Signal transduction pathways are chemical communication channels embedded in biological cells, and they propagate information from the environment to regulate cell growth and proliferation, among other cell's behaviors. Disruptions in the normal functionalities of these channels, mostly resulting from mutations in the underlying genetic code, can be leading causes of diseases, such as cancer. Motivated by the increasing availability of public data on genetic code expression in cell tissue samples, i.e., transcriptomics, and the emerging field of molecular communication, a novel data-driven approach based on experimental data mining and communication theory is proposed in this paper. This approach is an alternative to existing computational models of these pathways in the context of cancer, which often appear to oversimplify the complexity of the underlying mechanisms. In contrast, a computational methodology is here derived to estimate the difference in information propagation performance of signal transduction pathways in healthy and diseased cells, solely based on transcriptomic data. This methodology is built upon a molecular communication abstraction of information flow through the pathway and its correlation with the expression of the underlying DNA genes. Numerical results are presented for a case study based on the JAK-STAT pathway in kidney cancer, and correlated with the occurrence of pathway gene mutations in the available data.