Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Expanding Chemical Representation with k-mers and Fragment-based Fingerprints for Molecular Fingerprinting

Sarwan Ali, Prakash Chourasia, Murray Patterson

TL;DR

This work addresses the need for richer chemical representations from SMILES by expanding traditional fingerprints with substructure counting, $k$-mers, and Daylight-like features. The authors integrate these components into a Morgan fingerprint framework to produce enhanced molecular embeddings that better capture structural and contextual information. Empirical results on DrugBank-based drug-subcategory classification demonstrate that Morgan + $k$-mers and related hybrids often outperform MACCS, standalone $k$-mers, and Daylight fingerprints, with statistically significant gains. The proposed approach holds practical potential for molecular design, rapid similarity search, and improved predictive modeling in cheminformatics. Overall, the paper contributes a modular fingerprint fusion strategy that improves supervised analysis and could accelerate drug discovery workflows.

Abstract

This study introduces a novel approach, combining substruct counting, $k$-mers, and Daylight-like fingerprints, to expand the representation of chemical structures in SMILES strings. The integrated method generates comprehensive molecular embeddings that enhance discriminative power and information content. Experimental evaluations demonstrate its superiority over traditional Morgan fingerprinting, MACCS, and Daylight fingerprint alone, improving chemoinformatics tasks such as drug classification. The proposed method offers a more informative representation of chemical structures, advancing molecular similarity analysis and facilitating applications in molecular design and drug discovery. It presents a promising avenue for molecular structure analysis and design, with significant potential for practical implementation.

Expanding Chemical Representation with k-mers and Fragment-based Fingerprints for Molecular Fingerprinting

TL;DR

This work addresses the need for richer chemical representations from SMILES by expanding traditional fingerprints with substructure counting, -mers, and Daylight-like features. The authors integrate these components into a Morgan fingerprint framework to produce enhanced molecular embeddings that better capture structural and contextual information. Empirical results on DrugBank-based drug-subcategory classification demonstrate that Morgan + -mers and related hybrids often outperform MACCS, standalone -mers, and Daylight fingerprints, with statistically significant gains. The proposed approach holds practical potential for molecular design, rapid similarity search, and improved predictive modeling in cheminformatics. Overall, the paper contributes a modular fingerprint fusion strategy that improves supervised analysis and could accelerate drug discovery workflows.

Abstract

This study introduces a novel approach, combining substruct counting, -mers, and Daylight-like fingerprints, to expand the representation of chemical structures in SMILES strings. The integrated method generates comprehensive molecular embeddings that enhance discriminative power and information content. Experimental evaluations demonstrate its superiority over traditional Morgan fingerprinting, MACCS, and Daylight fingerprint alone, improving chemoinformatics tasks such as drug classification. The proposed method offers a more informative representation of chemical structures, advancing molecular similarity analysis and facilitating applications in molecular design and drug discovery. It presents a promising avenue for molecular structure analysis and design, with significant potential for practical implementation.
Paper Structure (15 sections, 3 figures, 3 tables, 4 algorithms)

This paper contains 15 sections, 3 figures, 3 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Approach
  4. Integration of $k$-mers in Morgan Fingerprint
  5. Daylight Fingerprint
  6. Experimental Setup
  7. Baseline Models
  8. MACCS Fingerprint
  9. $k$-mers
  10. Weighted $k$-mers
  11. Evaluation Metrics
  12. Data Visualization
  13. Results and Discussion
  14. Statistical Significance
  15. Conclusion

Figures (3)

  • Figure 1: Molecular structure for the drug named "Loperamide", with solubility AlogPS (Aqueous solubility and Octanol/Water partition coefficient) value of 0.00086, and the following SMILES string: CN(C)C(=O)C(CCN1CCC(O)(CC1)C1=CC=C (Cl)C=C1)(C1=CC=CC=C1)C1=CC=CC=C1
  • Figure 2: Different methods for Feature Vector generation using SMILE String
  • Figure 3: The t-SNE plots for different feature embedding methods. The DL stands for Daylight.