Analytical Evaluation of Whole Genome Sequencing for Acute Myeloid Leukemia

Article activity feed

1. 

   PREreview

   Sep 26, 2026

   This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/22983208.

   Short summary of the research and its contribution to the field

   This preprint evaluates a high-depth, tumor-only whole genome sequencing workflow (WGS TO) for comprehensive genomic profiling of acute myeloid leukemia (AML). The workflow combines commercially available DNA extraction and PCR-free library preparation with NovaSeq 6000 sequencing and DRAGEN Somatic tumor-only analysis, targeting approximately 220× genome coverage. The analytical evaluation includes 68 clinical AML samples, healthy donor samples, cancer cell lines, and reference material.

   The study assesses several clinically important classes of genomic alterations, including single-nucleotide variants and small …

   This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/22983208.

   Short summary of the research and its contribution to the field

   This preprint evaluates a high-depth, tumor-only whole genome sequencing workflow (WGS TO) for comprehensive genomic profiling of acute myeloid leukemia (AML). The workflow combines commercially available DNA extraction and PCR-free library preparation with NovaSeq 6000 sequencing and DRAGEN Somatic tumor-only analysis, targeting approximately 220× genome coverage. The analytical evaluation includes 68 clinical AML samples, healthy donor samples, cancer cell lines, and reference material.

   The study assesses several clinically important classes of genomic alterations, including single-nucleotide variants and small insertions/deletions, structural variants (SVs), copy-number alterations (CNAs), and loss of heterozygosity. The reported analytical sensitivities were 97.6% for small variants, 89.5% for structural variants, and 92.9% for copy-number alterations. The workflow also detected all seven FLT3 internal tandem duplications included in the clinical comparison.

   An additional strength is the evaluation of limit of detection as a function of sequencing depth and variant allele frequency. At 140× coverage, the authors estimate a 95% detection rate at approximately 5% VAF for small variants and 7.3% VAF for structural variants. The study also evaluates detection thresholds for copy-number alterations and loss of heterozygosity. The complete workflow is reported to have a turnaround time of approximately five days from DNA extraction to reporting.

   Overall, this work provides useful analytical evidence supporting high-depth WGS as a potentially comprehensive approach to AML genomic profiling and contributes to the continuing discussion about how genome-wide sequencing might be incorporated into routine molecular haematopathology.

   General assessment

   I found this to be a well-designed and clinically relevant analytical study. AML is particularly suitable for evaluating comprehensive genome sequencing because disease classification, prognostic assessment, and therapeutic decision-making depend on a broad spectrum of genomic abnormalities. In current practice, these abnormalities may require a combination of targeted sequencing, cytogenetics, FISH, and other molecular methods.

   An important contribution of this study is therefore the demonstration that a single WGS workflow can interrogate several different classes of genomic alterations simultaneously while maintaining a clinically relevant turnaround time.

   The manuscript is generally clear, and the analytical validation strategy is strengthened by the inclusion of different types of clinical and reference material and comparison with several orthogonal methods. I have several suggestions that could further improve the presentation and help readers understand the potential clinical implementation of the approach.

   Strengths

   One of the strongest aspects of this work is the breadth of the analytical evaluation. The assay is not limited to SNVs and indels but also evaluates structural variants, copy-number changes, and loss of heterozygosity. This is important for AML, where clinically meaningful alterations extend across multiple variant classes.

   The high analytical sensitivity for small variants is encouraging. In particular, detection of all seven FLT3-ITDs included in the comparison is noteworthy because FLT3 alterations are clinically important in AML and can present technical challenges for some sequencing approaches.

   The use of multiple orthogonal reference methodologies is another valuable feature. Comparison with sequencing-based approaches as well as FISH and karyotyping reflects the heterogeneous diagnostic environment in which a comprehensive WGS assay would ultimately be used.

   The analytical precision experiments are also useful. Fifteen clinical AML samples were examined across replicates, and the clinical samples showed very high intra-run and inter-run concordance. The additional evaluation using cell-line and reference materials provides further information about assay behaviour close to the detection limit.

   Finally, the approximately five-day turnaround time is potentially important from a translational perspective. The authors report less than one day for wet-laboratory procedures, approximately two days for sequencing, and around 13 hours for bioinformatic analysis. Such a workflow could make comprehensive genomic information available within a timeframe relevant to AML diagnostic work-up and treatment planning.

