quality scores in fastq files extracted from sea files

[efoss]

I extracted some fastq files from sra files. Here are some lines from one of them:

@SRR034473.2 X8097_104:6:1:881:909 length=39
TCAAAAAATGAAGAAGAAGAAAAAAATGAAAAGGGTGCA
+SRR034473.2 X8097_104:6:1:881:909 length=39
CC<7C<<CCC<<C<7C<?CC<<,;,7CC6:?0??:(??.
@SRR034473.3 X8097_104:6:1:900:876 length=39
TGAAGTTCTTGTGGTTCAACCAAGTGTATTGCCAGTACT
+SRR034473.3 X8097_104:6:1:900:876 length=39
C<?<CCCC77CCC?4?C<4C<<$C,C47?C<??7*<(44
@SRR034473.4 X8097_104:6:1:905:908 length=39
TTGATGTGACTTGAAGGCTTCATCTCCTTTTTAGTGATT
+SRR034473.4 X8097_104:6:1:905:908 length=39
CCCCCCCCACCCA??CAACAA-CCCA?ACCAA6<A7+??

In trying to figure out what quality scores are being used in a fastq file, I usually look for strings of Bs at the ends of reads (telling me that it is the old Illumina scoring system) or strings of #s at the end of reads (telling me that it is the normal Sanger scoring system). But in these NCBI-created fastq files, I don't see any #s or Bs. I have three questions:

1. Why are there no #s or Bs?
2. How can I figure out what scoring system was used here?
3. With what I consider a more normal fastq file (with lines like those pasted below), what is the best way to figure out what scoring system is being used?

@SEQ_ID
GATTTGGGGTTCAAAGCAGTATCGATCAAATAGTAAATCCATTTGTTCAACTCACAGTTT
+
!''*((((***+))%%%++)(%%%%).1***-+*''))**55CCF>>>>>>CCCCCCC65

For what it's worth, the Wikipedia page on the fastq file format implies that quality scores in NCBI-converted fastq files are automatically converted to the Sanger format.

Thanks.

Eric

[dpryan]

1. Maybe they quality controlled prior to uploading?
2. Look for numbers in the quality score line. If it has numbers, it's Phred+33
3. As above, look for numbers in the quality score. You could look for symbols too, but I always forget which ones come before the capital letters in ASCII and which ones are between the capital and lower case letters. Phred+64 and Solexa scores will look similar, except the latter can contain =.

[winsettz]

I'm sure you saw this on the SRA website, but for reference

Fastq acceptable to SRA

fastq has been narrowly defined with the goal of minimizing submission failures. However, this narrow definition may imply interpretation and conversion by the submitter.

Fastq produced by Illumina pipelines is supported.

The following terms are defined in general:

READNAME = Text string terminated by white space.

BASES::= [ACGTNactgn.]+

QUALITIES ::= [0-9]+ | <quality>\s[0-9]+

or

[\!\"\#\$\%\&\'\(\)\*\+,\-\.\/0-9:;<=>\?\@A-I]+

or

[\@A-Z\[\\\]\^_`a-h]+

The permissible fastq format is simply:

@READNAME
BASES
+[READNAME]
QUALITIES
where each instance of READNAME, BASES, QUALITIES are separated newline.

The QUALITIES string can be whitespace-separated numeric phred scores or an ascii string of the phred scores with the ascii character value = phred score + an offset constant used to place the ascii characters in the printable character range. There are 2 predominant offsets: 33 (0 = !) and 64 (0=@).

A "second opinion" of sorts might be gained by feeding your fastqs into fastqc and see what it thinks.
(http://www.bioinformatics.babraham.a...ojects/fastqc/)

Edit:

And
http://maq.sourceforge.net/fastq.shtml

...Although Solexa/Illumina read file looks pretty much like FASTQ, they are different in that the qualities are scaled differently. In the quality string, if you can see a character with its ASCII code higher than 90, probably your file is in the Solexa/Illumina format.

[efoss]

Thanks very much for the help. These suggestions are really useful.

Best wishes,

Eric