Homology-based inference sets the bar high for protein function prediction

BMC Bioinformatics

Table 1 Comparison of student methods.

	StudentA	StudentB	StudentC
Input Features	GO term counts	E-Values	GO term counts; percentage positives
Scoring Schemes	1	2	2
Propagation Types	maximum child	maximum child	maximum child; cumulative
Score Normalization Across Proteins	No	Yes	No

In this table, we have summarized the key differences between student methods. Input features include: the number of times a GO term appeared in the annotations of homologous proteins; the E-Values of the homologous proteins; and the percentage of 'positive' columns in their alignment matrices. Some groups used more than one way to score a GO term or differed during the propagation of a prediction by assigning a node the maximum value of its children or their sum. StudentB normalized the final score of a GO term to improve comparability among proteins.

ISSN: 1471-2105