   Comments and suggestions for the authors

   1. Clinical positioning of WGS TO

   It would be useful to expand slightly on how the authors envisage WGS TO being incorporated into the current AML diagnostic pathway. The manuscript demonstrates that several different classes of variants can be evaluated within one workflow, which raises an important practical question: should WGS TO be viewed primarily as a complement to existing cytogenetic and molecular testing, or could it eventually consolidate some of these assays?

   A short schematic showing the conventional AML testing pathway alongside a potential WGS-based pathway could make the translational implications of the study particularly clear.

   2. Tumor-only workflow

   The tumor-only design is a practical strength because it avoids the requirement for a matched normal sample and may simplify implementation. At the same time, readers would benefit from additional information about how potentially germline findings would be managed in a clinical setting.

   For example, the authors could briefly describe when confirmatory germline testing would be recommended and how the bioinformatic workflow distinguishes likely somatic events from variants that may warrant further investigation. This would help connect the analytical validation presented here with future routine clinical reporting.

   3. Structural variants and copy-number alterations

   The reported performance for structural and copy-number alterations is encouraging. The manuscript would be further strengthened by providing a concise summary of the types of SVs and CNAs that were most successfully detected and those that proved more technically challenging.

   Such information could be especially useful for laboratories considering implementation because it would identify the genomic contexts in which WGS performs particularly well and where complementary testing may still be useful.

   The authors also report additional variants detected by WGS TO that were not identified using the reference methods, including additional coding small variants, SVs, and CNAs. These observations are interesting and illustrate the potential breadth of high-depth WGS. Future orthogonal confirmation of a representative subset would further clarify the added diagnostic information provided by this workflow.

   4. Limit-of-detection analysis

   The LoD analysis is an important strength of the paper. The combination of titration, computational downsampling, and modelling provides useful information on the relationship between sequencing coverage and detection of low-frequency variants.

   For clinical readers, it might be helpful to summarise these findings in a single table showing, for each variant class, the recommended sequencing depth, estimated LoD, and corresponding analytical sensitivity. This could make the validation results easier to translate into laboratory practice.

   The observations close to the detection threshold are also informative. Reporting performance by VAF ranges, for example 2–5%, 5–10%, and above 10%, could provide additional practical information for interpretation of low-frequency subclonal alterations.

   5. Turnaround time and implementation

   The approximately five-day turnaround time is one of the most clinically attractive findings. It would be useful to provide some additional operational detail regarding whether this timing was consistent across sequencing runs and how much of the five-day period is attributable to laboratory processing, sequencing, bioinformatics, variant review, and final interpretation.

   Future prospective implementation studies could also assess turnaround time from receipt of a clinical specimen to a signed diagnostic report. This would complement the analytical workflow reported in the present study and provide an informative benchmark for routine laboratory practice.

   6. Future validation

   The study provides a strong foundation for further evaluation. A particularly valuable next step would be prospective testing of consecutive newly diagnosed AML cases, with WGS TO performed alongside the complete standard diagnostic work-up.

   Such a study could examine not only analytical concordance but also whether WGS provides information affecting WHO/ICC classification, ELN risk stratification, selection of targeted therapies, or other clinical decisions.

   Independent validation in additional clinical laboratories would also be informative. Many of the authors are employees of Illumina, which develops several of the technologies evaluated in the study, and this relationship is transparently disclosed in the manuscript. Replication in independent laboratories would provide useful evidence regarding reproducibility and generalisability across different clinical environments.

   Overall assessment

   This is a valuable analytical validation study of high-depth tumor-only whole genome sequencing for AML. The manuscript demonstrates strong performance across multiple clinically relevant variant classes and provides useful data on sensitivity, precision, sequencing depth, and limit of detection.

   The ability to evaluate small variants, structural variants, copy-number alterations, and loss of heterozygosity within a single sequencing workflow is particularly attractive. The approximately five-day turnaround time further supports the potential clinical applicability of this approach.

   The work provides an important foundation for moving from analytical validation toward prospective clinical implementation. Further evaluation in independent and prospectively recruited cohorts will help determine how WGS TO can best complement or potentially consolidate existing AML genomic testing strategies.

