Comments on: Protein Function: how do we know that we know what we know?

By: Iddo

Iddo — Mon, 26 Jul 2010 19:40:48 +0000

@shwu
Well, the most obvious is ligand-binding assays can detect multiple affinities (at different binding constants) of different ligands. This does not necessarily mean that all the ligands bound are physiological, and sometimes that does not even matter e.g. in the case of testing for a good pharmaceutical binder. So we may characterize a protein as “ATP binder” but also as “caffeine binder” (per example).

Taking it one step further, high throughput catalytic assays can check for catalysis of said ligands, at different Kcat constants. Again, checking only for an enzymatic aspect of function, without previous hypotheses.

But the same assays cannot be used to, say, test for physiological functions. Y2h or a protein array be used for discovering protein-protein interactions. This can inform us of the physiological aspects — again, no strict hypotheses, but usually no deep conclusions either.

By: shwu

shwu — Mon, 26 Jul 2010 19:26:57 +0000

cement_head brings up an interesting point. Computational methods can detect multiple functions in a protein in parallel, while experimental verification of function would presumably be done serially (if multiple experiments are even done, once an initial function is discovered) and so the additional functions would take much longer to be identified. We have to start assessing somewhere, but some “wrong” predictions may simply be “ahead of their time”.

Are there ways people use to probe protein function experimentally that allow for identification of multiple functions in a single set of experiments (without strict hypotheses about what those functions might be)?

By: Iddo

Iddo — Sat, 24 Jul 2010 19:57:23 +0000

@cement_head
We are using GO as the annotation standard. If a protein is found to have more than one GO term associated with it, then predicting the a subset or the full complement of GO terms will raise the prediction score. The rules are available here: http://biofunctionprediction.org/node/262

By: cement_head

cement_head — Sat, 24 Jul 2010 13:25:24 +0000

Are there any plans to deal with the multifunction aspect of proteins? I’m thinking in terms of the “experimentally” verified?

Looks like a really good effort.

By: widdowquinn

widdowquinn — Fri, 23 Jul 2010 10:02:46 +0000

Fantastic! Exactly what’s needed.

By: Farhat

Farhat — Fri, 23 Jul 2010 07:46:32 +0000

Shouldn’t that be an ‘exponentially increasing gap’ in the figure caption?