Consistent probabilistic outputs for protein function prediction

Table 2 Kernel transformations

Name	Linear	Normalized linear	Linear²	Normalized linear²	RBF	Diffusion	Parameters
MGI	✓	✓
OMIM	✓	✓			✓		σ = 1
Inpar	✓	✓			✓		σ = 1
Biomart	✓	✓			✓		σ = 1
Inter	✓	✓			✓		σ = 1
PfamA	✓	✓			✓		σ = 1
PPI	✓	✓				✓✓✓	τ ∈ {0.1, 1, 10}
Su		✓
Zhang	✓	✓	✓	✓	✓		σ = 1
SAGE	✓	✓			✓		σ = 1

Each row in the table corresponds to one of the ten data types listed in Table 1. Check marks indicate which kernels were computed for each data set. The kernels are described in the Kernels section in Materials and methods. 'Linear²' and 'Normalized linear²' refer to the squared version of the original kernel matrix, as described in Materials and methods. The three check marks for the PPI data indicate that three diffusion kernels were computed using τ ∈ {0.1, 1, 10}. RBF, radial basis function.

ISSN: 1474-760X