Optimised hybrid deep learning models for classifying high-dimensional RNA-sequencing data

Abstract

The field of genomics is moving closer to data-driven science. In the field of human genomics, the emergence of high-throughput sequence data-generating technology has resulted in an enormous amount of genomic data. Artificial intelligence, particularly Deep Learning techniques, has proved crucial in gaining knowledge and patterns from this genomic data. In the current study, we discuss the hybrid classification approach that combines Machine Learning and Deep Learning methods with Spider Monkey Optimisation (SMO) and Quantum-Inspired Immune Clone Optimisation (QIICO) techniques for improving the classification of RNA sequencing data. The extracted features are classified using Naive Bayes, Support Vector Machine (SVM), and Ada Boost Classifier on four diverse datasets to validate the models. The proposed hybrid model combines various meta-heuristic approaches to address the issues of overfitting and exploitation. Further, the thesis conducted a comparative analysis to assess the effectiveness of proposed hybrid classification models compared to existing models. The results obtained are compared with existing models using statistical tests and metric comparison analyses. The analysis may reveal that the hybrid models outperform existing models in accuracy, precision, recall, and F1-score or identify specific scenarios where they excel. Using Deep Learning approaches, we evaluated the genomics fields over and under charted areas. The genomic tools' underlying deep learning algorithms have been briefly described in the following sections of this research study. Finally, we briefly touched on the recent use of deep learning technologies in genomics. This study will certainly be beneficial for Biomedicine and human genome experts to understand what, why, when, and how this work is timely conducted and a hybrid visual analytic model with optimised features is developed to discover crucial genes for disease detection in patients, aiding in disease diagnosis and providing personalized treatment. Additionally, this study touched on the reasons why deep learning in genomics has just recently been applied, as well as potential remedies.

Date of Award	2024
Original language	English
Awarding Institution	Western Sydney University
Supervisor	Quang Vinh Nguyen (Supervisor)