[Genome] hg17 to hg18 liftOver and unmapped regions

[Arkady]

Hi folks.

I'm looking at microarray data, specifically from the Affymetrix whole
genome tile array (Kapranov et all 2007) from the May 2004 build. (It's got
a track on the genome browser as well, Affy Phase 3 or somesuch.)

I took Affy's original transcribed fragments list for HDFs (consisting of
ranges of genome locations in single-chromosome files) and converted them to
the position format. Then I ran them as one file through liftOver from hg17
to hg18, and mostly it seems to have worked. However, I also got around 20
lines of unmapped locations. The number of unmapped locations is reduced if
I lower -minMatch significantly, but there are a couple that don't seem to
go away.

The first one:
#Deleted in new
chr1:103788371-103788472

If I get the DNA for that fragment (an exon of RNPC3) off of May 2004 and
search for it with BLAT in March 2006, it shows up with 96.9% identity
(span=94, qsize=102, high score) at:
chr3:70,258,127-70,258,220
That aligns with several other species' RNPC3. Why is this not mapped in the
chain file?

Here are the rest (with -minMatch 0.4)
#Deleted in new
chr1:103789464-103789522 # doesn't map well
#Deleted in new
chr1:103789976-103790082 # does map well, also to chr3
#Deleted in new
chr1:103790783-103790844 # does map well, to chr3
#Deleted in new
chr1:103791990-103792116
#Deleted in new
chr1:103796005-103796100
#Deleted in new
chr1:103797938-103797998
#Deleted in new
chrX:114795577-114795630 # maps equally well to 7 locations on chrX

I didn't check the three above that have no comments.

I suspect I'm misunderstanding how the chains are created. I suppose I could
just write a script to BLAT the "deleted" transcribed fragments and get the
correct mappings, but I feel like this is a little bit of a hack.

Thanks for the help.

Cheers,
John Woods

Center for Systems and Synthetic Biology
Institute for Cellular and Molecular Biology
The University of Texas at Austin