Skip to main content

Table 6 The performance of NHMC in predicting gene function in combination with feature selection

From: Using PPI network autocorrelation in hierarchical multi-label classification trees for gene function prediction

 

Dataset

 

Struc

Homo

Feature set

No. of features

AUPRC ¯

No. of features

AUPRC ¯

All the features

19624

0.012

47034

0.012

Top 15%

2944

0.0115

7055

0.012

Top 10%

1962

0.0115

4703

0.012

Top 5%

981

0.0115

2351

0.012

Top 1%

196

0.0115

470

0.0115

  1. AUPRC ¯ of NHMC (α = 0.5) with BioGRID PPI network for struc and homo datasets, when working on all the features, top 15% of the features, top 10% of the features, top 5% of the features and top 1% of the features. The datasets struc and homo are chosen because of their very high number of features as compared to the other datasets.