Advancing HIV Patient Classification Through Contrastive Learning and Angular Distance in Graph Convolutional Networks Applied to Gene Expression Data

Accurate classification of HIV patients based on gene expression data is essential for personalized medicine, particularly in optimizing antiretroviral therapy (ART) strategies. This study proposes a novel framework that integrates Graph Convolutional Networks (GCNs) with contrastive learning and angular distance metrics to enhance classification accuracy in high-dimensional gene expression data. The model involves several stages: data preprocessing to normalize gene expression data, graph construction to represent gene interactions, GCN implementation to extract relational features, contrastive learning to improve feature separability, and the incorporation of angular distance metrics to enhance sensitivity to subtle variations. The combined loss function, integrating cross-entropy and contrastive losses, ensures robust optimization during training. When compared to traditional machine learning methods such as Random Forest, Naive Bayes, K-Nearest Neighbors, Support Vector Machines, and Artificial Neural Networks, the proposed model demonstrated superior 71.22% accuracy, 65.13% precision, and 64.7% recall. Despite these promising results, the study is limited by the relatively small dataset size, which may restrict the model's ability to generalize across broader patient populations and complex clinical scenarios. Future studies should incorporate larger and more diverse datasets to better capture the heterogeneity of HIV patients and further validate the proposed framework. While computationally intensive, the framework shows promise in advancing HIV patient profiling and guiding personalized treatment strategies, with future work focusing on improving efficiency and interpretability for broader clinical applications.