hgBlat does not use gene annotations. One difference between command-line blat and hgBlat is that our gfServer uses -stepSize=5 for extra sensitivity in order to support PCR use.
Adding that parameter to your command-line blat might help, at the cost of doubling RAM usage. However, even after that you will find differences in the output of the various blat tools. As I mentioned in another email, you probably want to use an RNA database target instead of genomic DNA to avoid the splicing issue. -Galt 8/30/2010 4:03 PM, [email protected]: > Hi, > I am using BLAT to map RNA-seq short reads (50bp) from next-gen sequencing > data. It works great except that I am seeing an inconsistency between the UCSC browser web-based BLAT results and the command line BLAT when mapping to splice junctions. > > In situations where the mapping to the splice junction is ambiguous, the web-based BLAT correctly maps to the precise annotated junction, whereas the command line BLAT does not. Here is an example: > > AGGCGCATG | *TCGA*-----------------------| *TCGA*CAAGCTAGC > > > (annotated splice junction marked by ' | ' ) > > If there is a read that spans the above junction, the *TCGA* at the beginning > of the intron and at the beginning of the 2nd exon cannot be differentiated. A read spanning the junction can either map to include the TCGA in the intron and then jump to CAAGCTAGC in the 2nd exon OR it can jump from the CATG at the end of the 1st exon to the TCGA in the 2nd exon. > > It seems that the web-based BLAT correctly maps to the annotated junction, > while the command-line does not. Does the web-based BLAT somehow have an additional input of gene annotations? It's not clear to me how the exact same read can map one way with the web-based BLAT, and another way with the command line BLAT. Regardless, is there a way to constrain these ambiguities using a known-gene annotation? > > Thanks. > > _______________________________________________ Genome maillist - [email protected] https://lists.soe.ucsc.edu/mailman/listinfo/genome
