally following contrib/reinvent-doc/BUILD.md

Comprehensive tutorial and guide for REINVENT4, a suite of generative AI for molecular design. github.com/MolecularAI/RE…

A web app for monitoring property optimization and chemical diversity in generative molecular design. #Cheminformatics #GenerativeAI pubs.rsc.org/en/content/art…

🚨 BREAKING: Claude can now build your entire resume and LinkedIn profile like a $500/hour executive recruiter from Robert Half. For free. Here are 12 prompts that get you interview calls within 7 days: (Save this before it disappears)

Molecular design for AI agents: announcing the Tamarind MCP Server. Today, scientists can use the @tamarindbio library of 250+ molecular design tools(Boltz, AlphaFold, RFdiffusion...) in any AI chat interface. We serve not just open-source models, but the internal protocols your team has onboarded to Tamarind. Any tool added to Tamarind is then available across the MCP server, Tamarind web app, and API, so it can be used in chat-based agents, multi-step pipelines, ELNs, and LIMS-connected workflows. Our goal is simple: make Tamarind the place where scientists can access the BioAI tools they need, wherever they want to work, while we handle the infrastructure. Many users have already incorporated our MCP into their internal AI agents, along with community efforts like Blatant-Why building apps on top of Tamarind. Try out our tooling for antibody design, small molecule virtual screening, developability/ADMET scoring and more!

@generativeai Novelty: It splits the generation into a high-level space group selection policy, followed by a low-level atom-lattice placement policy, optimized against a surrogate physics-informed or objective-driven reward toward a stable crystal structure with target properties.

A hierarchical generative flow network for generating 3D crystal structures constrained by space group symmetries and lattice parameters. pubs.rsc.org/en/content/art… #MaterialsDiscovery @GenerativeAI

Novelty: It overcomes the mismatch and posterior collapse common in continuous latent space with discrete binary space. It also fundamentally transforms molecular generation from a stochastic search into a physics-inspired quadratic unconstrained binary optimization problem.

A generative framework that maps molecules into a discrete binary space, enabling navigation in chemical space using a factorization machine to extract structure-activity relationship rules. pubs.acs.org/doi/10.1021/ac… #Cheminformatics #GenerativeAI #DrugDiscovery #MachineLearning

Novelty: It hierarchically fuses 2D interaction subgraphs into a 3D global graph using cross-attention and dynamically predicts binding sites instead of relying on predefined coordinates.

Diffusion model for 3D molecular generation that combines local binding patterns with global protein pocket structures. link.springer.com/article/10.118… #Cheminformatics #DrugDiscovery #MachineLearning #AI

AgentD, an agentic framework designed to automate and streamline the early-stage computational drug discovery pipeline by combining the reasoning capabilities of LLMs with specialized external tools. pubs.acs.org/doi/10.1021/ac… #DrugDiscovery #Cheminformatics #GenerativeAI

WeMol, a cloud-based, zero-code drug discovery platform that unifies docking, ADMET prediction, REINVENT molecular generation, DiffSBDD structure-based design, and GROMACS MD simulations into a browser interface. #Cheminformatics #MachineLearning #AI pubs.acs.org/doi/10.1021/ac…

Novelty: The authors bypass standard reward-shaping by embedding selection algorithms (e.g., Determinantal Point Processes (DPP) and MaxMin) into the RL framework to effectively optimize generative models for de novo molecular design.

This paper mitigates mode collapse and enhances sampling efficiency in reinforcement learning-based molecular design. by sampling a diverse subset of generated molecules by the agent for policy updates. arxiv.org/abs/2506.21158 #GenerativeAI #Cheminformatics #MachineLearning #AI