SGD liftOver chain files issue

[Slacanch]

SGD (Saccharomyces Genome Database), contains a repository of liftOver chain files that allow to convert form any version of the reference genome to the last one.

after trying, and failing multiple times, to convert between two assemblies, i decided to take a look at the chain files themselves and i compared the chain files provided from sgd with the same file provided by the ucsc genome browser (R61 to R64 from sgd, sacCer2 to sacCer3 from ucsc).

the result is rather odd; this is the beginning of the file in the two instances:

UCSC:

chain 21724089 chrI 230208 + 0 230208 chrI 230218 + 0 230218 16
3834	1	0
2091	0	1
527	1	0
10002	1	1
10	1	0
29	1	0

SGD:

chain 21724089 chr01_2008_03_05 230208 + 0 230208 chr01_2011_02_03 230218 + 0 230218 1
3834	1	0
2091	0	1
527	1	0
10002	1	1
10	1	0
29	1	0

most of the file is actually the same, but the definition of the chromosomes is really strange, it has dates attached! which render the file practically useless (i've managed to successfully liftover using the ucsc chain file, while i could not using th SGD one)

chrI 230208   vs     chr01_2008_03_05 230208

the strange pattern repeats for every chromosome in the sgd file.



is there something i missed? am i the only one having this issue? what is the meaning of this thing? i am confused

[raim]

sgd liftover problem solved?

Hi Slacanch,
Have you ever managed to solve this problem?
best,
Rainer

[raim]

Hi,

The SGD helpdesk said the date appendix in chromosome names can just be deleted.
Also, at least the liftOver reimplementation in R/bioconductor's rtracklayer can't handle the comment lines starting with ##.

So the following command-line processing seems to do the trick, although I have not tested the actual mapping result yet:

grep -v "##" V43_2004_07_26_V64_2011_02_03.over.chain | sed 's/\(chr..\)_[^ ]*/\1/g' > V43_2004_07_26_V64_2011_02_03_fixed.over.chain

Note that in your case you would still need to convert the roman chromosome numbers
to two digit arabic, e.g. chrI should be chr01. And watch mitochondrial chromosomes, chrM is not chr1000

Rainer