TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Text2Mol	55.4	# 11
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	BLEU	85.4	# 4
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Exact Match	30.2	# 5
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Levenshtein	16.07	# 14
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	MACCS FTS	83.4	# 14
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	RDK FTS	74.6	# 9
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Morgan FTS	68.4	# 11
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Frechet ChemNet Distance (FCD)	1.20	# 13
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Validity	90.5	# 8
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Parameter Count	770000000	# 17
Molecule Captioning	ChEBI-20	MolT5-Small	BLEU-2	51.9	# 19
Molecule Captioning	ChEBI-20	MolT5-Small	BLEU-4	43.6	# 19
Molecule Captioning	ChEBI-20	MolT5-Small	ROUGE-1	62.0	# 15
Molecule Captioning	ChEBI-20	MolT5-Small	ROUGE-2	46.9	# 15
Molecule Captioning	ChEBI-20	MolT5-Small	ROUGE-L	56.3	# 14
Molecule Captioning	ChEBI-20	MolT5-Small	METEOR	55.1	# 19
Molecule Captioning	ChEBI-20	MolT5-Small	Text2Mol	54.0	# 12
Molecule Captioning	ChEBI-20	MolT5-Base	BLEU-2	54.0	# 17
Molecule Captioning	ChEBI-20	MolT5-Base	BLEU-4	45.7	# 16
Molecule Captioning	ChEBI-20	MolT5-Base	ROUGE-1	63.4	# 11
Molecule Captioning	ChEBI-20	MolT5-Base	ROUGE-2	48.5	# 12
Molecule Captioning	ChEBI-20	MolT5-Base	ROUGE-L	57.8	# 12
Molecule Captioning	ChEBI-20	MolT5-Base	METEOR	56.9	# 16
Molecule Captioning	ChEBI-20	MolT5-Base	Text2Mol	54.7	# 11
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Text2Mol	59.0	# 2
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	BLEU	81.0	# 10
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Exact Match	31.4	# 4
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Levenshtein	16.758	# 13
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	MACCS FTS	87.2	# 7
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	RDK FTS	78.6	# 5
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Morgan FTS	72.2	# 6
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Frechet ChemNet Distance (FCD)	0.44	# 8
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Validity	99.6	# 3
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Parameter Count	770000000	# 17
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Text2Mol	48.2	# 13
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	BLEU	75.5	# 15
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Exact Match	7.9	# 17
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Levenshtein	25.988	# 3
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	MACCS FTS	70.3	# 18
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	RDK FTS	56.8	# 18
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Morgan FTS	51.7	# 18
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Frechet ChemNet Distance (FCD)	2.49	# 15
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Validity	72.1	# 18
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Parameter Count	60000000	# 5
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Text2Mol	49.6	# 12
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	BLEU	76.9	# 13
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Exact Match	8.1	# 16
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Levenshtein	24.458	# 5
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	MACCS FTS	72.1	# 17
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	RDK FTS	58.8	# 16
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Morgan FTS	52.9	# 16
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Frechet ChemNet Distance (FCD)	2.18	# 14
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Validity	77.2	# 17
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Parameter Count	220000000	# 10
Molecule Captioning	ChEBI-20	MolT5-Large	BLEU-2	59.4	# 8
Molecule Captioning	ChEBI-20	MolT5-Large	BLEU-4	50.8	# 8
Molecule Captioning	ChEBI-20	MolT5-Large	ROUGE-1	65.4	# 7
Molecule Captioning	ChEBI-20	MolT5-Large	ROUGE-2	51.0	# 8
Molecule Captioning	ChEBI-20	MolT5-Large	ROUGE-L	59.4	# 7
Molecule Captioning	ChEBI-20	MolT5-Large	METEOR	61.4	# 8
Molecule Captioning	ChEBI-20	MolT5-Large	Text2Mol	58.2	# 4

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/translation-between-molecules-and-natural/text-based-de-novo-molecule-generation-on)](https://paperswithcode.com/sota/text-based-de-novo-molecule-generation-on?p=translation-between-molecules-and-natural)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/translation-between-molecules-and-natural/molecule-captioning-on-chebi-20)](https://paperswithcode.com/sota/molecule-captioning-on-chebi-20?p=translation-between-molecules-and-natural)`

Translation between Molecules and Natural Language

25 Apr 2022 · Carl Edwards, Tuan Lai, Kevin Ros, Garrett Honke, Kyunghyun Cho, Heng Ji ·

We present $\textbf{MolT5}$ $-$ a self-supervised learning framework for pretraining models on a vast amount of unlabeled natural language text and molecule strings. $\textbf{MolT5}$ allows for new, useful, and challenging analogs of traditional vision-language tasks, such as molecule captioning and text-based de novo molecule generation (altogether: translation between molecules and language), which we explore for the first time. Since $\textbf{MolT5}$ pretrains models on single-modal data, it helps overcome the chemistry domain shortcoming of data scarcity. Furthermore, we consider several metrics, including a new cross-modal embedding-based metric, to evaluate the tasks of molecule captioning and text-based molecule generation. Our results show that $\textbf{MolT5}$-based models are able to generate outputs, both molecules and captions, which in many cases are high quality.

