Casava FASTQ from SRA

[JMFA]

Dear all,

I am currently trying to perform quality checks on a set of FASTQ files (downloaded from SRA) that (according to the authors) were generated using CASAVA 1.8.1 pipeline.

Since all files belong to a single sample, I am trying to use "--casava" option in FASTQC but I keep getting the "SRR*.fastq.gz didn't look like part of a CASAVA group" error.

To my understanding, FASTQC requires that all CASAVA generated files should be named <sample name>_<barcode sequence>_L<lane>_R<read number>.<0-padded 3-digitset number>.fastq.gz

However, the fastq files from SRA have a very different name: "SRR*_1/_2.fastq.gz"

Is there a way to change the name of these files so that FASTQC recognises them as a single sample or should I just analyse them independently?

Thank you very much in advance,
JMFA

[GenoMax]

Have you tried to run FastQC on the files without worrying about the casava option? What SRR # are you working with?

[JMFA]

Hi!
Thanks for the reply

Yes. Running w/out the "--casava" option was the first thing that I did. However, I got some weird Kmer profiles. FASTQC doesn't detect any adapter content / overrepresented sequences but it shows Kmer enrichments around the center and towards the end of the read...(I am attaching one example)



I have limited experience with NGS data so I have no idea whether this has anything to do with the way I am "using" the data. The pattern is actually similar for most fastq.gz files so perhaps this is a problem with the data itself.

[sarvidsson]

Not sure if FastQC includes all adapters - to me that looks like miRNA data before adapter clipping.

TCGTATGCCGTCTTC: http://www.biomedcentral.com/1471-2164/12/176

[JMFA]

This is actually 100bp PE DNAseq data.
I've clipped the adapter sequences and the Kmer profile looks much better (not perfect but better)!

However, I have an additional question.
While running cutadapt to remove the adapter sequences I am also setting the "-m" option (used to throw away processed reads shorter than N bases) to 100bp. However, I end up discarding approx. half of the reads this way.

Is there a problem (for instance, in the alignment step) to use reads with varying length?
Again, thank you very much for the input.