Despite the critical role of DNA methylation, clinical implementations harnessing its promise have not been described in acute myeloid leukemia. Utilizing DNA methylation from 3314 leukemia patient samples across 11 harmonized cohorts, we describe the Acute Leukemia Methylome Atlas, which includes robust models capable of accurately predicting AML subtypes. A genome-wide prognostic model as well as a targeted panel of 38 CpGs significantly predict five-year survival in our pediatric and adult test cohorts. To accelerate rapid clinical utility, we develop a specimen-to-result protocol that uses long-read nanopore sequencing and machine learning to characterize patients’ whole genomes and epigenomes. Clinical validation on patient samples confirms high concordance between epigenomic signatures and genomic lesions, though uniquely rare karyotypes remained challenging due to limited available training data. These results unveil the potential for increased affordability, speed, and accuracy for patients in need of complex molecular diagnosis and prognosis.

Acute myeloid leukemia (AML) is a clonal hematopoietic disease that arises from chromosomal and genetic aberrations in myeloid precursor cells. AML is one of the most common types of acute leukemia in adults; however, it is relatively rare overall, comprising about 1% of all cancers. In the last decade or so, numerous genome-wide association studies (GWAS) have been conducted to screen between hundreds of thousands and millions of variants across many human genomes to discover genetic polymorphisms associated with a particular disease or phenotype. In oncology, GWAS has been performed in almost every commonly occurring cancer. Despite the increasing number of studies published regarding other malignancies, there is a paucity of GWAS studies for AML. In this review article, we will summarize the current status of GWAS in AML.

Cytarabine arabinoside (Ara-C) has been the cornerstone of acute myeloid leukemia (AML) chemotherapy for decades. After cellular uptake, it is phosphorylated into its active triphosphate form (Ara-CTP), which primarily exerts its cytotoxic effects by inhibiting DNA synthesis in proliferating cells. Interpatient variation in the enzymes involved in the Ara-C metabolic pathway has been shown to affect intracellular abundance of Ara-CTP and, thus, its therapeutic benefit. Recently, SAMHD1 (SAM and HD domain–containing deoxynucleoside triphosphate triphosphohydrolase 1) has emerged to play a role in Ara-CTP inactivation, development of drug resistance, and, consequently, clinical response in AML. Despite this, the impact of genetic variations in SAMHD1 on outcome in AML has not been investigated in depth. In this study, we evaluated 25 single nucleotide polymorphisms (SNPs) within the SAMHD1 gene for association with clinical outcome in 400 pediatric patients with newly diagnosed AML from 2 clinical trials, AML02 and AML08. Three SNPs, rs1291128, rs1291141, and rs7265241 located in the 3′ region of SAMHD1 were significantly associated with at least 1 clinical outcome: minimal residual disease after induction I, event-free survival (EFS), or overall survival (OS) in the 2 cohorts. In an independent cohort of patients from the COG-AAML1031 trial (n = 854), rs7265241 A>G remained significantly associated with EFS and OS. In multivariable analysis, all the SNPs remained independent predictors of clinical outcome. These results highlight the relevance of the SAMHD1 pharmacogenomics in context of response to Ara-C in AML and warrants the need for further validation in expanded patient cohorts.

In this study, we leveraged machine-learning tools by evaluating expression of genes of pharmacological relevance to standard-AML chemotherapy (ara-C/daunorubicin/etoposide) in a discovery-cohort of pediatric AML patients (N = 163; NCT00136084) and defined a 5-gene-drug resistance score (ADE-RS5) that was predictive of outcome (high MRD1 positivity p = 0.013; lower EFS p < 0.0001 and OS p < 0.0001). ADE-RS5 was integrated with a previously defined leukemic-stemness signature (pLSC6) to classify patients into four groups. ADE-RS5, pLSC6 and integrated-score was evaluated for association with outcome in one of the largest assembly of ~3600 AML patients from 10 independent cohorts (1861 pediatric and 1773 adult AML). Patients with high ADE-RS5 had poor outcome in validation cohorts and the previously reported pLSC6 maintained strong significant association in all validation cohorts. For pLSC6/ADE-RS5-integrated-score analysis, using Group-1 (low-scores for ADE-RS5 and pLSC6) as reference, Group-4 (high-scores for ADE-RS5 and pLSC6) showed worst outcome (EFS: p < 0.0001 and OS: p < 0.0001). Groups-2/3 (one high and one low-score) showed intermediate outcome (p < 0.001). Integrated score groups remained an independent predictor of outcome in multivariable-analysis after adjusting for established prognostic factors (EFS: Group 2 vs. 1, HR = 4.68, p < 0.001, Group 3 vs. 1, HR = 3.22, p = 0.01, and Group 4 vs. 1, HR = 7.26, p < 0.001). These results highlight the significant prognostic value of transcriptomics-based scores capturing disease aggressiveness through pLSC6 and drug resistance via ADE-RS5. The pLSC6 stemness score is a significant predictor of outcome and associates with high-risk group features, the ADE-RS5 drug resistance score adds further value, reflecting the clinical utility of simultaneous testing of both for optimizing treatment strategies.

Purpose: Cytarabine (also known as ara-C) has been the backbone of acute myeloid leukemia (AML) chemotherapy for more than five decades. Recent pharmacogenomics-based 10-SNP ara-C (ACS10) scores showed low ACS10 (≤0) to be associated with poor outcomes in patients with AML treated with standard chemotherapy. Here, we evaluated the ACS10 score in the context of three different induction I regimens in patients with pediatric AML. Experimental Design: ACS10 score groups (low, ≤0, or high, >0) were evaluated for association with event-free survival (EFS) and overall survival (OS) by three randomized treatment arms in patients treated on the AML02 (NCT00136084) and AML08 (NCT00703820) clinical trials: AML02 low-dose ara-C (LDAC arm, n = 91), AML02 + AML08 high-dose ara-C (HDAC arm, n = 194), and AML08 clofarabine + ara-C (Clo/ara-C arm, n = 105) induction I regimens. Results: Within the low-ACS10 score (≤0) group, significantly improved EFS and OS were observed among patients treated with Clo/ara-C as compared with LDAC (EFS, HR = 0.45; 95% CI, 0.23–0.88; P = 0.020; OS, HR = 0.44; 95% CI, 0.19–0.99; P = 0.048). In contrast, within the high-ACS10 score group (score >0), augmentation with Clo/ara-C was not favorable as compared with LDAC (Clo/ara-C vs. LDAC, EFS, HR = 1.95; 95% CI, 1.05–3.63; P = 0.035; OS, HR = 2.10; 95% CI, 0.96–4.59; P = 0.063). Personalization models predicted 9% improvement in the outcome in ACS10 score–based tailored induction (Clo/ara-C for low and LDAC for high-ACS10 score groups) as compared with nonpersonalized approaches (P < 0.002). Conclusions: Our findings suggest that tailoring induction regimens using ACS10 scores can significantly improve outcomes in patients with AML. Given the SNPs are germline, preemptive genotyping can accelerate matching the most effective remission induction regimen.