This is a test version of Biostars. For the public version, visit https://www.biostars.org.
What's the best way to map CAGE-seq data to the genome?

Hi,

I've just downloaded several mouse CAGE-seq data from FANTOM5 database.

I tried using bowtie2 default setting to map those rdna.fa files to mm10. And I found a quite low mapping rate around 40% with the majority reads hit more than 1 locus.

I'm totally new to CAGE-seq data. Please forgive me if I ask something silly.

  1. When I looked into the files, the reads seemed being trimmed already. Is this mapping rate normal?
  2. Due to the short length of each read, it's reasonable to hit multiple genomic locations. But won't that raise false positive result when measuring which transcripts are 'really' expressed?
  3. Is there any specific parameter I should apply in Bowtie2?
bowtie cage-seq
• 5,417 views • link
updated by Ram Ram • written by dustar1986 dustar1986

Hi! Could also be a stupid question .. but where did you manage to download the .bam files for the FANTOM5 data, I can only find bed files so far (also in teh CAGEr package). Thanks in advance.

• 0 views • link
written by Vandelnokk • Vandelnokk

2 answers

Hi Dadi,

What you mean exactly with rDNA.fa files? rDNA normally stands for ribosomal DNA... I would recommend downloading the .bam files or just the ctss files.

  1. Normally CAGE reads are trimmed the same way as normal sequences reads beads on sequence quality (default = q=20) When aligning reads to the human genome we normally have at least 80% up to 95%... So I expect to see same ranges in mouse genome.
  2. Yes, by default these multimappers are thrown out. But calculations are made for this and it is only a very small percentage that is missed, because the reads are still at least 27nt long. And indeed there are some problems with pseudogenes for example, so normally these are thrown out of the analysis.
  3. We just use bwa default parameters and works pretty good.
• 0 views • link
updated by Ram Ram • written by Floris Brenk Floris Brenk

Thanks for your detailed explanation, Floris. That is extremely helpful for me. I think I really should download the bam file (mm9) and re-map them to mm10.

• 0 views • link
written by dustar1986 dustar1986

Worked out. Fantom5 uses its own aligner Delve: http://fantom.gsc.riken.jp/5/suppl/delve/delve.tgz

• 0 views • link
updated by Ram Ram • written by dustar1986 dustar1986

I am trying to run delve as well, but always getting segmentation fault error. where you able to run delve?

• 0 views • link
written by sinhashruti • sinhashruti

Log in to answer this question.

More posts like this