PDF Abstract

Code

Add Remove Mark official

blender-nlp/MolT5 official

133

Tasks

Add Remove

Drug Discovery

Molecule Captioning

Self-Supervised Learning

Text-based de novo Molecule Generation

Translation

Datasets

ChEBI-20

Results from the Paper

Edit

Ranked #4 on Text-based de novo Molecule Generation on ChEBI-20

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large	Text2Mol	55.4	# 11	Compare
			BLEU	85.4	# 4	Compare
			Exact Match	30.2	# 5	Compare
			Levenshtein	16.07	# 14	Compare
			MACCS FTS	83.4	# 14	Compare
			RDK FTS	74.6	# 9	Compare
			Morgan FTS	68.4	# 11	Compare
			Frechet ChemNet Distance (FCD)	1.20	# 13	Compare
			Validity	90.5	# 8	Compare
			Parameter Count	770000000	# 17	Compare
Molecule Captioning	ChEBI-20	MolT5-Small	BLEU-2	51.9	# 19	Compare
			BLEU-4	43.6	# 19	Compare
			ROUGE-1	62.0	# 15	Compare
			ROUGE-2	46.9	# 15	Compare
			ROUGE-L	56.3	# 14	Compare
			METEOR	55.1	# 19	Compare
			Text2Mol	54.0	# 12	Compare
Molecule Captioning	ChEBI-20	MolT5-Base	BLEU-2	54.0	# 17	Compare
			BLEU-4	45.7	# 16	Compare
			ROUGE-1	63.4	# 11	Compare
			ROUGE-2	48.5	# 12	Compare
			ROUGE-L	57.8	# 12	Compare
			METEOR	56.9	# 16	Compare
			Text2Mol	54.7	# 11	Compare
Text-based de novo Molecule Generation	ChEBI-20	MolT5-Large-HV	Text2Mol	59.0	# 2	Compare
			BLEU	81.0	# 10	Compare
			Exact Match	31.4	# 4	Compare
			Levenshtein	16.758	# 13	Compare
			MACCS FTS	87.2	# 7	Compare
			RDK FTS	78.6	# 5	Compare
			Morgan FTS	72.2	# 6	Compare
			Frechet ChemNet Distance (FCD)	0.44	# 8	Compare
			Validity	99.6	# 3	Compare
			Parameter Count	770000000	# 17	Compare
Text-based de novo Molecule Generation	ChEBI-20	MolT5-small	Text2Mol	48.2	# 13	Compare
			BLEU	75.5	# 15	Compare
			Exact Match	7.9	# 17	Compare
			Levenshtein	25.988	# 3	Compare
			MACCS FTS	70.3	# 18	Compare
			RDK FTS	56.8	# 18	Compare
			Morgan FTS	51.7	# 18	Compare
			Frechet ChemNet Distance (FCD)	2.49	# 15	Compare
			Validity	72.1	# 18	Compare
			Parameter Count	60000000	# 5	Compare
Text-based de novo Molecule Generation	ChEBI-20	MolT5-base	Text2Mol	49.6	# 12	Compare
			BLEU	76.9	# 13	Compare
			Exact Match	8.1	# 16	Compare
			Levenshtein	24.458	# 5	Compare
			MACCS FTS	72.1	# 17	Compare
			RDK FTS	58.8	# 16	Compare
			Morgan FTS	52.9	# 16	Compare
			Frechet ChemNet Distance (FCD)	2.18	# 14	Compare
			Validity	77.2	# 17	Compare
			Parameter Count	220000000	# 10	Compare
Molecule Captioning	ChEBI-20	MolT5-Large	BLEU-2	59.4	# 8	Compare
			BLEU-4	50.8	# 8	Compare
			ROUGE-1	65.4	# 7	Compare
			ROUGE-2	51.0	# 8	Compare
			ROUGE-L	59.4	# 7	Compare
			METEOR	61.4	# 8	Compare
			Text2Mol	58.2	# 4	Compare

Methods

Add Remove

No methods listed for this paper. Add relevant methods here

Edit Social Preview

Translation between Molecules and Natural Language

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove