AI-Driven Deep Learning Approach for Pan-Cancer Immune Profiling

Minh Huu Nhat Le, Ha-Hieu Pham, Huy Quoc Nguyen, Hong Xuan Ong, Hien Quang Kha, Phat Ky Nguyen, Thanh-Huy Nguyen, Han Hong Huynh, Dang Nguyen, Thanh-Minh Nguyen, An Thuy Vo, Thuy Vu Minh Nguyen, Lam Huu Phuc Nguyen, Trung Minh Tu Tran, Nguyen Quoc Khanh Le

Abstract: The tumor immune microenvironment (TME) influences cancer progression and treatment. RNA-Seq has identified six immune subtypes: Wound Healing (WH, C1), IFN- Dominant (IFNG, C2), Inflammatory (INF, C3), Lymphocyte Depleted (LD, C4), Immunologically Quiet (IQ, C5), and TGF-β Dominant (TGFb, C6). This study uses a Convolutional Neural Network (CNN) to classify these subtypes from RNA-Seq data. The model, with ReLU activation and dropout, achieved a 10-fold F1-score of 0.9483 and AUC of 0.9969. Results show CNN’s effectiveness in handling class imbalance and modeling complex gene interactions, outperforming XGBoost, Random Forest, and TabNet. Future work will validate results on independent datasets and incorporate multi-omics data to improve accuracy.

External IDs:doi:10.3233/shti250903