Figure 1

A schematic illustration of prediction rationale applied in GUMP. (A) Overview of the data integration and the validation process. (B) GO structure and the definition of normal prediction and aggressive prediction. Normal Prediction seeks matches to less specific nodes (such as node c) whereas Aggressive Prediction seeks more specific nodes further down the graph (such as node b). Less specific nodes selected by Normal Prediction are generally more reliable since more data is associated with these nodes.