programs for filtering low complexity

[swarbre]

This question was asked elsewhere but I have the same question as below, any help.

1) I have many "low-complexity" reads. Some are simply polyA, polyC,
> etc. But some others are runs of "ATATAT" or "CACACACA", etc. Previously
>
> I would have used "dust" on the command line to filter out this kind of
> read in a fasta file. Any ideas on how to achieve similar functionality
> in the ShortRead world?

[olus]

Look at :

https://stat.ethz.ch/pipermail/bioc-sig-sequencing/2009-March/000191.html

http://www.mail-archive.com/[email protected]/msg00148.html

Though not a ShortRead package:

http://genome.gsc.riken.jp/osc/english/software/src/tagdust.tgz

Cheers

[mdaskal]

Hi,

I am looking for a definition of low complexity reads for reads of variable lengths (about 100 nucleotides long).

Right now, I am using the following definition:
- Divide a read in subsegments of 32 nucleotides. (last subsegment is overlapping one before last)
- Count number of unique tri-nucleotides in each segment.
- If number of unique tri-nucleotides is smaller than 5, then the segment is of "low complexity"
- If there is at least one "low complexity" segment, the read is considered "low complexity"

Comments regarding the relevancy of this definition would be appreciated.

Regards, Michael.

[dan]

Hi,

I am looking for a definition of low complexity reads for reads of variable lengths (about 100 nucleotides long).

Right now, I am using the following definition:
- Divide a read in subsegments of 32 nucleotides. (last subsegment is overlapping one before last)
- Count number of unique tri-nucleotides in each segment.
- If number of unique tri-nucleotides is smaller than 5, then the segment is of "low complexity"
- If there is at least one "low complexity" segment, the read is considered "low complexity"

Comments regarding the relevancy of this definition would be appreciated.

Regards, Michael.

Why 5?

How does your operational definition compare to SEG / Dust?

[mdaskal]

Low complexity reads

Hi Dan,

I am not familiar with the definition of SEG / Dust.
Where can I find some details about it?

Would you suggest another limit than 5 unique tri-nucleotides (higher?lower?)?

Regards, Michael.

[dan]

I guess they are both in PubMed? Internet is slow here atm or else I'd link...

I don't suggest an alternative without data... so my question really was, did you analyse / benchmark your metric? i.e. what fraction of reads are low complexity at 3, 4, 5, 6, etc...