Abstract 2262: Inference of kinase activity for cancer phosphoproteomics using substrate prediction scores

Abstract

Abstract Background: Quantification of global protein phosphorylation abundance levels with residue-level resolution is feasible with mass spectrometry and phosphopeptide enrichment methods. Researchers have inferred kinase activity from phosphoproteomics data using a variety of methods to extract interpretable and biologically relevant information, but analysis of phosphoproteomics data is limited by the fact that detected phosphorylation sites often have unknown function and relevance. Methods: We developed a novel method for aggregating and interpreting phosphoproteomics data that combines kinase prediction scores from NetworKIN with experimental summary statistics to estimate abundance changes in predicted kinase substrates. By weighting student t-statistic scores by kinase prediction scores and summing cumulative scores across all phosphopeptides detected in a given screen, we generated kinase activity scores that reflect the overall direction, magnitude and significance of observed changes in predicted kinase substrates. These kinase scores were assessed using permutation testing to determine their significance versus changes that would be expected from random chance. To evaluate our method, we applied it to a shotgun phosphoproteomic dataset generated to study phosphorylation responses to rapamycin, dasatinib and combination treatment in MDA-MB-231 breast cancer cells. Results: We were able to confirm phosphoproteomic perturbations in expected “hallmark” kinase activities using prediction-based kinase activity inference scores, including decreased phosphorylation of Src and Abl substrates in response to dasatinib treatment, as well as decreases in p70S6K substrate phosphorylation in response to rapamycin. Compared to KSEA, we identified changes in substrate abundance for a variety of potential downstream kinases including DNAPK, PAK2, CDK3, and multiple members of the MAPK family that were observed to be downregulated in single-agent and in combination treatment. Conclusions: Our novel method provides a representation of phosphoproteomic changes based on abundance of predicted kinase substrates. Using predicted substrates allows for a global view of phosphoproteomic changes that extends beyond the relatively small number of known kinase-substrate relationships. Using a drug combination with potential application in breast cancer as an example, we demonstrate that this method can be used to identify phosphoproteomic changes and can potentially be used to rationally design and investigate novel drug combinations. Citation Format: Peter Liao, Jennifer Yori, Ruth Keri, Mehmet Koyuturk, Jill Barnholtz-Sloan. Inference of kinase activity for cancer phosphoproteomics using substrate prediction scores [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2262.