Picard: EstimateLibraryComplexity

[Deb]

Hello,

I've been using EstimateLibraryComplexity in Picard and have read the documentation but want to make sure I am interpreting the results accurately. We recently ran an estimate on our genomic reads and our library size was reported as 660,102,280 while the number of read pairs was 170,440,606 (our genome size is ~1GB). From my interpretation, this means that we have 660,102,280 unique fragments within our library of which 170,440,606 have been sequenced. Therefore, we have sequenced ~1/3 of the library and in theory some additional HiSeq runs would be needed to increase complexity?

Any insight would be helpful.

Thank you!

[timydaley]

If you want to get more distinct genomic molecules from the library, then sequencing more will give you more. But there is a law of diminishing returns since you have already sequenced the most common molecules in the library, so that most of the additional sequencing is duplicates. I am not sure how Picard estimates the library size, looking at the java code now, but it doesn't answer the question of how much more you stuff you will get from sequencing more.

Our lab does a lot of bisulfite sequencing analysis. To save money on sequencing poor libraries, we developed a tool to answer this question, "How many more reads are you going to get?" The website is here. Since we were looking at mapped reads, this is what the input is. We're looking for users, so maybe this can be of help to your problem.