RSII mapper benchmarking

[dovah]

Hi all,

I have RNA sequencing data (D. melanogaster) from 3 libraries (1-2Kb, 2-3Kb, 3-7Kb). What would be your suggestion for a mapper? I'm mostly interested in benchmarking isoform detection compared to illumina HiSeq2500. I'd like to do a benchmarking of the "established" tools (since I realized this knowledge is missing), but you can suggest new ones.

Thanks in advance for suggestions.

[Magdoll]

Do you mean how to map PacBio transcriptome (Iso-Seq) reads back to the reference genome or back to the reference transcript?

It also depends on what you have already done for the Iso-Seq data. If you have the results from running the classify + cluster pipeline of Iso-Seq, you will be getting something called "high-quality, quiver-polished, full-length sequences". These are expected to be at least >= 99% accurate. You can align them to the genome using GMAP/STAR or to the reference transcript using BLAST or BLASR.

If you have something called reads_of_insert.fasta (CCS reads) or isoseq_flnc.fasta (CCS reads, but specifically, full-length ones), they are of variable quality ranging from 85-99%+. They can be mapped with GMAP/STAR (but careful with low quality alignments, may need filtering) and BLAST/BLASR too (but again, careful with quality).


Please refer to this tutorial for details on how to use each aligner:

https://github.com/PacificBiosciences/cDNA_primer/wiki/Aligner-tutorial%3A-GMAP%2C-STAR%2C-BLAT%2C-and-BLASR

Additionally the wiki contains may useful information for analyzing Iso-Seq data downstream.

https://github.com/PacificBiosciences/cDNA_primer/wiki/

Also look into Iso-Aux for combining short read + long read data and doing comparisons:

https://github.com/bowhan/isoaux

--Liz

[lingling huang]

Do you mean how to map PacBio transcriptome (Iso-Seq) reads back to the reference genome or back to the reference transcript?

It also depends on what you have already done for the Iso-Seq data. If you have the results from running the classify + cluster pipeline of Iso-Seq, you will be getting something called "high-quality, quiver-polished, full-length sequences". These are expected to be at least >= 99% accurate. You can align them to the genome using GMAP/STAR or to the reference transcript using BLAST or BLASR.

If you have something called reads_of_insert.fasta (CCS reads) or isoseq_flnc.fasta (CCS reads, but specifically, full-length ones), they are of variable quality ranging from 85-99%+. They can be mapped with GMAP/STAR (but careful with low quality alignments, may need filtering) and BLAST/BLASR too (but again, careful with quality).


Please refer to this tutorial for details on how to use each aligner:

https://github.com/PacificBiosciences/cDNA_primer/wiki/Aligner-tutorial%3A-GMAP%2C-STAR%2C-BLAT%2C-and-BLASR

Additionally the wiki contains may useful information for analyzing Iso-Seq data downstream.

https://github.com/PacificBiosciences/cDNA_primer/wiki/

Also look into Iso-Aux for combining short read + long read data and doing comparisons:

https://github.com/bowhan/isoaux

--Liz

How to filter out all GMAP alignments that have less than 90% of the read aligned or less than 80% identity, by using "filterBAM" ?

[Magdoll]

GMAP has two parameters to filter by coverage and identity. Do gmap --help for more details.

The two parameters are --min-trimmed-coverage and --min-identity.