CatEmb Resource Hub: Catalyst Embedding Datasets and Model Weights

LiCheng Xu's CatEmb project provides resources for generating stereoelectronic-aware molecular descriptors from 2D graphs. The repository includes a 2.4 GB dataset of ligands, catalysts, and metal-ligand complexes with 2D and processed 3D information, alongside pre-trained model weights for descriptor generation. It was last updated on 2026-05-29.

Use Cases

• Generate stereoelectronic-aware molecular descriptors for catalyst similarity analysis based on the pre-trained CatEmb model weights.
• Reproduce or benchmark catalyst embedding models using the provided CatCompDB dataset of ligands and metal-ligand complexes.
• Accelerate catalyst design workflows by using the pre-trained descriptors for downstream machine learning tasks.

Strengths

• Includes a dataset with associated 2D structures and processed 3D information for ligands, catalysts, and complexes.
• Provides pre-trained model weights for multiple descriptor dimensions (8 to 4096), enabling immediate inference.
• Released under a permissive CC-BY-4.0 license, facilitating reuse and modification.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the dataset is unknown, which may limit suitability assessment.
• Data may reflect the specific chemical scope and biases inherent to the CatCompDB collection.

Provenance

Source

LiCheng Xu via figshare

Collection Method

Likely compiled and processed for the CatEmb (Catalyst Embedding) research project.

Freshness

Last updated 2026-05-29 07:28:54; freshness should be verified.