Protein-protein interaction databases

Name	Description	Link
String	database of experimental and predicted interactions; excellent graphics	http://string-db.org/
IntAct	derived from literature curation or direct user submissions; tables	http://www.ebi.ac.uk/intact/
iHOP	= information hyperlinked over proteins; built by literature mining; results as text clippings	http://www.ihop-net.org/
BioGRID	from literature & high-throughput datasets; best for yeast & A. thaliana	http://thebiogrid.org/
MIPS	Mammalian Protein-Protein Interaction Database (didn't work when tested in Sept '11)	http://mips.helmholtz-muenchen.de/proj/ppi/
add another	database here with description here	link

String

• collects evidence from genomic Context, high-throughput experiments, conserved coexpression, and text mining
• created by CPR, EMBL, SIB, KU, TUD, and UZH
• latest references: [1], [2]

Opinion: easy to use, great graphical interaction map, detailed explanation of evidence

IntAct

• data from the literature or from direct data depositions by expert curators
• <300k binary interactions in 2011
• developed by the EBI's Proteomics Services Team
• recent reference: Aranda et al. '10

Opinion: The default is an oversized results table (>2 screen widths on a 13"). Use "Change table presets" to compact. Separate columns for links/alias/pubID could be integrated or hover over to make the default table easier on the eye. Overall, the presentation of results seems to be more geared towards bioinformaticians than users with the table full of codes instead of names (Do you know what 9606 is?). Also, no easy species selection and, more importantly, no graphical overview.

iHOP

• genes and proteins converted into hyperlinks between sentences from PubMed & turned into a conceptual network
• originally developed at CNB-CSIC, authors now at CNIO & MIT
• references: Hoffmann & Valencia 2004 Fernández et al. 2007

Opinion: All information is presented as highlighted text in sentences which makes the information slower to take in than from a structured table and much slower than from a graphical representation. Hypothetical and possibly indirect interactions (A regulates B), non-existent interactions (A does not bind B), and experimental (A co-precipitates with B) are not differentiated. Easy to use, clean species separation.

BioGRID

• >400k interactions curated from both high-throughput datasets and primary literature
• mostly developed by researchers from Edinburgh, Princeton, and Toronto [3]
• recent references: The BioGRID Interaction Database: 2011 update, A global protein kinase and phosphatase interaction network in yeast. Breitkreutz et al. '10

Opinion: Good interface but generally fewer hits than other databases for mammalian genes. This is probably because this database contains mainly yeast protein interactions where it is very strong. It is currently being extended towards other metazoans. Nice representation of evidence as high & low throughput with reference and description.

See also

On OpenWetWare

• The Prince lab present on OWW works on protein-protein interactions.
• OWW's Keating lab also works on protein-protein interactions.

Lists of protein interaction resources:

• huge collection of protein interaction resources at the FLI
• interaction database list by the Finley lab
• MIPS (scroll down to other resources)

Related literature searches

• search PubMed for MeSH term Protein Interaction Mapping [4], major topic [5], and reviews only [6]

Recent reviews:

• Databases of protein-protein interactions and complexes. Ooi et al. 2010
• Computational prediction of protein-protein interactions. Skrabanek et al. 2008
• ..the Human Proteotheque Initiative (HuPI). Coulombe et al. 2008

Wikipedia entry protein interaction, interaction prediction