Alex Palazzo is an Assistant Professor in the Department of Biochemistry at The University of Toronto.
How come no one told me about this site?
STRING is a database of known and predicted protein-protein interactions. The interactions include direct (physical) and indirect (functional) associations; they are derived from four sources:
-Genomic Context
-High-throughput Experiments
-(Conserved) Coexpression
-Previous Knowledge
STRING quantitatively integrates interaction data from these sources for a large number of organisms, and transfers information between these organisms where applicable. The database currently contains 1,513,782 proteins in 373 species.
Check this out. An interaction map:
A grading of conservation:
Go and play with it now.
[HT: Joolya]
Comments
Great!! There used to be another, older version of this on EMBL but it was more reliant on text matching. This version is far advanced of the previous site... already learned something too. So why are metastatic cancer genes always associated with synaptic plasticity genes? Evolution -- a neat surprise at every turn.
Posted by: Theodore Price | April 9, 2007 10:05 AM
Cool site.
Posted by: Acme Scientist | April 9, 2007 3:58 PM
Well maybe.
How much should we trust "interactions" derived from text mining of papers? Frankly, even a lot of two hybrid data is of dubious biological relevance.
Sometimes more is less.
Posted by: JK | April 11, 2007 5:21 AM
Sometimes more is indeed less - or perhaps rather less is more. I would thus like to point out two features of STRING:
1) All interactions have confidence scores associated with them. The typical interaction from a high-throughput yeast two-hybrid experiment would have a lower confidence score than one from a small-scale study. Similarly, interactions derived by automated literature mining will receive a very low confidence score unless it is supported by multiple papers. Low-confidence interactions can easily be eliminated from an interaction network by specifying a cutoff for the confidence score.
2) It is possible to entirely turn off particular types of evidence. If, for example, text-mining results are unwanted (even those of high confidence) then one can simply turn of the text-mining channel.
Posted by: Lars Juhl Jensen | April 11, 2007 11:31 AM
This was really helpful and interesting.
Posted by: bala | April 27, 2007 10:04 AM