Skip to content
DraconiticResearch
Search Sources API
203,948 studies31 sourcesUpdated 31 min ago
Peer-reviewedCC BYJournal of cheminformatics

Transformer-based molecular fragment prediction using SMILES and DeepSMILES representations in a fragment-based drug discovery pipeline.

  • Aayush Kothari1
  • Amish Gupta1
  • Nisarg Shah2
  • Thomas Reed2
  • Anvita Nattuva2
  • Rahima Nazirudeen2
  • Harman Brah2
  • Marx Akl3
  1. 1Department of Computer Science & Engineering, Aspiring Scholars Directed Research Program, Fremont, CA, 94539, USA.
  2. 2Department of Chemistry, Biochemistry, and Physical Science, Aspiring Scholars Directed Research Program, Fremont, CA, 94539, USA.
  3. 3Department of Computer Science & Engineering, Aspiring Scholars Directed Research Program, Fremont, CA, 94539, USA. [email protected].
Read full textPDFDOICite

Abstract

Fragment-based drug discovery (FBDD) uses small molecular fragments as starting points for drug development, and machine-learning models that operate on molecular string representations are increasingly applied to fragment-related tasks. SMILES is the dominant such representation, but its paired ring closures and balanced parentheses introduce syntactic complexity that can affect model behavior. We present a controlled study of how molecular string representation influences transformer-based fragment recovery, using an FBDD-motivated label pipeline: reference fragments are derived from known drugs via RECAP fragmentation and docking-based ranking, and the model is scored on recovering them. We introduce DeepBERTa, a ChemBERTa-derived transformer pretrained on DeepSMILES, and evaluate it against a matched SMILES baseline under identical architecture, data splits, and optimization (approximately 34,000 drug-fragment pairs). DeepSMILES produces syntactically valid predictions more often than SMILES (54.2% vs. 43.5%) and a higher full test-set mean Tanimoto similarity (0.36 vs. 0.29). A per-sample selection between the two representations raises mean Tanimoto to 0.43, and a deployable variant that selects on model confidence rather than the reference recovers most of this gain. Among molecules where one representation strictly wins, DeepSMILES wins more often than SMILES (27.8% vs. 16.3% of test molecules). These results show that string representation meaningfully influences transformer fragment recovery and that representation-aware selection outperforms either representation alone.

Related

  1. Geometry-Aware Line Graph Transformer Pretraining for Molecular Property Prediction
    Peizhen Bai, Xianyuan Liu, Wenrui Fan +3
    IEEE Transactions on Neural Networks and Learning SystemsPeer-reviewedLicence not statedFull text8 citedMedicinal chemistry & drug discoveryOpenAlex
  2. Hybrid-CatBoost: an empirical comparison of molecular representations for aqueous-solubility prediction on the TDC benchmark
    Ruming Li
    ChemRxivPreprintCC BYFull textGeneral chemistryChemRxiv
  3. Task-adaptive multimodal molecular representations for structure-sensitive property prediction
    Shaolong Lin, Silong Zhai, Shihang Wang +11
    Chemical SciencePeer-reviewedCC BY 3.0Full textOrganic chemistryPMC
  4. DeMiR: Density-Matrix-Informed Representations for Molecular Property Prediction
    Donghyeon Kim, Dongsup Kim
    ChemRxivPreprintCC BYFull textGeneral chemistryChemRxiv
  5. Scaffold-Controlled Evaluation of Molecular Representations for Antimalarial Activity Prediction
    Jean-Pierre TCHAPET NJAFA, Penabei SAMAFOU, Fon Wilfred MBACHAM +2
    ChemRxivPreprintCC BYFull textGeneral chemistryChemRxiv

Cite

BibTeXRIS
BibTeX
@article{kothari2026transformerbased,
  title = {{Transformer-based molecular fragment prediction using SMILES and DeepSMILES representations in a fragment-based drug discovery pipeline.}},
  author = {Aayush Kothari and Amish Gupta and Nisarg Shah and Thomas Reed and Anvita Nattuva and Rahima Nazirudeen and Harman Brah and Marx Akl},
  journal = {Journal of cheminformatics},
  year = {2026},
  month = jul,
  doi = {10.1186/s13321-026-01255-w},
  url = {https://europepmc.org/article/MED/42958454},
  note = {CC BY},
}
RIS
TY  - JOUR
TI  - Transformer-based molecular fragment prediction using SMILES and DeepSMILES representations in a fragment-based drug discovery pipeline.
AU  - Aayush Kothari
AU  - Amish Gupta
AU  - Nisarg Shah
AU  - Thomas Reed
AU  - Anvita Nattuva
AU  - Rahima Nazirudeen
AU  - Harman Brah
AU  - Marx Akl
PY  - 2026
DA  - 2026/07/31
JO  - Journal of cheminformatics
DO  - 10.1186/s13321-026-01255-w
UR  - https://europepmc.org/article/MED/42958454
AB  - Fragment-based drug discovery (FBDD) uses small molecular fragments as starting points for drug development, and machine-learning models that operate on molecular string representations are increasingly applied to fragment-related tasks. SMILES is the dominant such representation, but its paired ring closures and balanced parentheses introduce syntactic complexity that can affect model behavior. We present a controlled study of how molecular string representation influences transformer-based fragment recovery, using an FBDD-motivated label pipeline: reference fragments are derived from known drugs via RECAP fragmentation and docking-based ranking, and the model is scored on recovering them. We introduce DeepBERTa, a ChemBERTa-derived transformer pretrained on DeepSMILES, and evaluate it against a matched SMILES baseline under identical architecture, data splits, and optimization (approximately 34,000 drug-fragment pairs). DeepSMILES produces syntactically valid predictions more often than SMILES (54.2% vs. 43.5%) and a higher full test-set mean Tanimoto similarity (0.36 vs. 0.29). A per-sample selection between the two representations raises mean Tanimoto to 0.43, and a deployable variant that selects on model confidence rather than the reference recovers most of this gain. Among molecules where one representation strictly wins, DeepSMILES wins more often than SMILES (27.8% vs. 16.3% of test molecules). These results show that string representation meaningfully influences transformer fragment recovery and that representation-aware selection outperforms either representation alone.
SN  - 1758-2946
ER  -