TW

@tw80086

PhD, EPFL
Generative/agentic AI for chemistry

Renens (VD), Suisse preghosh.github.io/personal_web/ Joined May 2023

39 Following 6 Followers 48 Posts

Latest posts 5

AutomaticMolCraft: A Web Interface for 3D Molecular Generation, Dataset Curation, and Structure–Property Analysis” is now available as a preprint.

Source code:
github.com/lcmd-epfl/Auto…
Preprint:
chemrxiv.org/doi/full/10.26…

github.com GitHub - lcmd-epfl/AutomaticMolCraft: A locally deployed browser platform for running and organis... A locally deployed browser platform for running and organising 3D molecular discovery campaign - lcmd-epfl/AutomaticMolCraft

TW @tw80086 Jul 2

Embedding molecular quality information in all stages of the 3D generative workflow, from data to training and generation.
Source code: github.com/lcmd-epfl/MolC…
Preprint: chemrxiv.org/doi/full/10.26…

github.com GitHub - lcmd-epfl/MolCraftDiffusion: A 3D Molecular Generation Framework for Data-driven Molecular... A 3D Molecular Generation Framework for Data-driven Molecular Applications. - lcmd-epfl/MolCraftDiffusion

TW @tw80086 Apr 16

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…

www.nature.com Disentangled autoencoding equivariant diffusion model for controlled generation of 3D molecules Nature Communications - Controlled generation of 3D molecules with desired properties can speed up drug discovery. Here, the authors propose a semantics-guided diffusion model to achieve precise...

TW @tw80086 Apr 10

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

chemrxiv.org Right into the Saddle: Stereochemistry-Aware Generation of Molecular Transition States | ChemRxiv Determination of transition state structures is fundamental to the mechanistic understanding of chemical reactivity and selectivity, yet conventional approaches require extensive expert intervention...

TW @tw80086 Apr 7

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…

pubs.rsc.org Coupled fragment-based generative modeling with stochastic interpolants Fragment-based drug design (FBDD) has become a key approach in structure-based drug discovery, allowing researchers to systematically develop molecular fra