Machine learning-based prediction of the first-line AML treatment response

• Applicant: Dr. Piyush More, Ph.D.
• Subject Area: Hematology, Oncology ; Bioinformatics and Theoretical Biology ; Pharmacology
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 448405134

AML in the elderly has a dismal prognosis, in a large part due to the toxicity of the initial standard "7+3" induction treatment, and to the limited efficacy of the alternative hypomethylating therapy. Our recently published work on AML-derived cell lines showed that gene expression changes in response to the AML drug etoposide determine drug efficacy and constitute targets for add-on treatment refinements. Herein, we will investigate whether gene expression changes could (I) guide the choice between the two aforementioned standard first-line AML therapies and (II) improve their efficacy and safety. To this end, (1) we will determine the ex vivo sensitivity to either first-line treatment in blasts from a panel of AML patients. (2) In parallel, we will assess the accompanying gene expression changes and analyze them for correlation with the ex vivo and clinical responses, and for the presence of druggable targets. (3) We will then assess the impact of these druggable targets as an add-on to the respective standard AML treatment ex vivo, and in vivo using mice xenografts. Specifically, we will use the add-on drugs to reduce the intensity and thereby toxicity of the "7+3" therapy and to improve the efficacy of the hypomethylating therapy. (4) The results will be combined with all available clinical and molecular data to develop a machine learning-based algorithm for selecting the optimal first-line induction therapy, its intensity, and add-on drugs. The innovative part of the application is that the gene expression changes will be predicted from the basal expression from AML patient samples using a machine learning algorithm. Predicted changes will then be used to formulate better drug combinations. These results will set the stage for a prospective validation in ex vivo patient samples. The entire approach may further be applicable to other cancer entities to make broad use of available drugs to improve the efficacy and safety of cancer treatments.

• DFG Programme: Research Grants
• Co-Investigator: Professor Dr. Thomas Kindler