Write a comment

Short summary of the research and contribution to the field

This preprint evaluates the concordance between droplet digital PCR (ddPCR) and next-generation sequencing (NGS) for detecting three clinically important EGFR mutations in non-small cell lung cancer: L858R, exon 19 deletions, and T790M. The study addresses a practical clinical problem: EGFR testing often uses different specimen types, including plasma cfDNA, FFPE tissue, fresh tumor tissue, pleural effusion DNA, and pre-capture NGS libraries, but the concordance of these materials and methods is not always clear. The authors enrolled 789 NSCLC patients, with a final cohort of 952 cases, and evaluated 936 samples by both NGS and ddPCR after 17 ddPCR failures were excluded.

The main findings show high agreement between NGS and ddPCR, with an overall concordance of 98.72%, subtype-specific positive percent agreement of 98.93% for L858R, 99.23% for Ex19del, and 97.14% for T790M, and strong VAF correlation between methods. The study also reports strong concordance between original DNA samples and paired pre-capture NGS libraries, suggesting that residual pre-capture libraries may be useful for additional testing when source material is limited.

Overall, this work moves the field forward by providing a large-scale, multi-specimen comparison of ddPCR and NGS for EGFR mutation detection and by exploring a practical sample-sparing strategy using pre-capture libraries. This is clinically meaningful for NSCLC molecular diagnostics, where limited tissue, low-abundance cfDNA, repeat testing, and treatment-resistance monitoring are common challenges.

Positive feedback / strengths

1. Clinically relevant and practical question. The study focuses on EGFR L858R, Ex19del, and T790M, which are highly relevant to treatment selection and resistance monitoring in NSCLC. The introduction clearly explains that L858R and Ex19del are primary targets for first-line TKIs, while T790M determines eligibility for third-generation TKI therapy.

2. Large and diverse sample set. The study includes multiple specimen types encountered in clinical practice, including cfDNA, FFPE DNA, fresh tumor tissue DNA, pleural effusion DNA, and corresponding pre-capture NGS libraries. The flowchart on page 11 is especially useful because it shows sample screening, ddPCR-tested cases, failed ddPCR cases, and the distribution of plasma, FFPE, fresh tissue, and pleural effusion specimens.

3. Good use of agreement statistics. The authors do not rely only on Pearson correlation. They also include Bland–Altman analyses, mean bias, limits of agreement, and VAF-stratified comparisons. This is important because correlation alone does not prove quantitative equivalence.

4. Mutation subtype-specific analysis is valuable. The stratified analysis shows that L858R and T790M had stronger agreement than Ex19del, while Ex19del showed higher bias and wider limits of agreement. This is an important observation because deletion variants can behave differently from point mutations in both NGS and PCR-based assays.

5. The pre-capture library analysis is a useful clinical contribution. The finding that pre-capture NGS libraries preserve VAF information with strong correlation to original source material could be valuable for laboratories handling limited FFPE or cfDNA specimens. This is one of the most practically useful parts of the manuscript.

Major issues

1. The study lacks an NGS-negative / healthy control cohort, limiting specificity assessment

The study mainly evaluates EGFR-positive cases, which supports positive concordance but does not fully assess specificity, false-positive rates, or negative percent agreement. The authors acknowledge this limitation and state that the lack of NGS-negative and healthy control samples prevents assessment of specificity and NPA.

Suggested improvement: The authors should include, or clearly plan as a follow-up, a negative control cohort consisting of:

• NGS-negative NSCLC samples

• healthy donor plasma samples

• EGFR wild-type FFPE samples

• pleural effusion samples without known EGFR mutation

• low-input / degraded DNA samples

This would allow calculation of specificity, negative percent agreement, false-positive rate, and limit of blank, which are necessary for a complete clinical validation.

2. ddPCR LOD of 0.01% needs deeper validation and statistical framing

The authors report that the ddPCR assays achieved a 0.01% LOD with 100 ng input DNA and that all 10 replicates of the 0.01% reference standards were detected. This is promising, but the data show high coefficients of variation at very low VAF levels, especially for L858R and Ex19del at 0.01%.

Suggested improvement: The manuscript should clarify:

• whether LOD was defined as 95% detection probability or 100% detection in 10 replicates

• limit of blank and false-positive droplet threshold

• total DNA copy number assumptions

• number of mutant copies expected at 0.01% with 100 ng DNA

• whether LOD was independently validated across specimen types, not only reference standards

• whether separate LODs should be reported for L858R, T790M, and Ex19del

Given the high CV at 0.01%, the authors should be cautious about presenting 0.01% as a routine clinical performance level unless robust false-positive and reproducibility data are shown.

3. The Ex19del ddPCR assay design requires more explanation

EGFR exon 19 deletions are heterogeneous. The manuscript states that Ex19del was detected using specific primers/probes designed around deletion regions, with the mutant probe fully or partially overlapping known deletion regions. However, the study does not clearly show whether all clinically relevant Ex19del subtypes are equally detected.

Suggested improvement: The authors should add:

• a list of Ex19del subtypes detected by NGS

• which Ex19del subtypes are covered by the ddPCR assay

• whether rare or complex Ex19del variants were excluded

• whether discordant Ex19del cases were due to deletion subtype coverage

• representative droplet plots for common and uncommon Ex19del variants

This is important because Ex19del showed the largest measurement bias and wider limits of agreement compared with L858R and T790M.

4. Discordant cases should be analyzed in a dedicated table

The discussion states that 12 discordant samples were identified, including 4 L858R, 3 Ex19del, and 5 T790M cases. The authors also note that 7 of 12 discordant cases were pre-PCR processed samples, suggesting pre-amplification artifacts may be a risk factor.

