Genomic amplification is not a frequent event in uveal melanomas.

We read with interest the article by Lake et al and were surprised by the interpretation of their results. Uveal melanomas (UMs) are rare tumors associated with negative prognosis. Risk of recurrence can be predicted by pathological and genetic findings like monosomy of chromosome 3 and gains of 8q. Nonetheless, the identification of new genetic markers is warranted to better stratify patients and unravel the pathogenesis of this tumor. Lake et al performed a copy number and genotyping analysis on formalin-fixed, paraffin-embedded (FFPE) samples of 58 cases with the genome-wide SNP arrays version 6.0 (Affymetrix, Santa Clara, CA). They reported an amplification of the CNKSR3 gene in 11 of 58 cases (19%) and described numerous other genes as amplified. These findings caught our attention because UMs are known for having simple genome profiles defined by very few copy number alterations (involving chromosomes 3, 6, and 8) and recurrent point mutations (GNA11, GNAQ, BAP1, and SF3B1 genes). Since 2005, several hundred cases of UM have been analyzed and actual recurrent amplifications (ie, related to focal amplicons) have not been previously described. To extend this literature series, we reviewed 300 comparative genomic hybridization profiles from our data obtained on NimbleGen 4x72K whole genome arrays (Roche NimbleGen, Reykjavik, Iceland) and did not detect any CNKSR3 amplification. As no threshold for genomic amplification was defined in the article by Lake et al, we hypothesized that amplification was defined with a low copy number threshold corresponding in fact to a chromosomal gain. To test this hypothesis, we retrieved the publicly available raw data used for this publication from Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo; accession number GSE37259). We reanalyzed the data using the standard manufacturer pipeline with Genotyping Console and Birdseed version 2 algorithm (Affymetrix) as recommended for SNP6.0 arrays rather than Birdseed version 0 as used by the authors. Some have recommended to modify the final in silico data analysis procedure to extract most of the biological information from the signal measured. Using the above mentioned rigorous procedure,