A review from 2 years ago (though probably still considered recent) offers ~40 tools for function prediction categorized by different approaches.

Trends Biotechnol. 2009 Apr;27(4):210-9. Epub 2009 Feb 27. Protein function prediction--the power of multiplicity. Rentzsch R, Orengo CA.

Also, join the coming Automated Function Prediction SIG2011 at ISMB 2011 to discover the bleeding edge of computational function prediction.

I would use a webservice that wraps a whole bunch of predictors. A good one is e.g. http://www.predictprotein.org which is developed in our group. It is sequence based and returns a variety of structure- and function predictions.

I would say:

1. a PFAM search
2. Blast against Swiss-prot

There is similar thread.

I'm happy with InterPro scan. It integrates multiple databases: PROSITE, PRINTS, Pfam, ProDom, SMART, TIGRFAMs, PIR superfamily, SUPERFAMILY, Gene3D, PANTHER and HAMAP. This means, often you will get functional annotation from multiple, independent sources. In addition, another tools like SignalP can be optionally launched by InterPro. But it's quite computationally extensive - you will need a cluster to annotate whole genome.

There is a webtool named ProtoNet: http://www.protonet.cs.huji.ac.il/. ProtoNet provides automatic hierarchical classification of proteins according to sequences. So you can get many kind of annotations of related proteins. Try the "Classify Your Protein" tool within ProtoNet.

Well, that clearly depends on what you define as protein function. Enzymes-only? Every protein that binds something? And what is that something? Small ligands? RNA? DNA? Other proteins?

Chris

well in my case I like to use the entire xl pharmacy products line, this people offer a so extensive variety of reactive sustance to work in a lab, in fact in this moment I worked in a personal project to discover what kind of proteins causes necrosis effect.