Predicting Response to BET Inhibitors Using a Computational Model and Its Validation: A Beat AML Project Study

Background: Despite advances in our understanding of the molecular pathogenesis of acute myeloid leukemia (AML), the prognosis remains poor, mainly due to high relapse rates. Therefore, novel treatment strategies are needed to improve therapeutic outcomes for AML patients. Epigenetic regulators are promising targets in cancer therapy, since dysregulations and mutations in chromatin modulators and transcription factors (TF) are commonly detected in many malignancies. Additionally, recurrent genetic events involving epigenetic regulators have been established in AML. Bromodomain (BRD) and extra-terminal protein family (BET) inhibitors interfere with transcriptional regulators, such as BRD4, by preventing them from interacting with acetylated regions of the genome, inhibiting the transcriptional activation of BRD4 target genes. Selective BET inhibitors, such as JQ1, have been shown to be highly effective against multiple subtypes of leukemia, by downregulating MYC transcription. Most BET inhibitors are in early-phase clinical trials (Phase I), where defined biomarkers and methods to predict drug response will expedite the development process towards FDA approval.

Aim: To create a digital drug model for BET inhibitors using computational modeling, and validate the model using primary AML patient samples treated with a BET inhibitor (JQ1) in an ex vivo drug sensitivity assay.

Methods: The Beat AML project (supported by the Leukemia & Lymphoma Society) collects clinical data and bone marrow specimens from AML patients (>800). Bone marrow samples were analyzed by conventional cytogenetics, whole-exome sequencing (WES), RNA-seq, and an ex vivo drug sensitivity assay. For 100 of randomly selected Beat AML patients, every available genomic abnormality was entered into a computational biology modeling (CBM) program (Cellworks Group) that uses PubMed and other online resources to generate patient-specific protein network maps of activated and inactivated protein pathways. Digital drug simulations with BET inhibitor JQ1 were conducted by quantitatively measuring drug effect on a composite AML disease inhibition score (i.e., cell proliferation, viability, and apoptosis). Computational predictions of drug response were compared to JQ1 IC₅₀ values from the Beat AML ex vivo chemosensitivity assays.

Results: Eighty-six percent (n=86) of the patient samples were predicted by computer simulation to show sensitivity to JQ1, of whom 84 had the lowest ex vivo JQ1 IC₅₀ values in the cohort (<0.0946). Fourteen percent (n=14) of patients were predicted to show resistance to JQ1, of which 9 had the highest IC₅₀ values (>1). Ex vivo JQ1 responses were correctly matched to their computer simulation prediction in 93 of 100 cases (93%), and incorrectly matched in 7 cases (7%), resulting in 94.38% positive predictive value, 81.82% negative predictive value, 97.67% sensitivity, and 64.29% specificity for the computational method.

From this analysis, the patient samples harboring chromosomal aberrations del(7q) or monosomy 7 (n=4), trisomy 8 (n=9), and del(5q) (n=7) responded to BET inhibitors, whereas patients with del(22q) were resistant, as this chromosome region contains EP300 , which plays a key role in histone acetylation and transcriptional regulation via chromatin remodeling. All simulated profiles with trisomy 8 had over-amplification of MYC, and were sensitivie to JQ1 treatment. One profile had both trisomy 8 and del(22q), resulting in resistance to JQ1 due to increased EP300 expression.

Conclusion: We developed a digital drug model of JQ1, a BET inhibitor, and validated the CBM predictions by an ex vivo drug sensitivity assay. This computational platform can identify sensitivity and resistance mechanisms for pipeline drugs, and inform the patient population for future clinical trials of BET inhibitors. Further analysis of this dataset will provide genomic and molecular signatures of response/resistance to JQ1 and other BET inhibitors.