AI-driven transcriptomic encoders: From explainable models to accurate, sample-independent cancer diagnostics

Danilo Croce, Artem Smirnov, Luigi Tiburzi, Serena Travaglini, Roberta Costa, Armando Calabrese, Roberto Basili, Nathan Levialdi Ghiron, Gerry Melino

Abstract: Highlights•Integrated Transformer-based learning with Explainable AI for cancer detection.•Identified key gene subsets for precise genomic analysis.•Maintained high accuracy across diverse patient cohorts.•Reduced model reliance on specific data encoding techniques.