Predictive Modeling Case Study on Statistical Difficulties in Combining Genomic and Proteomic Data

A key challenge in modern biology is integrating different types of molecular data. This study examines the specific relationship between gene copy number and protein expression levels. Using data from the Cancer Cell Line Encyclopedia (CCLE), we find that this relationship varies significantly by gene. For the MYC oncogene, copy number strongly predicts protein levels (R2 = 0.37), indicating that more gene copies generally lead to more protein. However, for the TP53 tumor suppressor, copy number poorly predicts protein abundance (R2 = 0.08), suggesting that other regulatory mechanisms dominate. These results show that simple statistical models are often insufficient for biological data, and more advanced approaches are needed to understand complex gene-protein relationships.