Abstract
The use of Hypomethylating agents combined with Venetoclax (VH) for the treatment of Acute Myeloid Leukemia (AML) has greatly improved outcomes in recent years. However not all patients benefit from the VH regimen and a way to rationally select between VH and Conventional Chemotherapy (CC) for individual AML patients is needed. Here, we developed a proteomic-based triaging strategy using Reverse-phase Protein Arrays (RPPA) to optimize therapy selection. We evaluated the expression of 411 proteins in 810 newly diagnosed adult AML patients, identifying 109 prognostic proteins, that divided into five patient expression profiles, which are useful for optimizing therapy selection. Furthermore, using machine learning algorithms, we determined a set of 14 proteins, among those 109, that were able to accurately recommend therapy, making it feasible for clinical application. Next, we identified a group of patients who did not benefit from either VH or CC and proposed target-based approaches to improve outcomes. Finally, we calculated that the clinical use of our proteomic strategy would have led to a change in therapy for 30% of patients, resulting in a 43% improvement in OS, resulting in around 2600 more cures from AML per year in the United States.