Despite extensive DNA sequencing of patient tumors, it remains challenging to translate the immense landscape of heterogeneous genetic alterations into function and clinical outcomes due to a limited understanding of cancer specific molecular network architectures. To bridge this gap, we have used affinity purification-mass spectrometry to generate protein interaction networks for 31 proteins with significant alterations in head and neck squamous cell carcinoma. This network includes 771 interactions covering both cancer and non-tumorigenic cell states, as well as wild-type and mutant proteins. Differential analysis across these dimensions reveals a strong interaction between PIK3CA and ERBB3 (HER3), dependent upon mutations in PIK3CA. We show that this interaction correlates with ERBB3 activity in vitro and can be targeted in vivo using a clinical ERBB3 inhibitor, CDX3379, to prohibit growth of tumors with common PIK3CA mutations. This study provides a roadmap for elucidating genetic complexity through multidimensional maps of cancer cell biology.