FastQC with the ENCODE RNASeq data

[per_ngs]

Hello,
I just downloaded the LHCN RNASeq data generated at Caltech. I merged the fastq files from the 3 runs to generate a single file and ran fastqc and I an a little bit confused about the output I have got. The per base quality graph in fastqc is showing quality score going upto 70 (attached) and the per sequence graph is showing peaks at approx 38 and 68 (also attached). According to the ENCODE documentation, the quality scores are phred 33, so how come the quality score graphs look like this?
Apologies if my question is silly and if i am not understanding the way fastqc works.

Thanks for help.
NGSnewbie

[GenoMax]

That does look odd. Can you tell us where exactly you downloaded this data from? I can have a look to see if I can reproduce this.

[per_ngs]

I got the 2 X 75 LHCN (cycling and 7 day diff) fastq files from the following link

http://genome.ucsc.edu/cgi-bin/hgFileUi?db=hg19&g=wgEncodeCaltechRnaSeq

Attached is the snapshot of the files that I got

[GenoMax]

I tried one of the files from the dataset and I am seeing this plot. Data appears to have been submitted on 05/05/2011. Based on that date it is most likely in illumina format.

[per_ngs]

I tried one of the files from the dataset and I am seeing this plot. Data appears to have been submitted on 05/05/2011. Based on that date it is most likely in illumina format.

Thanks for trying fastqc one of the datasets. Other than the fact that your plot looks much better, I notice that the fastqc run that i did has identified the illumina encoding as 1.9 whereas your run has identified it as Illumina 1.5. I ran fastqc on the same file that you did and got the same results! Attached is the plot
Could my earlier results have happened because I combined the data from three runs and performed fastqc on the combined dataset? Or did you specify any paramters during the fastqc run?



Also, would it be better to keep the runs separate and also do the alignment etc. accordingly?

Thanks a lot for your help.
NGSnewbie

[GenoMax]

The only explanation seems to be that something happened when you combined the files (did you just "cat" them together?).

You could keep the lanes separate and then combine the results later.

Could my earlier results have happened because I combined the data from three runs and performed fastqc on the combined dataset? Or did you specify any paramters during the fastqc run?


Also, would it be better to keep the runs separate and also do the alignment etc. accordingly?

Thanks a lot for your help.
NGSnewbie

[pbluescript]

Illumina 1.3+ uses Phred+64 while Illumina 1.9+ uses Phred+33.
You can't combine them without adjusting the quality scores to match. You will have to treat each version separately or convert the quality scores.

[per_ngs]

Thanks GenoMax and pbluescript. I will keep the data separate and process it that way.