Comparative analysis of computational approaches for predicting human neuronal Transthyretin (TTR) transcription activators and human dopamine D1 receptor antagonists

Lists

Ivanova, Mariya, Russo, Nicola, Mihaylov, Gueorgui and Konstantin, Nikolic ORCID logoORCID: https://orcid.org/0000-0002-6551-2977 (2025) Comparative analysis of computational approaches for predicting human neuronal Transthyretin (TTR) transcription activators and human dopamine D1 receptor antagonists. Journal of Cellular Biochemistry. ISSN 0730-2312 (Submitted)

[thumbnail of PDF/A] PDF (PDF/A)
Comparative analysis of computational approaches for predicting human neuronal Transthyretin_IvanovaM_accesssible.pdf - Submitted Version
Restricted to Repository staff only
Download (504kB)
[thumbnail of PDF/A] PDF (PDF/A)
Supplementary_materials_Comparative analysis of computational approaches for predicting human neuronal Transthyretin_IvanovaM_accessible.pdf - Supplemental Material
Restricted to Repository staff only
Download (1MB)

Abstract

This study is part of a larger, ongoing research effort to develop a global machine learning (ML) model for drug discovery and development. The methodology utilised the scikit-learn ML library and combined NMR spectroscopy data (derived from SMILES notations) with molecular features from PubChem to predict the functionality of small biomolecules. The approach's effectiveness was demonstrated using a human dopamine D1 receptor antagonist case study and compared to a case study predicting neuronal Transthyretin (TTR) transcription activators. Additionally, a CID_SID ML model was developed to predict TTR transcription activation capabilities of compounds, based solely on their PubChem CID and SID. The enhanced ML model predicting dopamine D1 receptor antagonists obtained 75.8% Accuracy, 84.2% Precision, 63.6% Recall, 72.5% F1-score and 75.8 % ROC, trained on 25,532 samples and tested on 5,466. The hypothetical ML model predicting neuronal TTR transcription activators achieved 67.4% Accuracy, 74.0% Precision, 53.5% Recall, 62.1% F1-score and 67.4 % ROC, if it could be trained with 25,532 samples and tested with 5,466. CID_SID ML model achieved 81.5% Accuracy, 94.6% Precision, 66.8% Recall, 78.3% F1-score and 81.5 % ROC. Overall, an improved machine learning (ML) approach was developed by incorporating additional molecular features. The comparison to another case study revealed that the effectiveness of the ML approach is dependent on the specific case study. Additionally, the CID_SID ML model yielded promising results, demonstrating its potential for predicting side effects related to TTR transcription activation.

Item Type: Article
Keywords: CID_SID ML model, neurodegenerative disorders, 13C NMR spectroscopy, drug discovery and development, machine learning
Subjects: Computing > Intelligent systems
Depositing User: Mariya Ivanova
Date Deposited: 16 Sep 2025 15:35
Last Modified: 16 Sep 2025 15:45
URI: https://repository.uwl.ac.uk/id/eprint/14070

Actions (login required)

View Item View Item
References
Tools
CORE (COnnecting REpositories)

Menu