   Overall, the study represents a useful contribution to the development of comprehensive genomic diagnostics for AML and provides encouraging evidence for the feasibility of integrating high-depth WGS into clinically relevant laboratory workflows.

   Competing interests

   The authors declare that they have no competing interests.

   Use of Artificial Intelligence (AI)

   The authors declare that they did not use generative AI to come up with new ideas for their review.

   Read the original source
2. 

   PREreview

   May 6, 2026

   This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/20047186.

   Short summary of the research and contribution to the field

   This preprint evaluates a tumor-only, high-depth whole genome sequencing workflow (WGS TO) for acute myeloid leukemia (AML), designed to detect clinically relevant small variants, structural variants (SVs), copy number alterations (CNAs), and loss of heterozygosity in a single assay. The authors describe a workflow using commercially available extraction methods, Illumina PCR-free library preparation, NovaSeq 6000 sequencing, and DRAGEN Somatic tumor-only analysis, targeting approximately 220× coverage with a reported turnaround time of approximately 5 days from DNA extraction to reporting.

   The study reports strong analytical …

   This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/20047186.

   Short summary of the research and contribution to the field

   This preprint evaluates a tumor-only, high-depth whole genome sequencing workflow (WGS TO) for acute myeloid leukemia (AML), designed to detect clinically relevant small variants, structural variants (SVs), copy number alterations (CNAs), and loss of heterozygosity in a single assay. The authors describe a workflow using commercially available extraction methods, Illumina PCR-free library preparation, NovaSeq 6000 sequencing, and DRAGEN Somatic tumor-only analysis, targeting approximately 220× coverage with a reported turnaround time of approximately 5 days from DNA extraction to reporting.

   The study reports strong analytical performance across variant classes: 97.6% sensitivity for small variants, 89.7% for SVs, and 92.9% for CNAs across 68 AML clinical samples, using multiple orthogonal reference methods including targeted NGS, WGS, optical genome mapping, FISH, and karyotyping. The authors also estimate limits of detection at 140× coverage: 5% VAF for small variants, 7.3% VAF for SVs, and CNA detection thresholds based on copy-number fold change and tumor purity.

   This work moves the field forward by supporting the feasibility of a rapid, high-depth WGS approach as a potentially comprehensive alternative or complement to the fragmented AML testing landscape, where cytogenetics, FISH, targeted panels, and structural-variant methods are often performed separately. If validated further, this approach could simplify AML genomic profiling and improve detection of clinically relevant alterations across multiple variant classes.

   Positive feedback / strengths

   1. Clinically important problem. AML classification, prognosis, and treatment selection depend heavily on molecular and cytogenetic findings. A single WGS assay capable of detecting SNVs, indels, SVs, CNAs, and LOH could reduce fragmentation across multiple testing platforms.

   2. Comprehensive variant-class coverage. The study evaluates small variants, FLT3-ITDs, SVs, CNAs, and CNLOH, which is appropriate for AML because clinically relevant alterations span several variant types.

   3. Use of multiple orthogonal comparator methods. The authors compare WGS TO against reference sets from targeted sequencing, 60× WGS, FISH, karyotype, and optical genome mapping. This is a strength because each modality captures different classes and resolutions of AML alterations.

   4. Practical turnaround-time goal. The reported ~5-day turnaround time is clinically meaningful for AML, where diagnostic and treatment decisions can be time-sensitive. The workflow breakdown—wet lab under 1 day, sequencing around 2 days, and bioinformatics around 13 hours—is helpful for clinical laboratories assessing feasibility.

   5. Analytical LoD modeling is useful. The use of titration, downsampling, and PROBIT modeling to estimate detection thresholds provides a structured analytical validation framework.

   6. Transparent discussion of discordant findings. The authors discuss missed variants and note that some lacked read support or had inconsistent support across orthogonal methods, which is important for interpreting sensitivity against imperfect reference standards.

   Major issues

   1. Tumor-only WGS interpretation needs more careful discussion

   The assay is explicitly described as a tumor-only WGS workflow. This is practical and faster, but tumor-only analysis can make it harder to distinguish somatic variants from germline variants, clonal hematopoiesis, sequencing artifacts, or rare inherited variants.

