What question did this study set out to answer?

This research aims to determine how different molecular string representations impact the effectiveness of transformer models in fragment recovery for drug discovery.

August 2, 2026Journal of CheminformaticsOpen Access

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

View Full Paper

Ask AI

Bookmark

Authors

AKAayush Kothari AGAmish Gupta NSNisarg Shah

Discussion

Member takes

Overview

Controlled study evaluates molecular string representations for fragment recovery in drug discovery, suggesting better predictions with DeepSMILES.

Key Points

This research aims to determine how different molecular string representations impact the effectiveness of transformer models in fragment recovery for drug discovery.
Controlled study comparing DeepBERTa trained on DeepSMILES against a baseline trained on SMILES;
Approximate analysis using 34,000 drug-fragment pairs with evaluations based on Tanimoto similarity;
Assessment of per-sample selection to enhance fragment recovery accuracy.
DeepSMILES predictions showed 54.2% syntactically valid predictions versus 43.5% for SMILES;
Mean Tanimoto similarity was higher for DeepSMILES (0.36) compared to SMILES (0.29);
DeepSMILES outperformed SMILES in winning instances (27.8% vs. 16.3% of test molecules).

Cite This Study

Kothari et al. (2026) studied this question.

synapsesocial.com/papers/6a6eeacb1b0468a7eeab34fc https://doi.org/10.1186/s13321-026-01255-w

View Full Paper

Ask AI

Bookmark

August 2, 2026Journal of CheminformaticsOpen Access

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

View Full Paper

Ask AI

Bookmark

Authors

AKAayush Kothari AGAmish Gupta NSNisarg Shah

Discussion

Member takes

Overview

Controlled study evaluates molecular string representations for fragment recovery in drug discovery, suggesting better predictions with DeepSMILES.

Key Points

This research aims to determine how different molecular string representations impact the effectiveness of transformer models in fragment recovery for drug discovery.
Controlled study comparing DeepBERTa trained on DeepSMILES against a baseline trained on SMILES;
Approximate analysis using 34,000 drug-fragment pairs with evaluations based on Tanimoto similarity;
Assessment of per-sample selection to enhance fragment recovery accuracy.
DeepSMILES predictions showed 54.2% syntactically valid predictions versus 43.5% for SMILES;
Mean Tanimoto similarity was higher for DeepSMILES (0.36) compared to SMILES (0.29);
DeepSMILES outperformed SMILES in winning instances (27.8% vs. 16.3% of test molecules).

Cite This Study

Kothari et al. (2026) studied this question.

synapsesocial.com/papers/6a6eeacb1b0468a7eeab34fc https://doi.org/10.1186/s13321-026-01255-w

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark