Hi,
Did anybody check the new v4 reagents quality scores against the actual error rates?
I read enthusiastic blogs e.g. http://genomics.ed.ac.uk/blog/hiseq-v4-here-and-it-delivers, but look at views of my data below.
The median! is at 38 which would translate to an error probability of 0.0001585 (0.01585%). Looking at the PhiX spike in we have however 0.1-0.4%. Quickly comparing with older runs shows that we have now 1/2 the error rate, but surely not 1/10th-1/40th. The cautious interpretation would be that illumina still has to improve the calibration of their v4 model.
fastqc:
in Basespace (with error rates between:
in Basespace:
It sure looks good. I have seen runs of comparable quality on a MiSeq but were the exception not the rule. In fact after alignment the matches seemed more accurate than even the FASTQ scores.