Acute myeloid leukemia (AML) is a rapidly progressing blood cancer in which immature blood cells called myeloid blasts accumulate in the bone marrow, crowding out healthy cells. Diagnosing and managing AML requires integrating information from multiple disciplines simultaneously - no single test is sufficient.
Traditional AML diagnostics relied on cytomorphology - physically examining blood and bone marrow cells under a microscope to count blast cells and assess their appearance. While indispensable, morphology alone cannot tell oncologists which specific mutations are driving the disease or which treatments will be most effective.
Modern AML diagnosis now layers multiple technologies: flow cytometry (identifying proteins on cell surfaces), cytogenetics (detecting chromosome abnormalities), PCR (measuring specific known mutations), and increasingly, next generation sequencing (NGS) - which reads the entire DNA sequence of hundreds of relevant genes in a single experiment.
The explosion of molecular knowledge about AML over the past two decades has transformed it from a single disease into a collection of dozens of molecularly distinct subtypes. This complexity demands molecular tools capable of mapping the full genetic landscape of each patient's leukemia.
Next generation sequencing (NGS) is a technology that reads DNA sequences in a massively parallel fashion - analyzing thousands to millions of DNA fragments simultaneously. Unlike older methods such as Sanger sequencing (which reads one region at a time), NGS can interrogate hundreds of genes in one experiment within a week.
The most common NGS approach in clinical AML practice is targeted panel testing: sequencing a predefined set of genes known to be mutated in AML. Panels can include the mutations needed for WHO disease classification, risk stratification, and identification of targetable lesions - all from one sequencing run, saving both time and sample.
Beyond panels, NGS encompasses three broader approaches: whole exome sequencing (WES) covers all protein-coding genes; whole genome sequencing (WGS) covers the entire DNA, including regulatory regions; and whole transcriptome sequencing (WTS) measures RNA expression levels and detects gene fusions. As costs fall and computing power grows, these genome-wide approaches are moving from research tools toward clinical reality.
The technical workflow is demanding - samples must be carefully processed, libraries prepared with unique molecular tags to reduce errors, sequenced to adequate depth (typically over 1,000 reads per region to detect mutations present in as few as 10% of cells), and analyzed with sophisticated bioinformatics software to distinguish true mutations from sequencing artifacts.
The 2017 WHO classification of AML requires testing three genes at minimum: NPM1 (mutated in 25-35% of AML, generally favorable prognosis), CEBPA (biallelic mutations confer favorable prognosis), and RUNX1 (mutations define a provisional disease category). These mutations determine which type of AML a patient has and influence initial treatment planning.
The European LeukemiaNet (ELN) risk stratification system - the primary tool oncologists use to decide who needs a bone marrow transplant - requires testing five additional genes beyond WHO classification: FLT3-ITD (internal tandem duplication, adverse), ASXL1 (adverse), TP53 (adverse), and updates to the NPM1/CEBPA context. Patients classified as high-risk are candidates for allogeneic stem cell transplantation, which is the only potentially curative option for many.
Comprehensive NGS testing finds at least one mutation in over 90% of AML patients - far more than cytogenetics alone (which detects abnormalities in only 50-60%). Most AML patients carry multiple mutations simultaneously, with up to 13 different mutations found in some patients. This genetic complexity makes NGS indispensable for capturing the full disease picture.
Several AML-associated mutations now directly guide therapy choice. FLT3-ITD mutations indicate eligibility for midostaurin or gilteritinib (approved FLT3 inhibitors). IDH1 and IDH2 mutations indicate eligibility for targeted IDH inhibitors. CD33 expression guides use of gemtuzumab ozogamicin. NGS simultaneously identifies all these therapeutic targets in one test.
Measurable residual disease (MRD) refers to the tiny number of leukemia cells that survive initial chemotherapy, undetectable by standard microscopy but detectable by sensitive molecular methods. The presence of MRD after treatment is one of the strongest predictors of relapse in AML.
The gold standard for MRD monitoring currently requires sensitivity of 1 in 10,000 to 1 in 1,000,000 cells - levels achievable by PCR-based methods and flow cytometry. Standard NGS typically reaches only about 1% sensitivity, making it less suitable for routine MRD quantification. However, with optimized protocols, NGS-based MRD monitoring has shown clinical value in several studies, particularly before stem cell transplantation decisions.
