Copy number analysis by Golden Helix

Question

[ytian]

Dear everyone I recently focused on the copy number analysis using SNP numeric data produced by illumina 2.5M platform. I have several questions for statistical experts who have experience in calculating the copy number by the SVS function CNAM which mainly creates a segment means to represent genome copy number. These questions may influence the hypothesis of my project

1. Can the SVS implanted function Segmentation, which only consider log2Ratio, be used to detect copy number variate in normal tissue, in our case, in the normal prostate tissue of prostate cancer patients?
2. If the answers above is yes, after calculation, we found little variation in our data. Do we have to turn to another method incorporating the B allele frequency to calculate the copy number variant?
3. Can the SVS method Segmentation outputs such as segment means used to make association tests because no clear threshold can be inferred from the histogram of segment means and so no clear definition of loss and gain can be decided?

Answer 1

[Gabe Rudy]

Hi there,

I realize this question is old, but I found it when search and thought I would weigh in (I have worked on this algorithm at Golden Helix).

1) The Log2Ratio based segmentation can detected germline CNVs, but it really matters how you normalize the data from your microarrays. We have a illumina based importer that normalizes samples, but not in a tumor-normal aware mode which may be helpful in this context.

2) Possibly, but more likely the signal is not there due to the normalization used to construct the Log2Ratios

3) Yes, it is pretty common for the segmentation output to simply bucket all Log2Ratios into segments detected in any sample so that you have a matrix of samples X detected_segments that can be used in association tests.

Also, feel free to email support at GoldenHelix as we can answer these questions promptly through that channel!

Thanks, Gabe