This research introduces GexBERT, a breakthrough transformer-based framework that analyzes gene expression data to improve cancer prognosis and classification.

• Addresses key challenges in gene expression analysis: data sparsity, high dimensionality, and missing values
• Learns context-aware gene embeddings through pretraining on large-scale transcriptomic data
• Demonstrates superior performance in pan-cancer classification and cancer-specific survival prediction
• Bridges the gap between AI advances in natural language processing and genomic medicine

Significance: This technology could revolutionize personalized cancer treatment by enabling more accurate prognosis and classification, ultimately improving patient outcomes through targeted therapeutic approaches.

Transformer-Based Representation Learning for Robust Gene Expression Modeling and Cancer Prognosis