ImmPort/Sketch

''' I think this page is a fork of page ImmPort/Demo sketch. Do not edit.'''

(Based on email from Alan to Jonathan 2009-05-12)

Here are key pieces of a feasible demo, assuming we continue along the pdb direction:

Have a clear, reasonable model for PDB structures. Compute (based of coordinates) or import (from IMGT) points of contact between HLA complexes and short polypeptides. We want this in the form peptide-to-peptide (i.e. AA-to-AA).

The 4-level partonomy is as follows: A structure has multiple molecules (some of which are polypeptide chains); a chain is a series of amino acids; an amino acid is made from atoms. An atom defined by coordinates and element. Sometimes two similar chains are described as "variants" of a single chain.

This involves reasonably steady work (Dan), mostly to understand the convoluted PDB representation. (Many quirks, e.g. the conversion to XML may be lossy.) Dan seems to have tools from the python library to do the manipulation if he can figure out what stuff means. [...]

For the contact points the most accessible reference seems to be the one I sent out on 5/12. (The other two references are the paper we got from Paula, and the IMGT documentation that Nishanth sent.) Essentially our model needs to be of features placed at sequence coordinates on the components of complexes. Relations such as contact points are in terms of pairs of these.

PDB will supply one such set of features, the contact points for some set of alleles, and some set of peptides.

The SFVTs that are supplied by them - these are features on just one allele. This is work I do.

Be able to use the IMGT alignments to go from positions on one HLA allele to another HLA allele. This will let us compute whether features are in areas of variation in specific alleles even if we don't have direct information. This is what they have asked for. I imagine that you would do this portion of the work in consultation with, which would include doing the parsing, translation, and query development demonstrating the capability. Example of such a query: By comparing structures of Allele X binding to peptides of length 9, which contact points in Allele Y (assuming the IMGT alignment used to map to X) have different amino acids than X.

We need some mapping of disease to allele. There are implicit mappings already via Mesh -> paper, IMGT(allele) -> paper, our 'grep' of Medline (allele) -> paper. We can also use the OMIM phenotype annotations from http://www.human-phenotype-ontology.org/index.php/downloads.html in combination with your allele finder run over the omim records, unless we can find a source for that already. Finally we can also look at the GO annotations to see whether there are mappings to the Uniprot entries for different alleles to see whether there is biological material there (essentially: what are the Uniprot ids for the HLA alleles. How many annotations are there. Rerun the GO annotations to get the Uniprot associations too).

With the above we can do a demo query: Start with disease (or anything else we can), follow to allele. For that allele find perturbed contact positions or SFVTs. If more than one allele is associated with the disease, ask which peturbations are common to both.

For deliverable at end of this month we try to get them to have an installation of neurocommons with as much of this as we can accomplish (it's possible we can do all of this) and have them running whatever portions of queries leading to this as we have available.

Extra credit would be to figure out how to drive that browser that I send the pic of today and display the result as labels on such a picture.

There is other work in progress to continue, e.g. the databases of peptides that bind to particular HLA variants. This could be worked on in the extension.

-


 * Best display of contacts: http://molvis.sdsc.edu/fgij/fg.htm?mol=1K5N
 * http://www.pubmedcentral.nih.gov/articlerender.fcgi?tool=pubmed&pubmedid=15118103
 * http://en.wikipedia.org/wiki/File:L-tryptophan-skeletal.png
 * http://en.wikipedia.org/wiki/File:L-tyrosine-skeletal.png
 * http://www.genome.jp/dbget-bin/www_bget?pdb+1K5N+withATOM
 * http://search.cpan.org/~itub/Chemistry-File-PDB-0.22/PDB.pm
 * http://mail.python.org/pipermail/tutor/2004-October/032850.html
 * http://mail.python.org/pipermail/python-list/2008-July/671001.html