Batch Mode

Specify Protein

Single Mode

Specify Proteins


Summary of interactions found in at % cutoff.

  • interactions.
  • genes.
  • Highest degree protein: with interactions.


ENTS, Elucidating Network Topology with Sequence, is a random forest-based program for the prediction of whole-genome interactions in a variety of species using only information derived from primary sequence data.

ENTS predicts physical protein-protein interactions using pairwise combinations of PFAM domains and predicted subcellular localization information supplied by the MultiLoc2 program. The ENTS webserver provides the set of predicted interactions and associated random forest confidence scores for a variety of plants, animals, and fungi.

Using the ENTS Webserver

PPI Window PPI Table

The ENTS webserver intends to provide sets of predicted interactions for model and non-model organisms across the eukaryotic domain. You can perform searches for specific proteins using the search box found at the top of each page or browse the proteins of a particular organism. The default protein names in the ENTS database conform to those found in Ensembl. However, you may also search using the Uniprot, RefSeq, GeneID identifiers. We also allow searches using the organism-specific database databases (i.e. TAIR for Arabidopsis thaliana or SGD for Saccharomyces cerevisiae).

Each protein brings up a page containing aliases, PFAM domains, and GO categories for the query protein and its interactors. Nodes are sorted according to the random forest confidence scores. You may click on links to find more information on each item, and you also have the option of checking for interaction evidence in the public PPI databases. There is also a pictoral representation of the interactions, which shows the top 25 interactors by default. You may customize the look of the graph by dragging the nodes and using the control panel on the right side of the page. Protein nodes are drawn with sizes proportional to their numbers of interactions, and edges are drawn in proportion to their confidence scores. You can mouse over the nodes to discover more information, single-click to highlight the node and its corresponding row in the interaction table, and double-click to get the full information on that protein.

You can also search a group of proteins for predicted interactions within that group using the batch mode. This also brings up a graph containing all predicted interactions between proteins in the batch. You can click on edges to highlight them and their corresponding rows in the interaction table in addition to peforming the functions described in the single-protein query graph.

Meaning of Confidence Scores

Each interaction is given a confidence score, which is assigned by the random forest predictor. Random forests work by constructing a large number of decision trees (in this case 400), each of which is somewhat different due to random sampling of the data and the predictors. The confidence score corresponds to the relative frequency of "yes" votes in the forest. Thus, a confidence score of 0.65 indicates that 65% of the trees in the forest predicted an interaction between the two proteins.


Initial protein interaction sets added
July 1, 2012
We have added interaction datasets for Arabidopsis thaliana, Populus trichocarpa, and Saccharomyces cerevisiae. We will soon add interaction datasets for Homo sapiens and Mus musculus.

Major updates
July 22, 2012
We have added the interaction datasets for Homo sapiens and Mus musculus. We have also made major improvements to the graph interface.

Design by