Referee 1's Review:

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication, the manuscript has been revised in light of these comments and to address other editorial requirements.
**********

Report on « Identification of novel tumor markers in prostate cancer by unbiased methylation profiling » by Chung et al.

To isolate new hypermethylated CpG island to be used as biomarkers, Chung et al. used MCA (Methylated CpG island Amplification) associated with RDA (Representational Difference Analysis) on prostate cancer cell lines and they isolated 34 clones corresponding to hypermethylated CpG island promoters.

Then, to confirm the accuracy of MCA-RDA, the authors analysed DNA methylation of 17 genes in a collection of 21 cancer (prostate, colon, breast, leukemia) cell lines using COBRA and pyrosequencing. This analysis allowed the authors to identify a number of genes that were very frequently hypermethylated in cancer cell lines.

Finally, DNA methylation of seven to nine frequently hypermethylated genes were tested in three panels of prostate, colon, breast tumors and paired healthy mucosa. Each panel comprised 20 to 24 different tissues.

This is a potential interesting paper as 1) it describes new potential epigenetic biomarkers for cancer diagnosis and 2) it shows the efficiency of MCA-RDA to identify new hypermethylated genomic sites.

However, if the aim of the work is to identify new tumor markers, I regret that the authors did not calculate the sensitivity and the specificity of the markers they found. It would be interesting to see the performance of individual markers and the combination of several markers to detect cancer. With this respect, I think that the discussion does not fit very well the aim of the work. The authors discuss data from the litterature on the possible role of these genes in tumor trasformation, whereas their work does not really add any information on the gene function.

Beside this, I think the manuscript lacks of accuracy in several points and a number of issues must be amended or improved.

1. Some genes are considered hypermethylated, but nowhere in the texte I found a definition of hypermethylation.

2. Before using COBRA to quantify DNA methylation, it is advisable to check whether the relation between DNA methylation and the COBRA scores is linear, in order to determine the quantitative accuracy of the assay. This can be done by using defined mixtures of M SssI methylated genomic DNA and untreated DNA. In this work, Chung et al. tested the COBRA assays on DNA from normal lymphocytes and on 100% methylated DNA, however they did not verify whether the COBRA assay yelded reliable quantitative results across a wide range of DNA methylation levels.

3. The authors do not specify whether the COBRA experiments were done in replicates and, if this is the case, they did not mention how reproducibles were the quantitative analyses of individual genes.

4. It is mentioned that some of the genes were also analysed by pyrosequencing, but no results for pyrosequencing are provided. Were they always consistent with the COBRA results?

5. Fig. 1B shows the COBRA digestion pattern of 17 genes in 8 cancer cell lines, in 3 prostate cancers and in 3 colon cancer and healthy tissues. I think the gels are not really informative. Instead, it would be more useful to include, as a supplementary data, a table containing the methylation percentage of each gene in each cancer cell line and tissue.

6. Methylation and gene expression. At the end of the work, the authors analysed the expression of eight hypermethylated genes in cancer cell lines and showed that they were silenced. After 5-Aza-dC treatment, gene expression was reactivated. It is not clear why these eith genes and not others were selected.
Moreover, in the material and method section, the authors described also TSA treatments, but these experiments do not appear in the results section.

7. In a similar way, in the material and method section (statistical analysis), the authors state that correlations were calculated using the Spearman non-parametric test, but in the result section I did not find any correlation study. I have the feeling that the authors did a copy paste of material and method from an other manuscript without verifing whether the content really fitted the present manuscript.

8. Two sentences at the end of the discussion do not make sense.

The manuscript should be accepted after major improvement of data presentation.