Advancing cancer care: unravelling genomic insights for precision medicine using meticulous predictive architecture

Advancements in genomic profiling have significantly enhanced oncology by enabling precise tumor classification. However, challenges such as high dimensionality and limited sample sizes persist. This study presents a predictive modeling framework integrating t-distributed stochastic neighbor embedding (t-SNE) with Kullback-Leibler divergence and Shannon entropy reduction for efficient dimensionality reduction. A hybrid decisive random forest classifier further enhances model robustness and generalizability. Evaluated on the TCGA Pancancer dataset encompassing five cancer types, the proposed model achieved 99% accuracy, demonstrating superior sensitivity and specificity. This approach provides a reliable and interpretable solution for cancer subtype classification, facilitating improved genomic-based diagnostics.