Description
There are two tracks in this composite collection:
 Gap Overlaps  Paired exactly identical sequence on each side of a gap
 Tandem Dups  Paired exactly identical sequence survey over entire genome assembly
The Gap Overlaps is thus a subset of the full Tandem Dups track.
This investigation began when an unusual number of paired sequences
around gaps was noticed during the mouse strain sequencing project.
This naturally raised the question, how common is this feature, and what
type of assemblies can it be found in.
The Gap Overlaps track indicates any pair of exactly identical sequence
on each side of gaps. Where a gap is any run of N's, including
a single N. The end of an upstream sequence before the gap
is duplicated exactly at the beginning of the downstream sequence
following the gap in the assembly.
The Tandem Dups track is a similar survey over the entire genome
assembly. The separation gap between these paired sequences
can range from 1 base up to 20,000 bases.
Methods
The Gap Overlap duplicate sequences were found by
extracting 1,000 bases before and after each gap and aligned to
each other with the blat command:
blat q=dna minIdentity=95 repMatch=10 upstreamContig.fa downstreamContig.fa
Filtering the PSL output for a perfect match, no mismatches,
and therefore of equal size matching sequence,
where the alignment ends exactly at the end of the upstream sequence,
and begins exactly at the start of the downstream sequence.
The Tandem Dups paired sequences were found with the following procedure:
 Generate 29 base kmers for the entire genome, allow only kmers
with bases: A C T G, no N's allowed.
 Pair up identical kmers with at least one base separation and
up to 20,000 bases separation.
 Collapse overlapping kmer pairs when they are the same size of sequence
and the same spacing between the pairs. This procedure preserves the
definition of duplicated identical pairs.
 The resulting pairs can now be longer sequences with smaller separation
then the constituent pairs
 Final result selects sizes of 30 bases or more for the size of the
paired sequence, and at least one base remaining as a separation gap.
 Collapsed pairs that close the gap are discarded. They appear to
indicate simple repeat sequences when this happens. It would be
interesting to have this result available, but that is not available
at this time.
The reason for starting with 29 base sized pairs and then selecting
results of at least 30 base sized pairs results in a reasonable
number of 30 base pairs. If the procedure starts with 30 base
sized pairs, it produces way too many 30 base kmer pairs for
a reasonable count.
See Also
Interactive tables of all results:
Credits
Thank you to Joel Armstrong and Benedict Paten of the
Computational Genomics Lab
at the
U.C. Santa Cruz Genomics Institute
for identifying this characteristic of genome assemblies.
The data and presentation of this track were prepared by
Hiram Clawson,
U.C. Santa Cruz Genomics Institute
