Firstly, we’ve redesigned the FuncNet site, which you’ve probably noticed, unless you’re reading this via RSS. Hope you like it.

Secondly, we’ve added a new suite of predictors using a method called GOSS. This is a GO-based semantic similarity measure, like engineDB, but allows individual queries by the three GO subtrees: biological process (BP), cellular component (CC) and molecular function (MF).

Thirdly, we’ve expanded the section of the site dedicated to human spindle protein prediction, with a new prediction service called DORA-Spindle (DORA = Domain Over-Representation Analysis). This spots proteins containing domains which also appear in known spindle proteins. The FuncNet-SpindleP integration package has now been replaced with a new version called just FuncNet-Spindle, which queries FuncNet, SpindleP and DORA-Spindle in parallel and integrates the results. See the spindle prediction page for more details.

We hope to bring them back online by around 10:00 tomorrow morning.

Apologies for any inconvenience.

The new service, iHOP-FuncNetConn, uses unweighted co-occurrences of gene and protein names to estimate the probability of a genuine connection between each pair found in the literature. This gives it broader coverage of the genome, but the scores it produces are slightly less reliable.

The original service, iHOP-FuncNet, is still available. This uses weighted connections between names found in the literature, calculated in advance over the whole human/mouse co-occurrence network. As a result, the scores are more accurate but the coverage is lower.

PIPs is a database of predicted human protein-protein interactions. The predictions have been made using a naïve Bayesian classifier to calculate a Score of interaction. There are 37606 interactions with a Score ≥1 indicating that the interaction is more likely to occur than not to occur.

PIPs is an interesting addition to FuncNet as both PIPs and FuncNet itself aggregate a variety of sources of evidence, but using different methods. We will be performing some analyses over the next few weeks to quantify the benefits this kind of hierarchical aggregation of classifiers can bring.

Thanks to Mark McDowall for developing the PIPs service implementation.