American Journal for Young Scientists ยท ISSN 3068-9015
Tumor Resource Allocation using Transcriptomes and AI
Authors: Bhavya Dwivedi
Published: July 20, 2026
Category: Biology
Abstract
Cancer cells survive by reprogramming their gene expression to support growth, metabolism, and immune evasion, but they must balance these competing demands. Here, we analyze transcriptome data from twelve cancer types to determine what tumors are biologically optimized for. We apply a machine-learning framework combining a variational autoencoder with classifiers to learn compact representations of gene expression and accurately distinguish cancer types, achieving about 98% accuracy. The learned latent features reveal dominant biological pathways, including metabolic, immune, and proliferative programs. By viewing cancer as a system that must make trade-offs to survive, this approach helps identify the biological pathways that each tumor relies on most and could be targeted by treatment.