BMC Bioinformatics

Table 4 The performance of NHMC and competitive methods in predicting gene function on weakly connected genes

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

	15 connections				5 connections
Dataset	CLUS-HMC	NHMC	FF	H	CLUS-HMC	NHMC	FF	H
seq	0.014	0.014	0.001	0.001	0.033	0.042	0.007	0.007
pheno	0.018	0.051	0.001	0.001	0.033	0.027	0.005	0.007
struc	0.012	0.078	0.001	0.001	0.093	0.093	0.000	0.007
homo	0.012	0.023	0.001	0.001	0.149	0.149	0.003	0.007
cellcycle	0.015	0.015	0.001	0.001	0.041	0.023	0.007	0.007
church	0.013	0.025	0.001	0.001	0.031	0.022	0.007	0.007
derisi	0.015	0.015	0.000	0.001	0.024	0.026	0.007	0.007
eisen	0.020	0.020	0.000	0.001	0.039	0.040	0.007	0.002
gasch1	0.015	0.015	0.001	0.001	0.023	0.025	0.007	0.006
gasch2	0.018	0.023	0.001	0.001	0.028	0.028	0.007	0.007
spo	0.015	0.015	0.000	0.001	0.022	0.022	0.007	0.007
exp	0.015	0.015	0.001	0.001	0.026	0.044	0.007	0.003
Average:	0.015	0.026	0.001	0.001	0.045	0.045	0.006	0.006

We report the $\bar{AUPRC}$ of the CLUS-HMC (α = 1), NHMC (α = 0.5), FunctionalFlow (FF), and Hopfield (H) methods, when predicting gene function in yeast, using GO annotations and the BioGRID PPI network. The models are trained on the subset of highly connected genes and tested on the subset of weakly connected genes.

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com