Hi all,
I have a fasta file containing thousands of gene sequences. I want to create a GTF annotation file that describes the locations of these genes, their exons etc.
I tried to turn the fasta into a fake fastq file with BBMap, then upload to galaxy, use BWA aligner (all of which worked) and then use Cufflinks to assemble a GTF, but it didn't work. I think the problem might be that the sizes of the fake reads varies greatly; When I use the default parameters I get a gtf file with only one gene in it, and when I try to take the parameters to the extreme so that all lengths of reads will be taken into account, it gives me an error (return code -11, if that helps).
Does anyone have an alternative strategy, or a way to make this one work?
Thanks!
I think the strategy you outlined (BWA -> Cufflinks) seemed to work better if you had started with short reads of RNA-seq. It also seem like BBMap generates fake read pairs from single reads (millions?), whereas you started with thousands of complete gene sequences.
If I understood you correctly and you had complete gene sequences in a fasta file, I can think of two alternative ways:
1. Use a short read simulator to generate millions of reads to simulate RNA-seq, then proceed with BWA and Cufflinks. The equivalent in BBMap should be randomreads.sh instead of bfakereads.sh
see also: http://alumni.cs.ucr.edu/~liw/rnaseqreadsimulator.html
see also: http://www.niehs.nih.gov/research/resources/software/biostatistics/art/
2. Blast the complete gene sequences against the genome, generate a bed file by post processing the output (maybe take the chromosome that best hits? the exons should be neighboring fragments), then convert this bed file to gtf.
Hi Alon
You could try looking into the tools used to map PacBio or Nanopore (RNA) reads - since these are typically very long reads (with different sizes) covering multiple exons they should have all the same properties as your data (sorry but I have not worked with those data so I cannot recommend any specific tools).
Just remember to turn off the error corrections in these tools as they are specific to the platform.
/Kristoffer
Try gmap gff3 output format. If needed you can create a GTF of sorts out of the GFF3. Gmap is excellent for this sort of thing.
For a much more complicated pipeline look at Maker.
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology