metadata

license: mit
task_categories:
  - text-generation
  - translation
tags:
  - chemistry
  - biology

ChEBI-20-MM Dataset

Overview

The ChEBI-20-MM is an extensive and multi-modal benchmark developed from the ChEBI-20 dataset. It is designed to provide a comprehensive benchmark for evaluating various models' capabilities in the field of molecular science. This benchmark integrates multi-modal data, including InChI, IUPAC, SELFIES, and images, making it a versatile tool for a wide range of molecular tasks.

Dataset Description

ChEBI-20-MM is an expansion of the original ChEBI-20 dataset, with a focus on incorporating diverse modalities of molecular data. This benchmark is tailored to assess models in several key areas:

Molecule Generation: Evaluating the ability of models to generate accurate molecular structures.
Image Recognition: Testing models on their proficiency in converting molecular images into other representational formats.
IUPAC Recognition: Evaluating the ability of models to generate IUPAC names from other representational formats.
Molecular Captioning: Assessing the capability of models to generate descriptive captions for molecular structures.
Retrieval Tasks: Measuring the effectiveness of models in retrieving molecular information accurately and efficiently.

Utility and Significance

By expanding the data modality variety, this benchmark enables a more comprehensive evaluation of models' performance in multi-modal data handling.

How to Use

Model reviews and evaluations related to this dataset can be directly accessed and used via the SLM4Mol link: SLM4Mol.

Data Visualization

We employ visualization techniques to analyze the suitability of data sources for language models and chemical space coverage. The figure below illustrates our use of different visualization methods to analyze text length distributions and token counts generated by each model's tokenizer across various text data types. This approach evaluates the adaptability of language models to the textual characteristics of our dataset.

We also focus on the top 10 scaffolds within the dataset, counting the number of molecules for each scaffold. Here, semi-transparent bars represent the total count, while solid bars indicate the quantity in the training set. On the other hand, for the analysis of \textbf{chemical space coverage}, we choose molecular weight (MW), LogP, the number of aromatic rings, and the Topological Polar Surface Area (TPSA) as descriptors. We examine the distribution and correlation of these descriptors within the dataset, providing insights into the chemical diversity and complexity present in our data.

Acknowledgments

The development of the ChEBI-20-MM dataset was inspired by the ChEBI-20 in molecule generation and captioning initiated by MolT5. Additional data information supplements are derived from PubChem.