iGenomes reference genome not accurate?

[Studentlost]

I grabbed the mm9 reference genome from iGenome. When running cuffcompare, I noticed that the genes.gtf refers to chr#_random whereas the chromosome folder only contained chr#, and no files for _random. I corrected this by downloading the chr#_random files from the UCSC mm9 build so now everything matches with the genes.gtf file.

However, the mm9 build came with a pre-index bowtie2 file to use when aligning to the genome (I used TopHat). My concern is that the bowtie2 index was not created with the chr#_random files. Is there a way to check this?

Does anyone know of a better place to grab an accurate mm9 build with bowtie2 index?

If you think the bowtie2 index might be questionable, how can I index all the chromosomes at once to create a version I trust?

Have you guys ran into similar issues?

Thanks a lot for your help.

[GenoMax]

Did you get the mm9 from cufflinks igenomes site?

Try the iGenomes mm9 directly from Illumina: http://support.illumina.com/sequenci...e/igenome.html

[Studentlost]

I am 90% sure that's where I grabbed it from but I'll download it again just to verify.

[Studentlost]

Downloaded it again:

Here's the chromosome list:
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chrY.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr5.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr3.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr2.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr6.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr16.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr15.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr12.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chrM.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr1.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr4.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr9.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr18.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr10.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr14.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chrX.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr11.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr13.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr19.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr8.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr17.fa
Mus_musculus/UCSC/mm9/Sequence/Chromosomes/chr7.fa

It's missing the _random.fa files for each chromosome, which are referenced in the genes.gtf file...

[GenoMax]

If you are interested in the "chr*_random" sequences that are not uniquely placed on the chromosome then you should build a genome file/index on your own.

[lh3]

*_random sequences could be unique sequences in heterochromatin or a large segdup where the flanking cannot be localized or placed. We use these sequences in mapping to reduce mapping artifacts, not really because we are interested in them. I am always curious why Illumina excludes them.

[Studentlost]

I downloaded the _random chromosomes from the UCSC website and built a new Bowtie2 index using this. I ran Tophat with the -G genes.gtf option and the new index.

However....
Here's an error I got from CuffLinks:
GFF warning: merging adjacent/overlapping segments (many of these)
Kept 32976 ref transcripts out of 33802
826 duplicate reference transcripts discarded.

Here's a similar error I got from CuffCompare
Kept 33035 transfrags out of 33262
227 redundant cufflinks transfrags discarded.

So now my GTF file isn't accurate? I got it from the Illumina iGenome mm9 build. I'm starting to think downloading anything from Illumina is more trouble than it's worth.

If I was going to download the unmasked genome, build an index, and download an accurate GTF file for the genome, where would I best go? Or is there a way to verify my GTF file against my index?

What has worked for you in this situation?