What is your preferred way to map cDNA sequences to genome?

Question

[jon.brate]

I have a set of cDNA-sequences representing mRNAs and I want to create a gtf/gff file of these. I have tried to use BLAST and BLAT, but I find it difficult to obtain a single reliabe hit that represents all the different exons. And especially with BLAT I get many small hits for each sequence. I also tried to map the sequences using TopHat, but for some reason it crashed, perhaps due to the long sequences.

Any suggestions on how to map such sequences to a genome with reliable information about the exon/intron junctions?

Answer 1

[Nicolas Rosewick]

Use gmap. It was build for that and outputs a gff file with each exon of every sequence in input. Very usefull !

[jon.brate]

Thanks, I'll check it out. Does it also create an entry representing the entire gene in the gff-file?

[Sean Davis]

Aligners do not really have the concept of a gene, only transcript. If you want to treat each transcript as a gene, that can be done pretty easily by adding a tag to the resulting GFF annotation column in a post-alignment step.

Answer 2

[Istvan Albert]

I never had to solve this problem myself but I am guessing that a long read aligner such as lastz executed with smaller (zero?) penalty for gap extension might work.

[jon.brate]

Thanks! I never heard of that program. Will check it out.

Answer 3

[Ram]

Have you tried blastx to compare against protein databases and then backtrack to the genes from the top protein hit? That could solve your problem with minimal human intervention, if you script the efetch to get gene from protein identifier. Exon junctions can then be an SQL query to UCSC Genome Browser.

[jon.brate]

I didn't mention that, but in my case I have non-protein coding sequences from a non-model species.

[Ram]

Oh, I see. My solution is not useful then. Looks like there are better tools anyway!