Not every mutation detected at diagnosis is suitable as an MRD marker. Some mutations - particularly DNMT3A, ASXL1, and TET2 - persist even when AML is in remission because they represent age-related changes in blood stem cells called clonal hematopoiesis, which is a normal aging phenomenon unrelated to leukemia recurrence. Using these as MRD markers would generate false alarms.
Established reliable MRD markers in AML include specific gene fusion transcripts (RUNX1-RUNX1T1, CBFB-MYH11, PML-RARA) and NPM1 mutations. Critically, at relapse the genetic landscape often changes from diagnosis - new mutations may emerge and others disappear - making comprehensive re-sequencing at relapse essential for identifying resistance mechanisms and new therapeutic targets.
Precision medicine in AML means selecting targeted drugs based on the specific genetic mutations driving each patient's disease, rather than applying the same chemotherapy regimen to all AML patients. NGS is the engine that makes precision medicine possible by identifying which targetable mutations are present at diagnosis.
The earliest and most successful precision medicine example in AML is acute promyelocytic leukemia (APL), caused by the PML-RARA fusion. Treatment with all-trans retinoic acid (ATRA) and arsenic trioxide achieves cure in the vast majority of APL patients - transforming a previously lethal leukemia subtype into a highly manageable condition. Molecular monitoring of PML-RARA levels guides pre-emptive therapy to prevent relapse.
Midostaurin became the first FDA-approved drug targeting a molecular mutation in non-APL AML - specifically FLT3-ITD mutations. Added to standard chemotherapy, it improved survival for FLT3-mutated patients. More recently, second-generation FLT3 inhibitors such as gilteritinib and quizartinib have shown activity even in patients who relapsed on midostaurin, reflecting an evolving precision medicine landscape in AML.
The identification of IDH1 and IDH2 mutations - present in about 20% of AML - opened another therapeutic avenue. Ivosidenib (IDH1 inhibitor) and enasidenib (IDH2 inhibitor) are approved for relapsed/refractory IDH-mutated AML, and are being incorporated into earlier treatment combinations. Each new approved targeted agent creates a new reason to test for the corresponding mutation, driving expansion of NGS panels.
The sheer volume of data generated by NGS creates an interpretive challenge. When a gene panel identifies dozens of variants in a patient's AML, determining which are disease-driving (pathogenic) versus harmless background variations requires expert review against multiple databases (ClinVar, COSMIC, gnomAD) - a time-consuming process in an acute disease where treatment decisions cannot wait.
Artificial intelligence (AI) is being applied to streamline variant interpretation. Machine learning models trained on large databases of known mutations can automatically predict variant pathogenicity, flagging likely disease-relevant mutations and deprioritizing likely benign variants. This reduces the burden on human experts and accelerates clinical decision-making in time-sensitive AML diagnosis.
As NGS moves toward whole-genome approaches, the data volumes become orders of magnitude larger. Storing, processing, and extracting clinically meaningful patterns from whole-genome sequences of hundreds or thousands of AML patients requires sophisticated bioinformatics infrastructure and data mining techniques that can identify novel genetic interaction patterns invisible to traditional statistical analysis.
Future AI applications in AML genomics will likely go beyond variant calling to pattern recognition at the disease level - identifying combinations of mutations, expression changes, and epigenetic alterations that collectively predict response to specific therapies or risk of transformation, enabling truly individualized treatment planning that no single biomarker can provide alone.
NGS panel testing is now the standard of care in AML diagnostics, providing in a single experiment all the genetic information needed for WHO classification, ELN risk stratification, and targeted therapy planning. The era of testing one gene at a time has ended - comprehensive parallel sequencing is both more efficient and more clinically informative.
The next frontier is expanding NGS toward whole-genome and whole-transcriptome approaches in routine diagnostics. These broader methods can detect chromosomal rearrangements, copy number changes, fusion transcripts, and expression abnormalities beyond what targeted panels capture, potentially revealing therapeutic vulnerabilities in patients who currently lack targetable mutations.
Falling sequencing costs and increasing computing power suggest that genome-wide NGS will enter routine AML diagnostics within the next five years, according to the authors. This transition will require substantial investment in bioinformatics infrastructure and computational expertise, as the limiting factor is no longer sequencing itself but interpreting what the sequences mean.
The convergence of comprehensive NGS, AI-powered data interpretation, and an expanding arsenal of targeted therapies represents a fundamental transformation of AML care - from a largely uniform treatment approach to a molecularly guided, patient-specific strategy that matches each person's leukemia biology to the interventions most likely to achieve lasting remission or cure.