Machine learning integrates genomic signatures for subclassification beyond primary and secondary acute myeloid leukemia

Abstract

While genomic alterations drive the pathogenesis of acute myeloid leukemia (AML), traditional classifications are largely based on morphology and prototypic genetic founder lesions define only a small proportion of AML patients. The historical subdivision of primary/de novo AML (pAML) and secondary AML (sAML) has shown to variably correlate with genetic patterns. Perhaps, the combinatorial complexity and heterogeneity of AML genomic architecture have precluded, so far, the genomic-based subclassification to identify distinct molecularly-defined subtypes more reflective of shared pathogenesis. We integrated cytogenetic and gene sequencing data from a multicenter cohort of 6,788 AML patients that were analyzed using standard and machine learning methods to generate a novel AML molecular subclassification with biological correlates corresponding to underlying pathogenesis. Standard supervised analyses resulted in modest cross-validation accuracy when attempting to use molecular patterns to predict traditional pathomorphological AML classifications. We performed unsupervised analysis by applying Bayesian Latent Class method that identified 4 unique genomic clusters of distinct prognoses. Invariant genomic features driving each cluster were extracted and resulted in 97% cross-validation accuracy when used for genomic subclassification. Subclasses of AML defined by molecular signatures overlapped current pathomorphological and clinically-defined AML subtypes. We internally and externally validated our results and share an open-access molecular classification scheme for AML patients. Although the heterogeneity inherent in the genomic changes across nearly 7,000 AML patients is too vast for traditional prediction methods, however, machine learning methods allowed for the definition of novel genomic AML subclasses indicating that traditional pathomorphological definitions may be less reflective of overlapping pathogenesis.

How to Improve Prognostication in Acute Myeloid Leukemia with CBFB-MYH11 Fusion Transcript: Focus on the Role of Molecular Measurable Residual Disease (MRD) Monitoring

“…In recent years, the advances in molecular diagnostic and monitoring, with the simultaneous incoming of new therapeutic agents, have led to significant improvements in clinical AML management. Awada et al [122] recently integrated cytogenetic and gene sequencing data from a multicenter cohort of nearly 7000 AML patients that were analyzed using standard and machine learning methods to generate a novel AML molecular subclassification with biological correlates corresponding to underlying pathogenesis. Despite the heterogeneity of AML genomics, non-random genomic relationships were capable of identifying four novel unique genomic clusters with a distinct prognosis, regardless of the availability of pathomorphological or anamnestic information.…”

A Focus on Intermediate-Risk Acute Myeloid Leukemia: Sub-Classification Updates and Therapeutic Challenges

“…These approaches may be further employed to define ambiguous definitions by unmasking unexplored clinico-morphologic and genomic characteristics unique to intermediate-risk AML patients. Indeed, ML-based methods demonstrated a 97% accuracy rate when reclassifying AML patients according to four genomic signature clusters [ 4 ]. While the clusters exhibited a certain degree of overlap with the 2017 ELN groups, significant differences in survival across the two classification systems emphasized the need for a more subtle distinction of AML risk groups.…”

“…In the ELN 2017, the allelic burden and the pattern of co-mutations (in particular in NPM1 and DNMT3A ) modulated the prognostic significance of the more common and prognostic FLT3 mutation, FLT3 -ITD [ 7 , 61 ]. For instance, FLT3-ITD with a high allelic ratio (>0.5) conferred adverse prognosis; however, the simultaneous presence of an NPM1 mutation made it instead recognized as intermediate-risk [ 4 ]. The new 2022 revision abated this differentiation by assigning all FLT3 mutants to the intermediate-risk category [ 6 ].…”

“…Patient characteristics such as age and performance status (overall, summarized by the concept of fitness [ 2 ]) and heterogeneous cytogenetic and genomic underpinnings constitute the main determinants of outcomes [ 3 ]. Recent advances in genome scanning techniques have increased our understanding of AML pathogenesis, unveiling a plethora of driver gene lesions and possible disease-outcome modifiers [ 4 ]. Despite the abundance of genomic information on AML, current diagnostic and prognostic classifications somehow remain inexact.…”

The Anti-Leukemia Effect of Ascorbic Acid: From the Pro-Oxidant Potential to the Epigenetic Role in Acute Myeloid Leukemia

“…Acute myeloid leukemia (AML) is a heterogeneous clonal disorder characterized by the uncontrolled proliferation of undifferentiated myeloid progenitor cells in the bone marrow and peripheral blood. Advances in DNA sequencing technologies have provided a detailed knowledge of the molecular landscape of AML, with a better understanding of the disease pathogenesis and prognosis (The Cancer Genome Atlas Research Network, 2013;Papaemmanuil et al, 2016;Awada et al, 2021). Despite high heterogeneity, the spectrum of genetic alterations have highlighted the presence of recurrent mutations in genes encoding epigenetic regulators (Gallipoli et al, 2015), including DNA methyltransferase 3A (DNMT3A), Ten-eleventranslocation 2 (TET2), Wilms' tumor 1 (WT1), and isocitrate dehydrogenase 1 and 2 (IDH1/2) (Kunimoto and Nakajima, 2017).…”