Built a data-driven framework to investigate tumor-specific signaling using large-scale phosphoproteomics data from CPTAC (~18k phosphosites, 165 paired samples).

Key contributions:

• Designed a paired tumor–normal statistical pipeline to reduce inter-patient variability
• Constructed a differential phosphosite co-regulation network using Spearman correlation
• Identified signaling modules via Louvain community detection
• Performed kinase enrichment analysis (KEA) to infer regulatory pathways
• Discovered a CDK-dominant signaling module and prioritized candidate “dark phosphosites” (ILF3, RFC1, NCL)

The entire workflow was implemented on Azure Machine Learning, enabling scalable computation and reproducible analysis of high-dimensional omics data.