Article Type
Article
Abstract
The antibacterial properties of nanoparticles play a crucial role in industries such as textiles, water disinfection, pharmaceuticals, and food packaging. Given the rising concerns over antibiotic resistance, metal oxide nanoparticles (NPs) have emerged as a promising alternative. While machine learning techniques in nano informatics have shown encouraging results in predicting the antibacterial efficacy of nanoparticles, the application of deep learning methods in this domain remains limited. This study introduces a novel predictive model based on a multi-branch recurrent neural network, incorporating three types of recurrent cells: LSTM, BiLSTM, and GRU, to estimate the antibacterial effectiveness of nanoparticles. The proposed model employs a three-stage preprocessing pipeline to enhance data quality and is trained on a dataset comprising 436 samples of metal-based nanoparticles. Furthermore, SHAP is utilized to explain the contribution of input features in the model's predictions. With an RMSE of 3.01, experimental data show that the suggested model performs better than deep neural networks like transformers and multilayer perceptron as well as conventional machine learning techniques. For use in lab settings, this model can function as an AI-powered assistance tool.
Keywords
Nanoparticles, Antibacterial efficacy, Multi-branch recurrent network, SHAP explainability, Nano informatics
Recommended Citation
Zainulabdeen, Shaimaa M.S.; Haifaa, Najee B.; and Atiyah, Iman Ali
(2025)
"An Advanced Multi-Branch Recurrent Deep Learning Model for Predicting the Antibacterial Capacity of Nanoparticles with SHAP-Based Explainability,"
Muthanna Medical Journal: Vol. 12:
Iss.
3, Article 2.
Available at:
https://muthmj.mu.edu.iq/journal/vol12/iss3/2
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
