[Fabrice ODEFREY]

SVA only support hg18 so no point using it for the moment if you aligned to hg19.

[laura]

ensembl works on both NCBI36 and GRCh37 (hg18/hg19)

[Papillon]

I'm an intern working with exome analysis and I am facing the same thing. I want to implement this annotation into my own java pipeline and although the solution may seem very easy, I am troubled in finding the right approach.

I have tried using the RefSeq annotation file of SeqCap EZ Exome v2 (matching UCSC genome browser with HG19), which holds information on cdsStart, -End en exon starts and endings. This file also holds an Ensembl gene reference for each RefSeq gene, which should make it easy to link with the cDNA fasta file of Ensembl and get exactly what I want...

... a few problems though:
1) RefSeq- and ensembl genes overlap and multiple of the same ensembl references may occur in the ensembl fasta file, making it hard to differentiate. This is most likely due to different isoforms.
2) Looking at a few cases, I noticed that some RefSeq genes show cdsStart and cdsEnd positions that can not be traced back to ensembl. In other words: when I read the ensembl reference from the RefSeq file and look them up in the ensembl file, I can find multiple isoforms, but none with the same cdsStart and/or cdsEnd. I already take into account that RefSeq and Ensembl differ 1 nuc. in cdsStart. Both files are based on HG19, so that can't be the problem either.

What would be the best approach on solving this puzzle? Should I just walk through the entire genome and annotate all the information to my SNPs as I go along? Any thoughts are welcome.

Thanks a bunch!

[laura]

If you have hg19/GRCh37 positions for all your snps I would suggest using a tool like the ensembl variant effect predictor to get the consequences of your snps and then tracing to refseq ids using the ensembl xref system rather than doing it the other way around

Refseq models and Ensembl models should be mostly the same for the cds coordinates (though not in all models) but to get the models which are identical across both sets it best to look at the ccds models http://www.ncbi.nlm.nih.gov/projects...CcdsBrowse.cgi

Do remember that utr coordinate may be different across both sets

[jnfass]

We're looking at using Ensembl's Variant Effect Predictor right now ... but with a non-model organism (or at least a model organism with some regions replaced by a "better" assembly). It seems like we'll need to set up a local database in order to provide the reference and gene models. Does anyone have experience setting up a db like this?

[apratap]

You could also try :

http://genome.sph.umich.edu/wiki/VcfCodingSnps

The requirement is that the input snps be in VCF format.

-Abhi

[jnfass]

Thanks, but it seems to only do SNPs ... we need to annotate indels, too.

[laura]

For help with setting up a custom ensembl database I would suggest emailing [email protected] to get help

[Michael.James.Clark]

Thanks, but it seems to only do SNPs ... we need to annotate indels, too.

Try Annovar. http://www.openbioinformatics.org/annovar/

[nexgengirl]

Check out the Polyphen-2 website. If you run your SNPs through the batch page:

http://genetics.bwh.harvard.edu/pph2/bgi.shtml

it will output a SNP file which is a mapping of all your SNPs. If you work with different settings for the advanced options on the input page you should be able to annotate the SNPs in many ways. Hope this helps!

[jnfass]

@Michael.James.Clark ... Annovar seems to require annotation databases from the UCSC Genome Browser. That doesn't exist for at least part of the genome I'm working with.

@nexgengirl ... Polyphen-2 seems to be restricted to human

@laura ... thanks for the suggestion; I might try contacting ensembl.

But in the meantime someone responded to my Biostar post and suggested snpEff (snpeff.sourceforge.net) and it seems to fit the bill. If anyone has had good or bad experience with it, I'd appreciate hearing about it.

[Michael.James.Clark]

@Michael.James.Clark ... Annovar seems to require annotation databases from the UCSC Genome Browser. That doesn't exist for at least part of the genome I'm working with.

@nexgengirl ... Polyphen-2 seems to be restricted to human

@laura ... thanks for the suggestion; I might try contacting ensembl.

But in the meantime someone responded to my Biostar post and suggested snpEff (snpeff.sourceforge.net) and it seems to fit the bill. If anyone has had good or bad experience with it, I'd appreciate hearing about it.

For your and others' information, Annovar doesn't require annotation databases from UCSC. You can make your own annotations and feed them to the program. Read the documentation.

[jnfass]

Ok ... thanks.

You're probably referring to this.

[Michael.James.Clark]

Yeah, I'm not saying it is necessarily the easiest solution for you, but it is incorrect to state that it requires annotation databases from UCSC.

[colindaven]

Thanks for the idea on SNPeff. It's working very well for me and even supports some bacterial genomes out of the box. Input format is also easy to generate, as well as VCF or pileup.