LTVAE: Hybrid LSTM–Transformer VAE Architecture for Molecular SMILES Representation

Abstract

A hybrid LSTM–Transformer Variational Autoencoder (LTVAE) framework is introduced in this paper for learning compact and chemically meaningful latent representations from SMILES strings, providing a SMILES‐to‐latent‐to‐SMILES representation model trained on ZINC15 and evaluated on ZINC15 and an out‐of‐distribution dye dataset. In the proposed model, sequential and structural dependencies in molecular strings are encoded using a bidirectional LSTM, while a Transformer‐based decoder is employed to reconstruct valid molecular structures in the evaluated settings. When trained on a curated ZINC15 dataset, the framework achieves reasonable reconstruction performance, with 98.58% token‐level accuracy, 92.19% string‐level accuracy, a 99.30% chemical validity ratio, and an average Levenshtein distance of 0.29 using a 64‐dimensional latent space. Reconstruction under an out‐of‐distribution evaluation setting is evaluated on a structurally distinct dye dataset, where 95.56% token‐level accuracy and 85.30% validity are achieved. The learned latent space preserves trends aligned with both general molecular descriptors and dye‐relevant structural descriptors, particularly conjugation length and aromatic surface fraction, with Fsp providing complementary information on molecular saturation. Controlled perturbations in the latent space suggest smooth and chemically consistent structural variations, illustrating the continuity and interpretability of the learned latent manifold.

Transformer-based molecular fragment prediction using SMILES and DeepSMILES representations in a fragment-based drug discovery pipeline.
Aayush Kothari, Amish Gupta, Nisarg Shah +5
Journal of cheminformatics31 Jul 2026Peer-reviewedCC BYFull textTheoretical & computational chemistryEurope PMCPMC
Explicitness in SMILES representation via ExACT: improved tokenization for aqueous solubility prediction.
Domagoj Begušić, Damir Pintar, Zlatko Smole +1
Journal of cheminformatics4 Aug 2026Peer-reviewedCC BYFull textTheoretical & computational chemistryEurope PMCPMC
Monte Carlo-based QSPR modeling of drop-weight impact sensitivity of energetic materials using hybrid SMILES–graph descriptors
Shahram Lotfi, Shahin Ahmadi, Parvin Kumar
RSC advances14 Oct 2026Peer-reviewedCC BYFull textGeneral chemistryEurope PMCPMC
Geometry-Aware Line Graph Transformer Pretraining for Molecular Property Prediction
Peizhen Bai, Xianyuan Liu, Wenrui Fan +3
IEEE Transactions on Neural Networks and Learning Systems11 Jun 2026Peer-reviewedLicence not statedFull text8 citedMedicinal chemistry & drug discoveryOpenAlex
Data-driven and feature-sifting based state-of-health estimation for li-ion batteries using KPCA and NRBO-optimized transformer-LSTM hybrid model
Xiangcai Luo, Jinwei Chen, Haijiang Zhang +6
PLOS One5 Oct 2026Peer-reviewedCC BY 4.0Full textElectrochemistry & energy StatisticsPLOS

Cite

BibTeX RIS

BibTeX

@article{mondol2026ltvae,
  title = {{LTVAE: Hybrid LSTM–Transformer VAE Architecture for Molecular SMILES Representation}},
  author = {Md Halim Mondol and Qingshan Wei and Edgar Lobaton},
  journal = {Journal of Computational Chemistry},
  year = {2026},
  month = nov,
  doi = {10.1002/jcc.70521},
  url = {https://doi.org/10.1002/jcc.70521},
  note = {CC BY 4.0},
}

RIS

TY  - JOUR
TI  - LTVAE: Hybrid LSTM–Transformer VAE Architecture for Molecular SMILES Representation
AU  - Md Halim Mondol
AU  - Qingshan Wei
AU  - Edgar Lobaton
PY  - 2026
DA  - 2026/11/03
JO  - Journal of Computational Chemistry
DO  - 10.1002/jcc.70521
UR  - https://doi.org/10.1002/jcc.70521
AB  - A hybrid LSTM–Transformer Variational Autoencoder (LTVAE) framework is introduced in this paper for learning compact and chemically meaningful latent representations from SMILES strings, providing a SMILES‐to‐latent‐to‐SMILES representation model trained on ZINC15 and evaluated on ZINC15 and an out‐of‐distribution dye dataset. In the proposed model, sequential and structural dependencies in molecular strings are encoded using a bidirectional LSTM, while a Transformer‐based decoder is employed to reconstruct valid molecular structures in the evaluated settings. When trained on a curated ZINC15 dataset, the framework achieves reasonable reconstruction performance, with 98.58% token‐level accuracy, 92.19% string‐level accuracy, a 99.30% chemical validity ratio, and an average Levenshtein distance of 0.29 using a 64‐dimensional latent space. Reconstruction under an out‐of‐distribution evaluation setting is evaluated on a structurally distinct dye dataset, where 95.56% token‐level accuracy and 85.30% validity are achieved. The learned latent space preserves trends aligned with both general molecular descriptors and dye‐relevant structural descriptors, particularly conjugation length and aromatic surface fraction, with Fsp providing complementary information on molecular saturation. Controlled perturbations in the latent space suggest smooth and chemically consistent structural variations, illustrating the continuity and interpretability of the learned latent manifold.
SN  - 0192-8651
SN  - 1096-987X
ER  -

Abstract

Related

Cite