ConforFormer: representation for molecules through understanding of conformers

2026/01/01 by Mas Klein, Irina Rudenko, Evgeny A. Pidko +1 · 1 voice
Computer Science · Materials Science · #Advanced Graph Neural Networks #Computational Drug Discovery Methods #Machine Learning in Materials Science

paper · pdf · doi:10.1039/d6dd00096g

openalex created_date 2025/12/16 · openalex publication_date 2026/01/01 · openalex updated_date 2026/07/15

Abstract

ConforFormer learns to align conformers while separating isomers, producing compact 3D molecular embeddings transferable across property prediction and similarity search tasks.

Discussions

• Are molecular graphs really necessary to represent molecules? Research from Klein, Rudenko, Bushmarinov and Pidko argues that powerful vector representations can be learned from 3D geometries alone, v [bsky, 1 points, 0 comments]

Related

• Swallowing the Bitter Pill: Simplified Scalable Conformer Generation
• Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks
• Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks
• Conformation Generation using Transformer Flows
• Molecular machine learning with conformer ensembles