Umesh Reclusa hominid indica
For those of you who have downloaded their BAM Files, have you ever tried using FastQC to check the Quality of your BAM?

I checked mine, and although my BAM's per Base Sequence Quality is very high, I have some statistics which have been labelled as "concerning' in terms of Quality Analysis. 

Most of the flags can be attributed to possible variations in individual sequencing processes adopted by different companies, although I would love to hear more details from a more qualified person. I am most concerned about the "Per Base Sequence Content" chart, which shows the percentages of bases to be same across a large part, rather than the expected uneven distribution as seen in the reference chart. The Long Reads in the "Sequence Length Distribution" Chart are missing, but that was expected, as I had not paid extra for it (plus maybe Dante has an issue deciphering Long Reads).  Have you tried running FastQA on your BAM? Plz share your observations.

Available for all platforms here: https://www.bioinformatics.babraham.ac.uk/projects/download.html

Fast QC0.jpg  Fast QC0 index.jpg  Fast QC1.jpg  Fast QC1 index.jpg  Fast QC2.jpg  Fast QC2 index.jpg  Fast QC4.jpg  Fast QC4 index.jpg  Fast QC3.jpg  Fast QC3 index.jpg  Fast QC5.jpg  Fast QC5 index.jpg  Fast QC6.jpg  Fast QC6 index.jpg  Fast QC7.jpg  Fast QC7 index.jpg  Fast QC8.jpg  Fast QC8 index.jpg 
Quote 0 0
Umesh Reclusa hominid indica
Sorry the Sequence Duplication Levels Report got truncated.
Fast QC.jpg
As also the Per Base Sequence Content Chart
Fast QC.jpg 
Quote 0 0
Erik Chapman
So much depends on what people mean by quality and the assumptions that go into it.   One of the big problems in gene sequencing is there are in fact long sequences that are relatively monotonous rather than random and that can make it more difficult to make a call.   Some services will avoid making a call to lower risk of a miss.   But is that better quality?  I'd think it would be better to have a system to flag or asterisk the less certain areas rather than omitting them, to my mind omitting information over a slight increase in error risk is actually a misrepresentation and distortion of the results rather than higher quality.    But there should be aggressive flagging of the at risk areas- in my view 99% certainty is more than enough for a call, but over tens or hundreds of millions of base pairs with the same issue that's going to result in a lot of false positives.  
Quote 0 0