Run GO Annotation.
This is the process of selecting GO terms from the GO pool obtained by the Mapping step and assigning them to the query sequences.
GO annotation is carried out by applying an annotation rule (AR) on the found ontology term candidates. The rule seeks to find the most specific annotations with a certain level of reliability. This process is adjustable in specificity and stringency.
For each candidate GO an annotation score (AS) is computed. The AS is composed of two additive terms.
The first, direct term (DT), represents the highest hit similarity of this GO weighted by a factor corresponding to its EC.
The second term (AT) of the AS provides the possibility of abstraction. This is defined as an annotation to a parent node when several child nodes are present in the GO candidate collection. This term multiplies the number of total GOs unified at the node by a user-defined GO weight factor that controls the possibility and strength of abstraction. When GO weight is set to 0, no abstraction is done.
Finally, the AR selects the lowest term per branch that lies over a user-defined threshold. DT, AT and the AR terms are defined below.
To better understand how the annotation score works, the following reasoning can be done: When EC-weight is set to 1 for all ECs (no EC influence) and GO-weight equals zero (no abstraction), then the annotation score equals the maximum similarity value of the hits that have that GO term and the sequence will be annotated with that GO term if that score is above the given threshold provided. The situation when EC-weights are lower than 1 means that higher similarities are required to reach the threshold. If the GO-weight is different to 0 this means that the possibility is enabled that a parent node will reach the threshold while its various children nodes would not.
The annotation rule provides a general framework for annotation. The actual way annotation occurs depends on how the different parameters at the AS are set. These can be adjusted in the properties file.