Questions about solexa quality score！

[baohua100]

reads.fq file:

@4:1:518:715
GATACCATAAAAGCTGGATCCTTCTTCAAGCATAA
+4:1:518:715
hhhhhhhhhhhhhhhdhhhhhhhhhhhdRehdhhP

1. How to change character (like 'e' or 'h') to quality score?

2. What's the meaning of this score? How to compute this score （ formula ）？

[Farhat]

For a Fastq file, if the quality character is $q the corresponding Phred quality can be calculated with the following Perl code:

$Q = ord($q) - 33;

[SoupDragon]

This is correct if you are using quality scores encoded in "fastq" format. I believe the Illimina pipeline used a different ascii offset (64) according to their pipeline documentation. A value of zero = ascii 64 ('@'). The ascii value for a qv is therefore qv+64. So "h" = 104 - 64 = 40

[Farhat]

Dupe. Deleted.

[Farhat]

Originally Posted by SoupDragon

This is correct if you are using quality scores encoded in "fastq" format. I believe the Illimina pipeline used a different ascii offset (64) according to their pipeline documentation. A value of zero = ascii 64 ('@'). The ascii value for a qv is therefore qv+64. So "h" = 104 - 64 = 40

You are right. 'h' would make the quality way beyond 40 by my calculation.

[baohua100]

Thanks.

what's the range of this score ? (0---40 ?)

what's the meaning of this score?

[sparks]

The range is from -5 to 40

If P is probability of base then Solexa quality is 10 log10(P/(1-P))

A quality of -5 corresponds to P=0.25

[Farhat]

Originally Posted by sparks

The range is from -5 to 40

If P is probability of base then Solexa quality is 10 log10(P/(1-P))

A quality of -5 corresponds to P=0.25

In my datasets the range has been from -40 to 40.

[sparks]

Farhats right for Solexa prb file formats from the base caller but for fastq format files the OP asked about, the range should be -5 to 40

[Farhat]

Originally Posted by sparks

Farhats right for Solexa prb file formats from the base caller but for fastq format files the OP asked about, the range should be -5 to 40

Yes, that's right, because for solexa PRB file the probability of A,C,G or T is given separately, and can be really low, whereas for fastq the lowest probability is 0.25 implying equal probability for any nucleotide.

[baohua100]

$sQ = -10 * log($e / (1 - $e))

when $sQ =40, $e=0.0001

when $sQ=0, $e=0.5

0.5>0.25


when $sQ=-4 $e=0.72



what's the probalibity of error?????????????????????????????

[sparks]

If you are talking fastq format and have a quality of -4 then the probability of the base called is 0.28 and probability it is anyone of the other 3 bases is 0.72.

If you see a -4 in a prb format file then the probability of the base is 0.28 and the other bases will each have their own prb/qual value.

[mikertesz]

The output of a Solexa run generated a "quala" file of the following format:

>sequence_0
40 40 40 19 7 40 40 40 40 40 31 40 40 40 40 40 40 40 40 40 40 11 40 40 40
36 40 12 40 21 39 1 4 40 40 15 40 40 4 40 40 10 40 40 40 40 40 2 4 10
1

>sequence_1
40 40 8 13 12 40 40 40 40 17 27 40 25 17 4 40 40 40 21 40 40 37 40 40 37
4 40 33 40 25 40 3 20 40 40 20 40 40 4 40 8 7 40 40 15 4 10 1 5 20
1

etc...

Does anybody know what those numbers mean? Are those simply the Solexa quality scores per base-pair? The range seems to be 1-40 --- why isn't it -5 to 40 as in fasq?

[vruotti]

Hi,
Does anyone know an easy way or an existing program to convert all the .prb files from one particular lane into one fastq file? Similar to the s_1_sequence.txt file but with no filters applied?
We have trying hacking around the Perl scripts within GERALD but looks like you need an intermediate seqpre.tmp file which I think gets deleted after the completion of GERALD.

We know this is possible by just running GERALD with the fastq parameter. However, we would like to generate a fastq file that is not affected by GERALD's filters. That way we can set up our own quality filters.

