Whole-Transcriptome Profiling on Small FFPE Samples: Which Sequencing Kit Should Be Used?

Round 1

Reviewer 1 Report

The report by Hilmi et al. compares the efficacy of several kits for RNA-seq analysis of FFPE samples. The report is overall well-written and has the potential to help others in the field, in the choice of methods. There are minor issues with presentation of methods and results that need to be addressed prior to publication.

1) It is not always clear what the p-values are referring to, especially when a correlation of different methods is presented - the type of test should be specified in every case.

2) The p-values are also stated in the text without connection to the measurements being estimated. On the other hand, the measurements are presented in figures without statistics (e.g. Fig.2). I would suggest always presenting the measurements and their statistical testing together, be it in the text or the figure.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Review of:

Whole-transcriptome profiling on small FFPE samples: which sequencing kit should be used?

Comments to the Authors

 General remarks:

The authors compare six different RNA sequencing library kits for in FFPE archived samples. The RNAseq data was then compared with data from Nanostring (as gold standard for quantification) and polyA-enriched RNA (as reference for RT-based RNAseq) from a paired flash-frozen sample. 20 breast cancer tumor tissues were used at various amounts for each kit.

The authors identify “Smarter” as the best choice and “Lexogen” as the most economical one.

 Specific remarks:

Adding a table with all the kits including the amounts of RNA used, number of samples (this was not clear to me), different methodologies and costs per sample since quality price ratio is mentioned would be extremely helpful.

What were the RIN numbers for each of the samples?

 Line79:

Genes with at least one count are shown. The quantification should include a cutoff of at least 10 cpm.

Figure 1:

Number of expressed genes seem extremely high with >35,000 genes. Gencode GRCh38.p13.40 contains 19,988 coding genes. From my experience no more than that number of genes is detectable using RNAseq. Adding a cutoff of 10cpm may reduce these numbers but make them more reliable (especially the correlation plots in fig. 4a). In the legend please add the parameters for the Box & Whiskers graph (also in figures 2,3,4). Add significance (p-value for each comparison) directly to the graph or in the legend.

Figure 2:

Y-axis label for 2b is missing.

Line 126:

What are the cutoffs for each tercile in trueseq?

Line 172 and 175:

There is no figure 5. Correct to Figure 4a and 4b, resp..

Figure4:

Please increase the fonts (graph axis labels and R and p-value are too small.

Where is the probability map in those images?

Line 194,195:

Write: techniques

Author Response

Please see the attachment.

Author Response File: Author Response.pdf