Papers › MolXPT: Wrapping Molecules with Text for Generative Pre-training

MolXPT: Wrapping Molecules with Text for Generative Pre-training

18 May 2023arXiv:2305.10688archive 2025-07-28

Zequn Liu, Wei zhang, Yingce Xia, Lijun Wu, Shufang Xie, Tao Qin, Ming Zhang, Tie-Yan Liu

Generative pre-trained Transformer (GPT) has demonstrates its great success in natural language processing and related techniques have been adapted into molecular modeling. Considering that text is the most important record for scientific discovery, in this paper, we propose MolXPT, a unified language model of text and molecules pre-trained on SMILES (a sequence representation of molecules) wrapped by text. Briefly, we detect the molecule names in each sequence and replace them to the corresponding SMILES. In this way, the SMILES could leverage the information from surrounding text, and vice versa. The above wrapped sequences, text sequences from PubMed and SMILES sequences from PubChem are all fed into a language model for pre-training. Experimental results demonstrate that MolXPT outperforms strong baselines of molecular property prediction on MoleculeNet, performs comparably to the best model in text-molecule translation while using less than half of its parameters, and enables zero-shot molecular generation without finetuning.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Language ModelingLanguage ModellingMolecular Property PredictionMolecule CaptioningProperty PredictionText-based de novo Molecule Generationscientific discovery

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Molecular Property Prediction BACE MolXPT ROC-AUC 88.4 #1 of 20 Archive leaderboard report
Molecular Property Prediction BBBP MolXPT ROC-AUC 80.5 ± 0.5 #11 of 29 Archive leaderboard report
Molecular Property Prediction HIV dataset MolXPT AUC 0.781 #6 of 11 Archive leaderboard report
Molecular Property Prediction SIDER MolXPT ROC-AUC 71.7 #3 of 19 Archive leaderboard report
Molecular Property Prediction Tox21 MolXPT ROC-AUC 77.1 #10 of 20 Archive leaderboard report
Molecular Property Prediction clintox MolXPT ROC-AUC 95.3±0.2 #3 of 20 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT BLEU-2 59.4 #16 of 33 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT BLEU-4 50.5 #16 of 33 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT METEOR 62.6 #16 of 33 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT ROUGE-1 66 #16 of 33 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT ROUGE-2 51.1 #16 of 33 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT ROUGE-L 59.7 #16 of 33 Archive leaderboard report
Molecule Captioning ChEBI-20 MolXPT Text2Mol 59.4 #16 of 33 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT Exact Match 21.5 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT Frechet ChemNet Distance (FCD) 0.45 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT MACCS FTS 85.9 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT Morgan FTS 66.7 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT Parameter Count 350000000 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT RDK FTS 75.7 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT Text2Mol 57.8 #20 of 20 Archive leaderboard report
Text-based de novo Molecule Generation ChEBI-20 MolXPT Validity 98.3 #20 of 20 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Absolute Position EncodingsAdamAttentionBPEDense ConnectionsDropoutLabel SmoothingLayer NormalizationLinear LayerMulti-Head AttentionPosition-Wise Feed-Forward LayerResidual ConnectionSoftmaxTransformer

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections