Fig. 6: KSTAR applied to breast cancer biopsies in three studies.
From: KSTAR: An algorithm to predict patient-specific kinase activities from phosphoproteomic data
HER2 is used when referring to clinical diagnosis and ERBB2-activity for ERBB2/HER2 activity predictions. a KSTAR predictions of ERBB2-activity for the 77 breast cancer patients in the CPTAC dataset5 and their clinical IHC/FISH HER2-status is given (samples are ranked by ERBB2-activity prediction score). The table gives the total number of HER2-positive and HER-negative patients and the KSTAR predictions for ERBB2/HER2 activity for the best of three cutoffs (score-based) considered for designation of ERBB2-active: FPR < = 0.05, FPR < = 0.1, and score >3. b Predictions of EGFR and ERBB2 activities for the patient-derived xenograft (PDX) models published in Huang et al.8 subset that were treated with lapatinib (EGFR/HER2 targeted therapy), where WHIM14 is a HER2-negative tumor that showed a surprising response to lapatinib treatment. The table reports the HER2-status of all 25 PDX tumors and KSTAR ERBB2/HER2-activity predictions. c The ERBB2-activity predictions for tumor biopsies of patients enrolled in a HER2-positive study by Satpathy et al.4. Five patients were non-pathologically complete responders (non-pCR) and the remainder were pathologically complete responders (pCR). Biopsies were taken pre-treatment and most patient's also had an on-treatment biopsy taken with phosphoproteomic profiling. The first three non-responders were reclassified for HER2-status upon additional analysis in Satpathy et al. and results are shown as one false positive and two classified as “Pseudo-positives”. Full KSTAR results for data in this figure available in Supplementary Note 6. Source data are provided with this paper.
