unencumbered haplotype estimation

[Richard Finney]

No. Its' one of about 6 projects I could get done if I had 2 more hours a day.
I didn't get very far with it.

Though if I were doing it now I'd use webassembly.

[greekkey]

Hi, Richard, Are you still going on with your haplotype phasing project?

[Richard Finney]

Thanks husamia. I checked out Stephens Lab. It does appear to be excellent work; but I want to avoid serious licensing or patent issues; I see this : http://depts.washington.edu/uwc4c/ex...ets/fastphase/
With words like : The conditions and financial terms of this license apply only to those who download the license, sign it and return it "as is" to the UW Center for Commercialization. Some Express Licenses may be modified for an additional cost. Other licensing arrangements may be available.

What'd I'd want is something more open, less cumbersome.

[husamia]

I think this is great project. Its going to take sometime to think about it.

The big question is phasing. Do you or not know the phase?
I seems from your post that you don't know the phase then do you want to infer it bioinformatically using basic assumptions? Refer to phase indifference problem Check out PHASE- Stephen lab for implementation of current methods.

the most simplistinc problem is you have child, and two parents then assumptions> then infer the phase for homozygous in either one the hets in both parents is not possible.

This is a big problem and I think best way is to assume you know phase and go from there.

If you interested in persuing this further need to spend more time and discussion.

Can you give a visual representation of what the page looks like in your mind as example for clarification?

[Richard Finney]

The problem is this:

A = homozygous reference
B = heterzygous
C = homozygous non-reference

rows are samples, columns are genomic position
header row is labels, data is ordered by genomic position
This data is contrived.

The data:

123456789
ABAAAAABC
ABCAAAABC
ABCAAAABC
BBAAAAABA
ABCAABABC
ABCAAAABC
BBCAAACBC
ABCAAAABC

The x and y dimensions can get very large (1000x10000 for example).

I want to develop an open source, graphical tool to browse this data.
The core code could be ported anywhere but the first implementation would be for cgi/html/html5 I want a user to be able to mouse over a column and see what's in "linkage disequilibrium" (LD) with that particular genomic position. Speed, usability and reasonable accuracy are critical specs.

The problem, is, of course "phasing" (or whatever it's called) the B (heterzygous) calls so that LD can be calculated.

The questions are: what is the best non-patented algorithm for this? "Un-encumbered" by patent is fine, that is if it is patented but unrestricted license is granted. Also, is there an implementation I could check my results with? GPL/LGPL/MIT/DWTFYW whatever is fine. I can port anything to the target language (C for absolute speed).

This would be a great tool for checking out GWAS results using the newly emerging NGS data sets: "what else is near that sneaky little problem snp?"