Analytical Validation of NETest2.0^®, a Novel Multigene Blood-Based Molecular Assay for Neuroendocrine Tumors

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

I enjoyed reading the article. The idea is clear and the goals are well outlined.

I have some questions – can the test be easily implemented globally for early detection of neuroendocrine tumours? Does the analysis depend on the individual laboratory to be carried out, or is it performed using a software tool?

Have you compared your proprietary RNA-stabilisation tubes with other RNA-stabilisation tubes?

What is the advantage of the RNA test compared to DNA?

How did you select the housekeeping genes for normalisation? Did you test a larger cohort to choose from?

What about the epigenetic aspect, particularly the methylation profile? Did you consider comparing the RNA molecular signature with the methylation profile?

Could the specificity of the test be improved? How did you select the 51 genes included in  the test?

Did you observe any differences between the various levels of malignancy in the NET samples tested?

How do you explain the positive result in some epithelial tumours from the non-NET group? Can it be minimized, perhaps with using a broader gene set?

Author Response

Response to Reviewer 1

We thank Reviewer 1 for the positive assessment of the manuscript and for the insightful comments. We have addressed each point below and revised the manuscript accordingly.

1. Global implementation and reproducibility

The study is interesting. Could the test be implemented globally? Is the assay dependent on laboratory-specific analysis or software interpretation?

Response #1:

We appreciate the reviewer’s interest in the translational applicability of NETest2.0®. The assay was specifically designed for standardized and scalable implementation across molecular diagnostic laboratories. Peripheral blood samples are collected in validated RNA stabilization tubes, followed by standardized RNA isolation, reverse transcription, and qPCR using pre-spotted TaqMan plates.

The resulting Ct values from 51 NET-related target genes and 4 housekeeping genes are processed using a cloud-based locked proprietary machine-learning algorithm, running as an API, that generates a continuous NETest2.0® score from 0–100, with a validated positive cutoff of ≥50. Because interpretation is algorithm-based rather than operator-dependent, inter-operator variability is minimized. The analytical validation data presented in this study demonstrated strong intra-assay and inter-assay reproducibility with low coefficients of variation, supporting robust implementation potential across laboratories.

We have clarified this workflow in the revised Methods and Discussion sections. Furthermore, we have expanded the discussion has been updated to identify next steps in validation including inter-laboratory evaluations. The assay workflow (protocol, kit, software) is designed to support scalable implementation across molecular diagnostic laboratories.

2. RNA stabilization tubes

Have the proprietary RNA-stabilization tubes been compared with commercially available RNA stabilization tubes?

Response #2:

This study focused specifically on analytical validation of the NETest2.0® assay workflow using the validated blood collection and stabilization system incorporated into the assay platform. Direct comparative studies with alternative commercial RNA stabilization tubes are currently ongoing e.g., vs. PAX RNA collection tubes. A preliminary evaluation of data from the Wren vs. Pax study identifies significantly better results (RNA quality, quantity, stability) for the Wren tubes. This separate study will be finalized shortly and then submitted for peer-review.

We have acknowledged this in the Discussion and clarified that the reported performance characteristics apply specifically to the validated NETest2.0® collection and processing workflow.

3. RNA versus DNA biomarkers

What is the advantage of RNA testing compared with DNA-based approaches?

Response #3:

We thank the reviewer for raising this important conceptual point. DNA-based assays primarily identify static genomic alterations, whereas RNA-expression profiling may better reflect dynamic biologic pathway activity. NETest2.0® measures transcriptional activity from multiple NET-associated pathways and therefore may provide real-time biologic insight into disease activity, treatment response, and progression risk.

This biologic responsiveness is particularly relevant for neuroendocrine neoplasms, which frequently demonstrate heterogeneous clinical behavior that may not be fully explained by genomic alterations alone.

4. Selection of housekeeping genes

How were the housekeeping genes selected?

Response #4:

The four housekeeping genes included in NETest2.0® (ALG9, ATG4B, RHOA, and TXNIP) were selected during prior assay development studies based on their stable expression across NET and control samples and their suitability for normalization in qPCR-based transcriptomic analyses (see Reference #16. Physiol Genomics. 2007;30(3):363-70).

In the current validation study, housekeeping gene Ct variability remained low, with coefficients of variation approximately 2.1–2.3%, supporting the robustness of normalization and assay reproducibility.

5. Epigenetic or methylation profiling

Were epigenetic or methylation-based approaches considered?

Response #5:

The current study was designed as an analytical validation of a multigene RNA-expression assay rather than a methylation-based assay. We agree that epigenetic profiling represents an important complementary area of investigation in neuroendocrine tumor biology. Future studies evaluating integration of transcriptomic and epigenetic biomarkers may further improve disease characterization and monitoring strategies.

6. General comments

Reviewer comment:

The manuscript is generally well written and clinically relevant.

