This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Incomplete filtration of reads by length in .fastq and in .bam files

Hi,

I want to filter my reads to have only 29-31 nt long ones. I tried these three methods:

## 1:
awk 'BEGIN {FS = "\t" ; OFS = "\n"} {header = $0 ; getline seq ; getline qheader ; getline qseq ; if (length(seq) >= 29 && length(seq) <= 31) {print header, seq, qheader, qseq}}' < in.fq > out_filtered.fq

## 2:
reformat.sh in=in.sam out=out_filtered.sam minlength=29 maxlength=31

## 3:
seqkit seq in.fq -j 10 -m 29 -M 31 > out_filtered.fq

At the end, none of them had reads longer than 31 nt but all of them had some shorter than 29 nt. Around 5% of the filtered reads are 28 nt long and there are less than 1% of them with less than 28 nt length.

Why is this happening and how can I fix this?

Or is it possible that the scripts 'psite' and 'phase_by_size' from the plastid package is doing something wrong to give this result?

Here is the link for two pics of results of 'psite' and 'phase_by_size': https://photos.app.goo.gl/sGhT2dvoNsAQHF7d7

Thanks.

Farkas

rna-seq
• 1,390 views • link
updated by Biostar Biostar • written by kerenyifarkas • kerenyifarkas

Can you try following two step procedure with BBtools?

reformat.sh in=in.sam out=stdout.sam maxlength=31 | reformat.sh in=stdin.sam out=out_filtered.sam minlength=29
• 0 views • link
written by GenoMax GenoMax

Following works for me using @ATPoint's test file.

$ reformat.sh in=../test.fq out=stdout.fq minlength=29 maxlength=31

Input is being processed as unpaired
@30
AAAAACCCCCGGGGGTTTTTGGGGGAAAAA
+
AAAAACCCCCGGGGGJJJJJGGGGGAAAAA
Input:                          4 reads                 120 bases
Short Read Discards:            3 reads (75.00%)        90 bases (75.00%)
Output:                         1 reads (25.00%)        30 bases (25.00%)

Time:                           0.072 seconds.
Reads Processed:           4    0.06k reads/sec
Bases Processed:         120    0.00m bases/sec

OR

$ reformat.sh in=../test.fq out=stdout.fq maxlength=31 | reformat.sh in=stdin.fq out=stdout.fq minlen=29 int=f

Set INTERLEAVED to false
Input is being processed as unpaired
Input is being processed as unpaired
Input:                          4 reads                 120 bases
Short Read Discards:            1 reads (25.00%)        56 bases (46.67%)
Output:                         3 reads (75.00%)        64 bases (53.33%)

Time:                           0.069 seconds.
Reads Processed:           4    0.06k reads/sec
Bases Processed:         120    0.00m bases/sec
@30
AAAAACCCCCGGGGGTTTTTGGGGGAAAAA
+
AAAAACCCCCGGGGGJJJJJGGGGGAAAAA
Input:                          3 reads                 64 bases
Short Read Discards:            2 reads (66.67%)        34 bases (53.13%)
Output:                         1 reads (33.33%)        30 bases (46.88%)

Time:                           0.084 seconds.
Reads Processed:           3    0.04k reads/sec
Bases Processed:          64    0.00m bases/sec
• 0 views • link
written by GenoMax GenoMax

To me, the awk seems to do the job (I didn't check the other methods).

For control, you could do {print length(seq), seq}.

And check manually the corner cases (e.g. len == 29).

Do you really convert the out_filetered.fq to *.bam and run phase_by_size on that?

• 0 views • link
written by massa.kassa.sc3na massa.kassa.sc3na

Agreed, you should check manually if this tool is not reporting something odd.

I did:

$ cat test.fq
@longer
AGCTCCGATCGATCGATCGATCGAGTAGCTAGCTACGATCGATCGATCGCACTAGC
+
AGCTCCGATCGATCGATCGATCGAGTAGCTAGCTACGATCGATCGATCGCACTAGC
@shorter
AATTGC
+
AAAAAA
@30
AAAAACCCCCGGGGGTTTTTGGGGGAAAAA
+
AAAAACCCCCGGGGGTTTTTGGGGGAAAAA
@28
AGCTAGCTAGCTAGCTAGCTAGCTAGCT
+
AGCTAGCTAGCTAGCTAGCTAGCTAGCT

$awk 'BEGIN {FS = "\t" ; OFS = "\n"} {header = $0 ; getline seq ; getline qheader ; getline qseq ; if (length(seq) >= 29 && length(seq) <= 31) {print header, seq, qheader, qseq}}' < test.fq 
@30
AAAAACCCCCGGGGGTTTTTGGGGGAAAAA
+
AAAAACCCCCGGGGGTTTTTGGGGGAAAAA
• 0 views • link
written by ATpoint ATpoint

Yes, checking manually it gives only reads 29-31 nt.

out_filtered.fq files were mapped and the psite and phase_by_size were run on the generated .bam files.

So now the question is why these scripts show this weird results... Maybe I should just skip this plastid stuff and use something else instead.

Thanks!

• 0 views • link
written by kerenyifarkas • kerenyifarkas

0 answers

No answers yet.

Log in to answer this question.

More posts like this