   Suggested improvement: The authors should discuss the limitations of tumor-only interpretation more explicitly, especially for AML where blood or marrow samples may contain malignant cells, preleukemic clones, clonal hematopoiesis, and germline predisposition variants. It would be helpful to describe:

   • whether matched normal testing is recommended in specific cases

   • how possible germline variants are filtered or flagged

   • how CHIP/preleukemic variants are handled

   • whether clinically important germline predisposition genes are excluded, reported, or referred for confirmatory germline testing

   • how tumor-only filtering affects sensitivity and specificity

   This is especially relevant because at least one small variant was detected but filtered as a possible germline variant.

   2. Discordant SV and CNA results require deeper orthogonal resolution

   The study reports lower sensitivity for SVs and CNAs compared with small variants, especially when benchmarked against karyotype and FISH. The authors appropriately note that some missed SVs and CNAs had inconsistent support across other methods, but additional adjudication would strengthen the conclusions.

   Suggested improvement: The authors should include a dedicated discordance table for all missed or WGS-only SV/CNA calls showing:

   • variant type and genomic coordinates

   • orthogonal method reporting the event

   • WGS TO read-depth / split-read / discordant-pair support

   • OGM/FISH/karyotype support if available

   • whether the event was clinically relevant

   • whether the discrepancy likely reflects WGS limitation, comparator limitation, tumor heterogeneity, sample differences, or reporting threshold differences

   Because the manuscript notes that additional testing is warranted to resolve conflicting calls, this should be elevated as a key limitation rather than a secondary point.

   3. The reference standard is heterogeneous and may not be equally appropriate for all variant classes

   The comparator set comes from multiple modalities, including targeted sequencing, 60× WGS, FISH, karyotype, and OGM. This is realistic, but it also complicates sensitivity estimates because each method has different resolution, false-positive risk, sensitivity, specimen requirements, and reporting conventions.

   Suggested improvement: The authors should more clearly define the "truth set" construction process. Specifically:

   • Were variants considered true positives if detected by one method only?

   • Were conflicting reference calls adjudicated?

   • Were methods performed on the same specimen or different aliquots/timepoints?

   • Were variant coordinates harmonized across platforms?

   • Were lower-resolution cytogenetic calls converted into genome coordinates?

   • Were WGS-only additional calls orthogonally confirmed?

   This would make the reported sensitivity values easier to interpret.

   4. Precision testing for low-VAF variants should be expanded

   The study reports 100% repeatability/reproducibility for clinical samples and Kasumi-1, but Seraseq showed 92.3% median concordance due to two ASXL1 variants near the LoD that were inconsistently detected across replicates. This is expected near detection limits, but it is highly relevant for AML subclonal detection.

   Suggested improvement: The authors should expand discussion of repeatability near LoD and provide more low-VAF replicate data across variant types. A table showing detection probability by VAF bin—especially 2–5%, 5–10%, and >10%—would be very helpful for clinical interpretation.

   5. The claimed ~5-day turnaround time needs operational context

   The workflow's fast turnaround time is one of the strongest clinical features. However, the study should clarify whether the 5-day TAT reflects a routine clinical workflow or an optimized validation workflow.

   Suggested improvement: Please clarify:

   • whether 5 days is from sample receipt, DNA extraction, or library start

   • whether batching is required to reach this TAT

   • number of samples per run

   • whether weekend/holiday processing was included

   • whether variant interpretation/reporting is included or only VCF generation

   • staffing and computational requirements

   • whether SV/CNA manual review adds additional time

   This would help laboratories assess implementation feasibility.

   6. Clinical reporting scope should be defined more clearly

   The authors state that the assay can identify variants indicative of AML treatment options, but the manuscript would benefit from a more explicit description of clinical reporting.

   Suggested improvement: The manuscript should clarify whether the final report includes:

   • WHO/ICC classification-relevant alterations

   • ELN risk stratification variants

   • therapy-associated mutations

   • MRD-relevant or subclonal variants

   • SV/CNA findings

   • germline-suspected findings

   • variants below LoD or low-confidence findings

   A mock report structure or reportable-variant framework would improve clinical interpretability.

