A targeted phosphoproteomic assay to enable the prediction of response to intensive chemotherapy in acute myeloid leukemia.

Abstract

6521 Background: Intensive chemotherapy (IC; cytarabine plus an anthracycline) remains a standard first-line therapy for fit patients with acute myeloid leukemia (AML), a heterogeneous hematologic malignancy with poor long-term survival. IC induces composite complete remission (cCR) in ~60–70% of cases, and IC response variability persists within genetic/ELN risk strata. As emerging regimens increasingly challenge IC as a default therapy, biology-informed predictors are needed to identify patients unlikely to benefit from IC to avoid unnecessary toxicity. Mass spectrometry (MS)-based phosphoproteomics enables quantitative profiling of phosphopeptides (PPs) from patient samples, providing a functional readout of tumor biology. Thus, we sought to develop a PP-based clinically-deployable assay, orthogonal to genetic risk stratification, to predict response to IC in AML. Methods: 261 retrospective samples were collected from patients with newly-diagnosed AML (Table) treated with standard “7+3” IC in 9 centers in Europe, Australia, and North America. Response was assessed at the end of induction by the treating physician. Samples underwent phosphoproteomics using: 1) global MS for biomarker discovery or 2) targeted, clinically-compatible MS for validation. Predictive model performance was evaluated using rebalanced leave-one-out cross-validation. Results: During discovery, 3205 PPs were detected. Using Bayesian approaches, we identified an 80-PP multi-analyte signature of IC response (refractory vs cCR), and achieved an AUROC of 0.68 (95% CI 0.54-0.83). The signature was enriched for DNA damage signalling and repair (DNA-PK-S2612, ATM/cohesin, nucleotide excision and double-strand break repair proteins), and cellular stress pathways (phospho-p38, IL-16, AP-1/c-Jun). In an independent validation cohort, the targeted assay reliably detected the biomarker set and preserved its association with outcome (AUROC 0.69; 95% CI 0.53-0.84). Biomarker-based classification was associated with improved event-free survival (HR 0.45; 95% CI 0.26-0.78). Conclusions: We identified and validated a signature of response to IC in AML, and translated it into a targeted, clinically-deployable assay. This approach captures signaling states relevant to IC mechanisms of action that are not directly inferred from standard clinical or genetic variables. Ongoing analyses are evaluating its relationship to established genetic and ELN risk stratification. These findings support diagnostic phosphoproteomics and suggest functional biomarkers may complement existing approaches for treatment selection in AML. Cohort characteristics. Cohort Discovery Validation Median age at diagnosis (years, quartiles) 54 (23, 65) 59 (47, 68) Median diagnosis year (range) 2014 (1999, 2023) 2013 (2001, 2024) No of patients/samples 135/165 102/106 No of cCR/Refractory 94/41 79/23 PPs 3205 80