Response #6:

We thank the reviewer for the positive evaluation and thoughtful suggestions, which have helped strengthen the clarity and translational relevance of the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

In this manuscript, the authors present an analytical validation of NETest 2.0 for detecting neuroendocrine tumours. The study is generally well designed, follows relevant guidelines, and includes a reasonably sized patient cohort, supporting its potential clinical utility. However, prior to acceptance, the overall quality and presentation of the manuscript require improvement. The following issues should be addressed:

Major comments

1. Tumour stage specification
   The manuscript should clearly define the stage of neuroendocrine tumours included in the cohort. Tumour stage is directly associated with circulating tumour-derived RNA levels in whole blood. It is therefore recommended to present Ct values and NETest 2.0 results stratified by tumour stage. Evaluating assay performance across stages is critical, particularly to assess its ability to detect early-stage disease.
2. Timing of imaging and sample collection
   Please provide detailed information regarding the timing of blood sampling relative to imaging procedures. The interval between these two events may influence diagnostic performance metrics such as positive predictive value (PPV) and negative predictive value (NPV), and should be discussed accordingly.
3. Table 1 and 2 formatting
   Table 1 should be reformatted to improve scientific presentation. Specifically, remove the colons in the second column and ensure consistency with standard reporting formats. Table 2 first column has problem to display information.
4. Figure formatting consistency
   Ensure consistent font size and formatting across all figures and sub-figures.
5. Figure quality
   The image quality of Figure 7 and Figure 9A should be improved to meet publication standards.
6. Please explain Figure 3 statistics what statistics are used in each subfigure? Is this the average of 55-gene Ct value? Please check with bio-statistician.

Minor comments

1. Remove citations from the Abstract.
2. In Table 3, replace the use of "*","**" and "***", as this is potentially confusing. Use clearer and more conventional notation, and ensure that symbols used for treated groups are distinct.
3. Provide more detailed information regarding the composition of the 55-gene panel.
4. Define all abbreviations at first use, including GI (Line 114), GIST, and RCC (Line 116). Please review the manuscript to ensure all acronyms are appropriately introduced.
5. Ensure consistency in formatting throughout the manuscript, including:
• Figure font styles (e.g., consistent use of bold formatting)
• Representation of sample size (use consistent notation for “n”)
• Units (replace “ul” with “µL”)

Author Response

Response to Reviewer 2

We thank Reviewer 2 for the careful evaluation and constructive recommendations. All comments have been addressed as outlined below.

Major Comments

1. Tumor stage specification

The manuscript should clearly define the stage of neuroendocrine tumours included in the cohort. Tumour stage is directly associated with circulating tumour-derived RNA levels in whole blood. It is therefore recommended to present Ct values and NETest 2.0 results stratified by tumour stage. Evaluating assay performance across stages is critical, particularly to assess its ability to detect early-stage disease.

Response #1:

We agree and have expanded the manuscript to clarify tumor staging characteristics within the NET cohort (see Table 2). The cohort included localized (Stage I–III) and metastatic (Stage IV) disease confirmed by imaging and histopathology. We additionally evaluated NETest2.0® scores across stages and observed detectable positive scores across all stages, including localized disease. AUCs were not different for Stage I-III vs. Stage (IV); 0.931 vs. 0.936 (see Figure 5D-E). This finding supports the assay’s ability to detect low-volume disease. Because histological grade may also impact scores, we have included an evaluation of this parameter too. The assay was equivalent in all grades, G1, G2 and G3 (see Figure 5A-C) confirming that it was an effective tool across all histological types. A supplementary analysis stratifying NETest2.0® scores by grade and stage has now been added to the Results section and incorporated into revised figures/tables (se Figure 5 and Table 2). The discussion has also been updated to reflect these results.

2. Timing of imaging and sample collection

Please provide detailed information regarding the timing of blood sampling relative to imaging procedures.

Response #2:

We thank the reviewer for highlighting this important issue. Blood samples used for diagnostic accuracy assessments were collected contemporaneously with routine clinical imaging evaluations, generally within a clinically accepted interval of ±30 days. We have clarified this in the Methods section. Because imaging represented the reference standard for disease status classification, this timing minimized potential discordance due to interval disease progression or treatment effects. We additionally note this timing consideration in the Discussion as a potential contributor to real-world variability in PPV and NPV estimates.

3. Table 1 and 2 formatting

Table 1 should be reformatted. Table 2 first column has display problems.

Response #3:

We thank the reviewer for identifying these formatting issues. Table 1 has been reformatted for consistency with journal style, including removal of unnecessary colons and alignment improvements. Table 3 (this is new, it was originally Table 2) formatting issues in the first column have also been corrected. We have ensured that all other tables (Table 2, Table 4 and Table 5) are appropriately formatted for the journal.

4. Figure formatting consistency

Ensure consistent font size and formatting across all figures and sub-figures.

Response #4:

All figures have been revised to ensure consistent font type, font size, labeling style, and formatting across panels and sub-figures (as far as applicable).

5. Figure quality

The image quality of Figure 7 and Figure 9A should be improved.

Response #5:

Higher-resolution versions of these two figures (old Figure 7 – new Figure 11 and old Figure 9A – new Figure 13A) have been generated and replaced in the revised manuscript to meet publication-quality standards. We have also taken the opportunity to update new Figures 8 and 9 (old Figures 6 and 7).

