Cookie Warning
This site uses cookies. Only cookies essential for site operation are used. These cookies prevent cross-site request forgery attacks, keep track of sessions and are deleted when the session ends.
DeepGOWeb SPARQL Endpoint
The SPARQL endpoint allows to call function prediction model in SPARQL query. We provide a custom function called "deepgo" which takes protein sequence and prediction threshold as an input and returns predicted functions along with the subontology, label and prediction score. The output can be downloaded in different formats such as json, xml, csv or text.
Two additional functions select the prediction model and expose its
internals (the original dg:deepgo function is unchanged):
dg:predict(?sequence ?threshold ?predictor)— likedg:deepgobut takes an extra predictor argument, e.g."dgpp-light"(the DeepGO-PlusPlus-Light ensemble) or"deepgoplus"(default).dg:components(?sequence ?threshold ?predictor)— returns each individual predictor substream (DIAMOND, STRING-Net, hierarchy-aware CNN, ESM2-kNN, ProteInfer) separately; the first variable binds the component name.
Example queries:
- Example 1: Simple DG++Light query
- Example 2: Federated query which runs deepgo on two sequences from UniProt SPARQL Endpoint
- Example 3: Run the
DeepGO-PlusPlus-Light model via
dg:predict - Example 4: Inspect the individual
predictor substreams via
dg:components