Again, from the authors' response:
measurements is exactly the same commercially available tissue cDNA as used for the RT-qPCR assays. If not, then the comparisons between array and PCR do not provide much information (Fig. 2B, and Fig. 2C).

Response: In our paper we compared the results which we obtained from real time PCR assays on a microfluidic platform to microarray data obtained from publicly accessible databases. We wanted to make three points with the data presented in Figure 2. First, that our approach enabled one to obtain high throughput gene expression data essentially identical in quality to conventional microliter RT-qPCR (r = 0.986, Figure 2D). Second that the reproducibility of our approach was very good (r >0.99) both across and between chips (Figures 2E and 2F respectively) in comparison to the reproducibility of similar tissues and assays on separate microarrays (which only showed a correlation coefficient value r = 0.933, Figure 2 A). Thirdly, even though we used different commercial preparations of cDNA from those used in the microarray experiments (although from the same tissue), on comparison, the variability we observed in a head-to-head comparison was only slightly worse than the variability from microarray to microarray (Figures 2B and 2C). We were not expressly comparing our data to that provided by the microarray approach, so we do not believe that the tissues examined have to be of the exact same commercial preparation to make the point that our platform provides more reliable data.
We therefore believe that the comparisons in question are both relevant and necessary for this study.

Comment: Is this normalization strategy (using multiple housekeeping genes) not possible for this application also?
Response: One could easily use more housekeeping genes for normalization purposes, and we have obtained similar results using a number of different housekeeping genes for this analysis. However, it is the convention in the real time PCR to normalize data using a single reference gene, using a calculation termed the 2-ΔΔCT method (see Livak KJ, Schmittgen TD <2001>Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 25: 402-408).


Comment:It would be of help to see more multiple determinations of the very same sample for the very same gene. Thus 2000 measurement sof the same would give a very exact determination of the variation.

Response:Figure S1 and Table S2 shows the results when multiple replicates of the same sample are analysed for the same sample. Each measurement shows the average and standard deviation from 288 replicates. For the highest concentration (Curve set 1, relative concentration 1 x 100), the mean Ct value was 7.9, and the standard deviation across the 288 Cts was 0.089. This is extremely high quality data. We routinely test our chips by using a similar concentration of sample across the entire chip (2304 replicates), and typically obtain a standard deviation of about 0.1 across 2304 replicates. We have included a description in the results section which highlights and stresses this point, so that it becomes a little more apparent to the reader (last paragraph in the section entitled Chip Design and Raw Data)

Comment:In the same way it would be helpful to see multiple linearity determinations using dilutions of -say 20 samples - analyzed in 50 assays of the same gene, for each dilution point.

Response:Again, this data is shown in Figure S1 and Table S2. A 10-fold serial dilution is shown across 288 replicates of the same gene. Again, we have reiterated this in the text, in the same paragraph mentioned above.

Comment:A determination of within run and between day variaiton and eventually between lot variation using conventional assay reagents would be on ineterst.

Response:This data is already present in the paper. Figure 2E shows a within run data set, and Figure 2F shows a between chip and between day data set. Additionally, Table S3 shows the data across 6 chips, which include two separate lots of chips. Conventional real time assay reagents were used in all cases.


Comment:Dilution curves and reproducibility should be determined on RNA extracted from "real" tissues and include conventional standardization molecules.
Response:The data shown in this paper includes analysis of cDNA extracted from 18 different “real” tissues obtained from OriGene Technologies (Rockville, MD). A complete list of the tissues was, and is still found in Table S1, Part II. It is also described in the Material and Methods section.

Comment:What is meant by digital PCR? This is not clear (end of section:" validation of gene expression data")
Response:We have added a description of digital PCR at the end of the section, along with appropriate references. Essentially, digital PCR involves the isolation and amplification of single DNA molecules, which is made possible by the use of the Fluidigm digital array chip.

Comment:Combined figures of the data described above would be welcome.
Response:We tried, but the ensuing figures lacked adequate resolution. We have restricted ourselves to the original figures.

Comment:It is difficult to see where the results described in Table S4 are described and discussed in the manuscript.

Response:We agree that it is difficult to determine where the results from Table S4 are discussed. Figure 3 is generated using the standard curves shown in Figure S2. The raw data for the standard curve in Figure S2 is presented in Table S4. We have added a line in the figure legend section (for Figure 3) which makes this clear.