6. Please explain Figure 3 statistics

What statistics are used in each subfigure? Is this the average of 55-gene Ct value?

Response #6:

We appreciate this important point and have clarified the statistical methodology in the Figure (old Figure 3, new Figure 7) legend and Methods section. The Ct values shown represent averaged Ct values for each of the individual 55 genes across the experiments. Statistical analyses included Mann–Whitney testing for operator comparisons and Pearson correlation analyses for reproducibility assessments. Coefficients of variation were calculated according to CLSI EP05 guidance. We additionally reviewed the statistical descriptions with our biostatistical team to ensure accuracy and clarity.

Minor Comments

1. Remove citations from the Abstract

Response 1: All citations have been removed from the Abstract.

2. Table 3 notation

Replace “”, “”, and “”.

Response 2: We agree and have replaced the notation with clearer superscript lettering and explanatory footnotes to avoid confusion (see new Table 4, old Table 3).

3. More detail regarding the 55-gene panel

Response 3: Additional detail regarding the composition and biological rationale of the 55-gene panel has been added to the Methods section. We clarified that the assay consists of 51 NET-associated target genes and 4 housekeeping genes selected through prior optimization studies and algorithmic training procedures.

4. Define abbreviations at first use

Response 4: All abbreviations including GI, GIST, RCC, and others have now been defined at first appearance and reviewed throughout the manuscript for consistency. The abbreviations list has been updated as well.

5. Formatting consistency

Response 5: The manuscript has been carefully reviewed and revised to ensure consistency in:

• figure formatting,
• bold font usage,
• notation for sample size (“n”),
• and units (all “ul” changed to “µL”).

Reviewer 3 Report

Comments and Suggestions for Authors

This is an interesting paper regarding clinical validation of the test called NETest2.0 - these procedures were conducted with reliability and meet the CLSI criteria. I have few comments that need to be precised prior to publication:

1. However, the results seem impressive, the authors could clearly address the 19.7% of patients with false-negative results. This is a major point. Do these cases correspond to specific NET subtypes?

2. It may be worth considering adding by the authors (if possible) general information about the type of machine learning model used, which would improve methodological transparency

3. As the authors mentioned in the limitations section - it is a single center study. In addition to stating this in the study limitations, it would be interesting to readers if  authors could describe how they plan to further develop and continue validating their diagnostic test

4. An economic aspect should also be considered - It would be useful to briefly mention the cost stability of the procedure compared with other methods 

5. Please correct figures - It would be clearer if the authors placed the labels A, B, C, etc. at the top of the figures. 

Author Response

Response to Reviewer 3

We thank Reviewer 3 for the positive assessment and valuable suggestions.

1. False-negative results

The authors should clearly address the false-negative results.

Response #1:

We agree this is an important point. The overall sensitivity of NETest2.0® was 91.5%, indicating approximately 8.5% false-negative cases (in the 219 other cancers assessed) rather than 19.7%. Our evaluation versus a control cohort (apparently healthy, no evidence of cancer: n=186) identified a sensitivity of 92.4%. A review of the NET cases suggested that false-negative results were more commonly associated with lower-volume or minimal residual disease and may reflect reduced circulating tumor-derived transcript abundance. Importantly, the assay still demonstrated positive detection across all disease grades (see Figure 5-C) and stages overall (see Figure 5D-E), including early-stage disease. We have expanded the Discussion to address potential biological explanations for false-negative findings.

2. Machine-learning transparency

Please provide general information about the machine-learning model.

Response #2:

We appreciate this suggestion. Additional information has been added describing the NETest2.0® algorithmic framework. Specifically, normalized Ct values from the 51 target genes and 4 housekeeping genes are processed through a locked proprietary machine-learning model trained to classify NET-associated transcriptional signatures and generate a continuous 0–100 risk score. While the proprietary nature of the algorithm limits disclosure of specific architectural details, we have improved methodological transparency by clarifying the analytical workflow (Figure 1).

3. Single-center study / future validation

Please describe future validation plans.

Response #3:

We agree and have expanded the Discussion accordingly. Although this analytical validation was conducted within a single laboratory under controlled conditions, future studies are planned to include formal multi-site reproducibility assessments and broader external validation cohorts. These studies will evaluate inter-laboratory reproducibility, workflow standardization, and performance across geographically diverse clinical settings.

4. Economic considerations

Please mention cost stability compared with other methods.

Response #4:

We appreciate this suggestion and have added a brief discussion regarding economic considerations. Because NETest2.0® utilizes standardized qPCR technology and centralized algorithmic analysis, the assay workflow is compatible with existing molecular laboratory infrastructure. Compared with repeated imaging studies or more complex sequencing-based approaches, the assay may offer a cost-effective strategy for longitudinal monitoring, although formal health-economic analyses remain warranted and are planned for future study.

5. Figure labeling

Place labels A, B, C at the top of figures.

Response #5:

All figures have been revised to reposition panel labels consistently at the upper portion of each subfigure for improved readability and compliance with publication formatting standards.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for addressing the comments. I don't have any further comments.