This tool prioritizes new genes in a pathway or process through the use of network guilt-by-association.
All genes connected to a query gene set in HumanNet are scored and reported. The quality of the predictions (essentially measuring the local clustering of the query genes in the network, which corresponds roughly to their modularity) is measured and reported as the area under a receiver operating characteristic (ROC) curve (AUC). AUC values range from near 0.5 (for random levels of modularity, and hence no predictability of new pathway members) to 1 (for a maximally coherent query gene set, and thus much better predictions of associated genes).
In general, higher AUC values (e.g., > 0.65) are considered more predictive; the higher, the better. So, high AUC values for a set of query genes suggests that their top-scoring neighbors are good candidates for follow-up.
Use this link for query gene sets smaller than 250 members, for which images of the local network are also provided. For larger genes sets, use the link below.
Same as above except allowing up to 2,000 query genes, but no graph layout.
This tool prioritizes new members of a pathway or process using the alternate Gaussian smoothing algorithm. This also allows queries of up to 2,000 genes, but no graph layout.
This predicts Gene Ontology (GO) biological process terms for the query genes based on their HumanNet gene neighbors.
Click for a list of the evidence supporting each interaction.