Suggested improvement: The authors should include a detailed discordance table with:

• mutation type

• specimen type

• original NGS VAF

• ddPCR VAF

• pre-capture library status

• DNA input amount

• droplet count

• NGS unique read depth

• strand support

• freeze-thaw history if available

• whether mutation subtype was covered by ddPCR

• likely cause of discordance

• any orthogonal resolution

This would make the findings much more actionable for clinical laboratories.

5. Pre-capture NGS library substitution is promising but should be framed carefully

The study reports strong concordance between native samples and pre-capture libraries, including r = 0.993 for cfDNA vs cfDNA-prePCR, r = 0.998 for tumor tissue DNA vs prePCR for L858R, and r = 0.991 for tumor tissue DNA vs prePCR for Ex19del. This is valuable, but pre-capture library use can be affected by PCR cycles, library complexity, degradation, storage duration, freeze-thaw cycles, amplification bias, and low-input artifacts.

Suggested improvement: The authors should clarify:

• number of PCR cycles used before capture for each library type

• minimum library concentration and quality thresholds

• storage time and storage conditions of pre-capture libraries

• whether multiple freeze-thaw cycles occurred

• whether library age affected concordance

• whether pre-capture libraries are recommended for routine testing or only rescue testing

• whether performance remains acceptable at low VAF

This would help clinical labs decide when pre-capture libraries are appropriate substitutes.

6. VAF correlation should not be overinterpreted as full interchangeability

The overall VAF correlation between ddPCR and NGS is strong, but Bland–Altman results show systematic bias, with ddPCR tending to report higher VAF values than NGS. The authors report an overall mean bias of 3.18%, with limits of agreement from -7.29% to 13.65%.

Suggested improvement: The authors should avoid implying complete quantitative interchangeability. Instead, they should frame the methods as highly concordant but not perfectly interchangeable for longitudinal monitoring unless the same platform is used consistently. This is especially important for high-VAF samples and Ex19del, where bias and variability were greater.

7. The NGS method needs clearer reporting of assay validation metrics

The NGS method is described as a custom 1326-gene panel with EGFR exons 18–21 coverage, hybridization capture, NovaSeq sequencing, and minimum depths of 1,000× for tissue and 2,000× for liquid biopsy samples. The bioinformatics section includes BWA, Gencore, Samtools, VarScan2, background polishing for cfDNA, paired normal filtering, and manual visual inspection.

Suggested improvement: The authors should provide more NGS analytical validation details:

• EGFR exon-level coverage uniformity

• unique molecular depth after deduplication for each sample type

• NGS LOD for each EGFR mutation type

• NGS false-positive controls

• whether UMIs were used

• background error rates for EGFR hotspot regions

• minimum VAF thresholds for reporting

• how paired normal data were used in routine cases

This would make the NGS comparator better defined.

Minor issues

1. Clarify cohort numbers. The abstract mentions approximately 1,000 EGFR-positive samples, methods mention 789 patients and 952 final cases, and results describe 936 samples evaluated by both NGS and ddPCR. The flowchart helps, but the manuscript should clearly distinguish patients, cases, samples, successful ddPCR tests, and paired comparisons.

2. Improve terminology around “prePCR.” The term “prePCR” could be confused with pre-amplification or pre-capture library. The authors should define it consistently as pre-capture NGS library and use one term throughout.

3. Correct formatting and typographical issues. There are several spacing, capitalization, and punctuation issues, such as “Backgrounds,” inconsistent “ttdna/ttDNA,” and “pedna/peDNA.” Standardizing these would improve readability.

4. Add representative ddPCR plots. Representative 2D amplitude plots for L858R, T790M, and Ex19del at low, intermediate, and high VAF would help readers assess assay performance visually.

5. Report ddPCR droplet QC metrics. The methods mention approximately 50,000–60,000 droplets per sample, but the manuscript should report accepted droplet thresholds, failed droplet criteria, and whether droplet counts differed by sample type.

6. Clarify sample failure causes. The flowchart shows 17 ddPCR method failures. The authors should describe why those samples failed and whether failures were associated with a specific sample type or processing condition.

7. Avoid overuse of “gold standard.” The discussion describes NGS as a gold-standard detection platform. Because the study uses NGS as the reference standard but lacks negative controls and full specificity assessment, the wording should be softened to “reference method” or “comprehensive profiling platform.”

8. Clarify clinical interpretation of VAF differences. The authors state that VAF discrepancies are clinically acceptable, but this may depend on clinical context. A clearer explanation is needed for monitoring, treatment resistance detection, and low-VAF ctDNA settings.

9. Add prospective validation as a future direction. The authors acknowledge the retrospective design. A prospective validation study would strengthen clinical utility, especially for monitoring treatment response and resistance emergence.

10. Consider adding a practical workflow recommendation. A figure or table showing when to use NGS, when to use ddPCR, and when pre-capture libraries are acceptable would increase the clinical utility of the manuscript.

Overall assessment

This is a clinically useful and well-motivated study that addresses an important practical question in NSCLC molecular diagnostics: how reliably NGS and ddPCR agree across EGFR mutation types, VAF ranges, biospecimen types, and residual pre-capture libraries. The large sample size, multi-specimen design, mutation subtype analysis, Bland–Altman evaluation, and pre-capture library comparison are major strengths.

The most important areas for improvement are the lack of negative controls, deeper LOD validation, clearer Ex19del assay coverage, dedicated analysis of discordant cases, and more careful framing of pre-capture libraries as substitute material. The data strongly support high positive concordance between NGS and ddPCR, but specificity, negative percent agreement, and routine low-VAF performance need further validation.

With these additions, the manuscript would provide a stronger and more actionable framework for integrated EGFR testing using NGS, ddPCR, and residual pre-capture library material in NSCLC molecular diagnostics.

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.