   7. WGS-only additional variants need follow-up validation

   The manuscript reports that WGS TO identified 339 additional small variants, 32 additional SVs, and 122 additional CNAs >5 Mb not detected by orthogonal methods, and the authors state that additional validation is warranted. This is potentially very important, but without orthogonal confirmation or clinical annotation, the meaning of these additional findings is uncertain.

   Suggested improvement: The authors should stratify these WGS-only findings by:

   • likely pathogenic/clinically relevant vs uncertain

   • variant type and size

   • VAF/tumor fraction

   • recurrence in AML genes or cytogenetic regions

   • manual review status

   • orthogonal confirmation status

   • potential clinical impact

   This would clarify whether WGS TO is identifying meaningful additional biology or lower-confidence findings.

   8. Funding and conflicts of interest should be discussed in relation to validation design

   The study includes substantial Illumina involvement, Illumina funding, and multiple Illumina-employed authors. This does not reduce the value of the work, but it increases the importance of transparent benchmarking and independent validation.

   Suggested improvement: The manuscript already provides disclosures and funding information. It would further strengthen the paper to include a statement on independent data review, whether analysis was blinded to comparator results, and whether any validation steps were performed independently of the platform developer.

   Minor issues

   1. Fix inconsistent sensitivity values for SVs. The abstract reports 89.5% SV sensitivity, while Table 2 reports 89.7% and earlier text includes 83.9% in one place. Please harmonize all SV sensitivity values across the abstract, results, tables, and discussion.

   2. Clarify WGS TO versus WGS 60× terminology. Because both are WGS-based methods, the paper should consistently distinguish high-depth WGS TO from the 60× WGS comparator.

   3. Define "clinically important AML genes." The manuscript states that small variants were filtered using a list of 156 clinically important AML genes. Please provide the full list prominently and explain how it was selected.

   4. Clarify CNA size thresholds. The manuscript excludes CNAs below 500 kb from comparator analysis and focuses on CNA calls larger than 5 Mb for some prioritization. Please explain the clinical rationale for these thresholds.

   5. Add confidence intervals for all major performance metrics. Sensitivity, precision, and LoD estimates should include confidence intervals where possible.

   6. Clarify how FLT3-ITD detection was validated. The 100% detection of seven FLT3-ITDs is clinically important. Please provide ITD size range, VAF range, read support, and comparator method.

   7. Improve figure/table readability. Figure 2 and Table 3 are useful, but a simplified summary table separating SNV/indel, FLT3-ITD, SV, CNA, and CNLOH performance would improve readability. The chart on page 13 clearly shows TP/FN distributions across variant classes and comparator methods, and this visual could be paired with a concise interpretation paragraph.

   8. Clarify whether RNA fusions are out of scope. AML testing may include fusion detection by RNA-based methods. Since WGS can identify many SVs but not expression-level fusion evidence, the authors should clarify whether RNA testing remains necessary.

   9. Discuss sample quality requirements. Please include minimum DNA input, DNA quality thresholds, acceptable specimen types, blast percentage/tumor purity requirements, and failure rates.

   10. Clarify clinical deployment status. The manuscript should clearly state whether WGS TO is research-use only, analytically validated for a specific clinical laboratory context, or intended for broader clinical implementation.

   Overall assessment

   This is a valuable and timely analytical validation study of a high-depth tumor-only WGS workflow for AML. The study is strong because it addresses multiple clinically relevant variant classes, uses a rapid workflow, compares performance across multiple orthogonal technologies, and provides LoD estimates for small variants, SVs, CNAs, and CNLOH. The reported 5-day turnaround time and broad variant coverage make the approach potentially important for modern AML diagnostic profiling.

   The major areas needing improvement are clearer handling of tumor-only interpretation, deeper resolution of discordant SV/CNA results, more transparent truth-set construction, expanded low-VAF precision analysis, and more operational detail around turnaround time and clinical reporting. With these additions, the manuscript would provide a stronger and more clinically actionable framework for laboratories evaluating high-depth WGS as a comprehensive AML testing modality.

   Competing interests

   The author declares that they have no competing interests.

   Use of Artificial Intelligence (AI)

   The author declares that they used generative AI to come up with new ideas for their review.

   Read the original source
3. 

   Version published to 10.1101/2025.10.30.25339095 on medRxiv

   Oct 31, 2025