A deep reinforcement learning hybrid algorithm for the computational discovery and characterization of small proteins utilizing mycobacterium tuberculosis as a model

• Main Author: Ouwabunmi, Babalola AbdulHafeez
• Format: Thesis
• Language: English
• Published: 2025
• Subjects: R Medicine; RA Public aspects of medicine
• Online Access: http://eprints.usm.my/62965/

The accurate prediction and characterization of small open reading frames (smORFs) are critical for understanding their functional roles in gene regulation and cellular processes. This study presents the development and evaluation of a novel hybrid machine learning algorithm that integrates the strengths of Random Forest and Gradient Boosting models to enhance the prediction of smORFs. The performance of the hybrid algorithm was rigorously assessed and compared to the standalone models using comprehensive evaluation metrics, including accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Results demonstrated that the hybrid model achieved superior performance, with an accuracy of 0.998, a sensitivity of 0.998, and a specificity of 1.00, significantly outperforming both the Random Forest and Gradient Boosting models individually. Additionally, transcriptomic data from Mycobacterium tuberculosis were utilized to validate the predictions, highlighting the biological relevance and potential applications of the proposed approach in computational biology. This study underscores the importance of combining machine learning techniques to improve prediction accuracy and provides a robust framework for advancing smORF discovery. While the focus was on comparing standalone and hybrid models, the study identifies opportunities for future benchmarking against external tools to further validate its contributions. The findings contribute to both computational and biological research, offering insights into smORF prediction methodologies and their applications.