Any ideas?
Do I go ahead and write one?

Thanks,
Victor

[swbarnes2]

I made my own very simple script, but here's a script of James Bonfield's here:

http://seqanswers.com/forums/showthread.php?t=282

The only problem ithat I see is this line

foreach (glob("$fn/*seq.txt")) {

which is going to get every single .seq in the directory, not just the ones from a single lane. So you'll have to fix that.

[kmcarr]

Victor,

Run GERALD including the following line in the GERALD configuration file:

QF_PARAMS '(1==1)'

This is a conditional which is true 100% of the time; in other words, GERALD passes every read.

(This technique comes from the Pipeline User Guide)

[ShaunMahony]

As said below (and also in the Solexa documentation), Solexa quality scores in their Fastq-like format are given by 10*log_10(P/(1-P)). I thought it might be useful for some people if I posted a lookup table based on this. Note I'm giving the probability that a base is erroneous, rounded to four decimal places. Please post a reply if you think this table is an incorrect translation:

Char ASCII Char-64 P(error)
; 59 -5 0.7597
< 60 -4 0.7153
= 61 -3 0.6661
> 62 -2 0.6131
? 63 -1 0.5573
@ 64 0 0.5000
A 65 1 0.4427
B 66 2 0.3869
C 67 3 0.3339
D 68 4 0.2847
E 69 5 0.2403
F 70 6 0.2008
G 71 7 0.1663
H 72 8 0.1368
I 73 9 0.1118
J 74 10 0.0909
K 75 11 0.0736
L 76 12 0.0594
M 77 13 0.0477
N 78 14 0.0383
O 79 15 0.0307
P 80 16 0.0245
Q 81 17 0.0196
R 82 18 0.0156
S 83 19 0.0124
T 84 20 0.0099
U 85 21 0.0079
V 86 22 0.0063
W 87 23 0.0050
X 88 24 0.0040
Y 89 25 0.0032
Z 90 26 0.0025
[ 91 27 0.0020
\ 92 28 0.0016
] 93 29 0.0013
^ 94 30 0.0010
_ 95 31 0.0008
` 96 32 0.0006
a 97 33 0.0005
b 98 34 0.0004
c 99 35 0.0003
d 100 36 0.0003
e 101 37 0.0002
f 102 38 0.0002
g 103 39 0.0001
h 104 40 0.0001

[vruotti]

Hello,
We are looking a little closer at the quality of one of our runs. Interestingly, we see a pattern in most of our runs right at the 30th cycle. The information from the graph below comes from the s_N_export.txt files. Please ignore the graph from lane 4. This was a failed lane. The others however, including our control (lane 8) show this pattern.

This was an IPAR run with the upgraded GAII and was one of our best runs. Other runs also show this pattern at the 30th cycle. Does anyone know the reason why the qualities drop so much after the 30th cycle? Have you seem this before in any of your runs?

Thanks in advance.
Victor

[TylerBackman]

Originally Posted by vruotti

Does anyone know the reason why the qualities drop so much after the 30th cycle?

This is most likely because the qualities you are looking at are alignment normalized, and a large number of your sequences failed to align to the reference genome (due to a ligated adapter, etc.)

Take a look at the un-normalized scores (s_<lane>_qraw.txt) instead, I think you'll find that the curve is more continuous between cycles.

[baohua100]

fastq file:
@I326_2_FC306FCAAXX:8:1:50:985
ATGTCCGAAGGGCAGTCTCAAGTGGTAAAATGGAT
+I326_2_FC306FCAAXX:8:1:50:985
hhhWhhhchhhhhahShh\PO]LgXZXPNLUTZNO


MAQ alignment output:

I326_2_FC306FCAAXX:8:1:50:985 1 1 + 0 0 99 99 99 0 0 1 0 35 ATGTCCGAAGGGCAGTCTCAAGTGGTAAAATGGAT ```W```````````S``\PO]L`XZXPNLUTZNO

what's the meaning of ```W```````````S``\PO]L`XZXPNLUTZNO ?

not the same as hhhWhhhchhhhhahShh\PO]LgXZXPNLUTZNO