"#" in illumina reads fastq quality line

[doublealice]

Hi all,

I am using prinseq to trim low quality tail of illumina reads. From the manual, I found -trim_qual_left and -trim_qual_right can trim seq by quality score from the 5' or 3' -end with certain threshold score. So, here are parameters I used:

#trim 
-trim_qual_right 33
-trim_qual_left 33

I am not sure whether I chose correct parameter and whether 33 is too high.

In addition, I also used below two to trim polyA/T

 
-trim_tail_right 10
-trim_tail_left 10

When program was running, I checked the filtered out reads in so called 'bad_reads" file, I found the quality lines of most of reads contain a very long "#########", some of them even have entire line of #, e.g:

@HWI-ST538:217:C0NFWACXX:4:1101:19625:1943 1:Y:0:GAGTGG
CNCGTCCCTTGATATGTTGTAATTCGTCTTTCATTTCCATTATGATGGCATCTGCAGCATCCTGCCAGAGACCTTTCAGATGAATATTTTCTTGCTGCAA
+
####################################################################################################
@HWI-ST538:217:C0NFWACXX:4:1101:20481:1941 1:Y:0:GAGTGG
TNCATACTTTCGTTCCTTTCTCTTTATACGGATCGACTTCGTTCCAAGCTGTGGGAATCTTGACCGTGTTGTGCATCAGGGGTCATCTGCTTCGGTCATT
+
3#02===@8<@?@7:=@?)>:>><>>@?9???8?4((--<(97<;):)7>7>???9?>???>)<>=99=?##############################
@HWI-ST538:217:C0NFWACXX:4:1101:20349:1946 1:Y:0:GAGTGG
CNGCGCTGCTGCCAACTAGTAAAGGAAGTATTCATTAAAATGCAGGGAGACCGCAGGAATGGGGACATGTTCCCCTTTGGGGACCCTTTTGGCAGCTTCG
+
;#0@-55=?<>>>??9??>.8=9>@>@<?>?=?>?>????>?>?<<=5=???<??<?9>?########################################

I am not quite clear the meaning of symbols in quality. Does multiple "#" really mean these sequences are bad? In "good reads" file, none of read contain "#" in its quality. I am afraid that I did anything wrongly and discard read which should be kept.

The version is illumina 1.9 based on fastQC.

Any advice is highly appreciated.

-alice

[maubp]

Since you have data from Illumina 1.9 this is using the Sanger FASTQ encoding, so '#' (ASCII 66) means PHRED quality 2 which has a special meaning with Illumina as the "Read Segment Quality Control Indicator". Under the old Illumina FASTQ encoding Q2 was a 'B'.

See http://seqanswers.com/forums/showpos...91&postcount=3

i.e. the run of PHRED 2 means the read failed in some specific way according to the Illumina software. Even without this knowledge, PHRED quality 2 is very bad and should be clipped/discarded.

[doublealice]

maubp, thanks very much. This is very helpful.