Technical Validation of a Fully Integrated NGS Platform in the Real-World Practice of Italian Referral Institutions

Round 1

Reviewer 1 Report

Dear authors

The paper is nicely written and the aim well outlined. The data presented it of great value for all facilities setting up this new platform.  It compares performance data of the newest ThermoFisher NGS Platform Genexus and its Oncomine Precision Assay to the self-designed NGS assay SiRe. The size of the data set is appropriate and the number of different mutations and fusions investigated represents the actual landscape of most detected alterations in tumors. This leads me already to the first question, which the authors may add in the discussion: since the OPA assay detects CNVs as well, were there really no CNVs detected in the whole data set? Furthermore, is the SiRe assay as well able to detect CNVs (I did not dig further in the specs of the SiRe assay).

There are several additional minor points, which should be described in more detail:

-        how was the DNA/RNA quantified: Qubit/NanoDrop/other?

-        What was the input concentration for OPA/Genexus in all cases: optimal for the assay according to Thermo or at the lower limit?

-        OPA mean depth in the table = raw read coverage or effective coverage or molecular coverage in table 3; or even base or amplicon coverage?

-        On target reads are ranging from 77.7%-93.7% in table 3 which it quite low. Any explanation for that?

-        Is the Uniformity in table 3 = base uniformity or amplicon uniformity?

-        In table 4 DNA19: why was mutation PIK3CA E545K in the SiRe assay missed / below LOD? Any ideas? Is it because of assay design? This should be discussed since the discrepancy of VAF is quite high (0.8 vs. 7.2%). Maybe add a third assay to verify the ‘correct’ VAF?

-        You mentioned in the line 176 ‘standardized clinical cut-off’. Which filter criteria were used in SiRe/S5 and OPA/Genexus. Examples in the Genexus setting would be: default filter criteria using ‘Oncomine Extended (5.16) Filter Chain’.

-        In table 7 RNA25 the MET Exon 14 skipping is missing in SiRe/S5. As far as I understood from the text, this was called.

-        Concerning RNA passing filter criteria in OPA/Genexus: were in all cases 7/7 expression controls reached?

Author Response

Comments to the Author

Reviewer#1

• The paper is nicely written and the aim well outlined. The data presented it of great value for all facilities setting up this new platform. It compares performance data of the newest ThermoFisher NGS Platform Genexus and its Oncomine Precision Assay to the self-designed NGS assay SiRe. The size of the data set is appropriate and the number of different mutations and fusions investigated represents the actual landscape of most detected alterations in tumors

           We really thank the Reviewer for the kind appreciation of our study.

• This leads me already to the first question, which the authors may add in the discussion: since the OPA assay detects CNVs as well, were there really no CNVs detected in the whole data set? Furthermore, is the SiRe assay as well able to detect CNVs (I did not dig further in the specs of the SiRe assay).

We really thank the Reviewer for these suggestions. SiRe panel was optimized for gene fusion rearrangements (with known and unknown partners) in clinically relevant genes for solid tumor patients. It was not designed for CNV analysis.

• how was the DNA/RNA quantified: Qubit/NanoDrop/other?

What was the input concentration for OPA/Genexus in all cases: optimal for the assay according to Thermo or at the lower limit (table 1 con DIN e conc)

We really thank the Reviewer for these suggestions. We have modified the latest version of the manuscript and Table 1 as recommended.

• OPA mean depth in the table = raw read coverage or effective coverage or molecular coverage in table 3; or even base or amplicon coverage?

We really thank the Reviewer for these suggestions. We have included in Table 3 the average number of reads of all targeted reference bases.

• On target reads are ranging from 77.7%-93.7% in table 3 which it quite low. Any explanation for that?

We really thank the Reviewer for these suggestions. In a single case DNA#1 a quite low on target reads were observed. Probably, it may depends from the low quality of nucleic acids in a batching analysis with high-quality samples.

• Is the Uniformity in table 3 = base uniformity or amplicon uniformity?
• We really thank the Reviewer for these suggestions. We have included in Table 3 amplicon uniformity.
• In table 4 DNA19: why was mutation PIK3CA E545K in the SiRe assay missed / below LOD? Any ideas? Is it because of assay design? This should be discussed since the discrepancy of VAF is quite high (0.8 vs. 7.2%). Maybe add a third assay to verify the ‘correct’ VAF?
• We really thank the Reviewer for these suggestions. This event may occur in residual scant sample where mutated allele may encounter decreasing VAF level. We have modified the latest versione of the manuscript, accordingly.
• You mentioned in the line 176 ‘standardized clinical cut-off’. Which filter criteria were used in SiRe/S5 and OPA/Genexus. Examples in the Genexus setting would be: default filter criteria using ‘Oncomine Extended (5.16) Filter Chain

We really thank the Reviewer for these suggestions. We have modified the latest version of the manuscript, accordingly.

• Concerning RNA passing filter criteria in OPA/Genexus: were in all cases 7/7 expression controls reached?

We really thank the Reviewer for these suggestions. We would confirm that all RNA samples successfully passed quality checks.

Author Response File: Author Response.pdf

Reviewer 2 Report

De Luca et al have reported a very interesting study on the implementation of large NGS panels in clinical practice. The work is very well done, and the results are truly fascinating. According to this reviewer, it would be interesting to emphasize a couple of concepts:

1. On one hand, the author could highlight the flexibility of GeneNexus in being able to analyze different panels on different lanes and the fact that it is possible to analyze even a few samples at a time by distributing multiple runs throughout the week. On the other hand, they could address the limit in optimizing an automated setup (both Chef Machine and GeneNexus), which requires 8 samples (or multiples) to optimize the costs of the run.

2. Is this automated approach using OPA/GeneNexus applicable to low-cellularity samples, such as cytological specimens?

3. It would also be interesting to mention the analysis time required for an operator to investigate a targeted panel versus a broader panel like OPA. How does the use of a broader panel impact the reading and interpretation of the results?

Author Response

De Luca et al have reported a very interesting study on the implementation of large NGS panels in clinical practice. The work is very well done, and the results are truly fascinating. According to this reviewer, it would be interesting to emphasize a couple of concepts.

           We really thank the Reviewer for the kind appreciation of our study.

On one hand, the author could highlight the flexibility of GeneNexus in being able to analyze different panels on different lanes and the fact that it is possible to analyze even a few samples at a time by distributing multiple runs throughout the week. On the other hand, they could address the limit in optimizing an automated setup (both Chef Machine and GeneNexus), which requires 8 samples (or multiples) to optimize the costs of the run

We really thank the Reviewer for these suggestions. We have modified discussion, accordingly.

Is this automated approach using OPA/GeneNexus applicable to low-cellularity samples, such as cytological specimens?

We really thank the Reviewer for these suggestions. In this study we have only selected tissue specimens in order to reach an adequate amount of starting amount of nucleic acids from residual archival sample. In our diagnostic experience routinely investigate cytological specimens following this workflow.

It would also be interesting to mention the analysis time required for an operator to investigate a targeted panel versus a broader panel like OPA. How does the use of a broader panel impact the reading and interpretation of the results?

We really thank the Reviewer for these suggestions. We have modified discussion, accordingly

Author Response File: Author Response.pdf