QC for Whole Exome Sequence Data

Question

[haiying.kong]

For quality control on whole exome sequence data, I first ran FastQC. Then I checked the proportion of the bases that are having coverage less than a threshold (10 or 15, ...). Ideally this proportion should be low.

Should I do any thing more?

I do not feel the uniformity of the coverage can be any reason for us to decide if the sample quality is good or bad. Usually total number of read is fixed when we sequence, so if it is not uniform, there is higher chance that the proportion of the bases that are having low coverage is large. But this is only a chance. Some times some how some samples have more reads. So the uniformity should not be a criteria to decide if a sample should be in or out.

[trausch]

For exome data, we usually check the on-target rate and the fraction of targets above a certain coverage level such as 30x. The error, mapping and duplicate rate is also important. We have implemented these metrics in a QC tool called Alfred and an example plot for whole-exome data is here (sorry for the self advertisement, I believe picard HsMetrics is reporting something similar).

Answer 1

[Brian Bushnell]

Exon capture always has low uniformity. For exon-capture QC, some of the criteria that are important are:

1) % of reads that map to the reference

2) % of reads that map to the baits

3) Error rate

4) Duplicate rate

[haiying.kong]

Thank you very much for your reply.

So how uniform the coverage is should not be a standard to decide if the data for the sample is usable or not, right? I think as long as the proportion of low coverage base is under control, how uniform the coverage is should not be a problem.

[Brian Bushnell]

To decide if you have enough coverage, yes, it's important to see what fraction of your target bases have coverage above the threshold that you are using for calling variants. I don't really consider that a "quality" metric, though.

[dyollluap]

I use picard tools to get some of these metrics as a general guide for comparison. In combination with fastqc it's a good QC overview.

[Shicheng Guo]

HWE cut-off ?