How To Mask Repeats In Ngs Data.

Question

[Daniel]

How can I mask repeats in Next Generation sequencing data? I several million NGS reads from a mammalian genome that was not sequenced yet. I would like to filter out those that have a significant hit against the RepBase or Repeatmasker databases. I would appreciate if anybody could give me more specific instructions.

Answer 1

[JC]

I simply filter reads from repetitive sequences using 2 approaches:

1) simple repeats and low complexity sequences can be filter with DUST or I just compute the complexity of the sequence using entropy or compression ratio.

2) interspersed repeats can be filters if you map the reads to the RepBase consensi with Bowtie, BWA or Blat (with -fastMap), this step can filter millions of reads in a few minutes.

If you are expecting a lot of repetitive sequences (as in genome genome sequencing), I strongly suggest to filter first before mapping/assembling, otherwise it doesn't gives you any advantage.

[brentp]

could you clarify #2? you mean you'd filter reads that map to multiple places?

[JC]

No, you can map the reads to the consensi sequences from known repeats obtained from RepBase or any other source filtering out those reads that match.

[pyKey]

Dear JC,

I have some very basic questions about how to map reads to the Repbase consensi, Could you please give me details on? - What is a Repbase consensus? Is it distinct for each repeat family? Is it distinct over species? - Where can I find it/them for Human? - Do I build a regular bowtie2 index from this consensus file?

Many thanks,

Answer 2

[Leonor Palmeira]

You could either:

• directly repeatmask your data : http://www.repeatmasker.org/
• map your data against a close mammalian genome and cross the matching positions with the repeatmasker positions

Answer 3

[Ian]

Just a thought (i.e. not sure it is practical/possible). But if you could obtain the repetitive sequences from RepBase you could use them as the reference sequences for an NGS sequence aligner, e.g. Bowtie. Any uniquely mapping reads could be excluded from your sample.

Answer 4

[Biojl]

A very simple and easy to use tool is SEG. It will replace the repeats and/or LCR in your sequences for 'XXXX'

[SES]

This is for protein sequences. It is also for masking mathematical repeats, not the type OP is interested in (which is identifying sequences using a reference library).

Answer 5

[Dattatray Mongad]

You can use "tantan" which is used by LAST in orderto mask genomes before comparing