Skip to main content
Figure 4 | BMC Bioinformatics

Figure 4

From: LS-NMF: A modified non-negative matrix factorization algorithm utilizing uncertainty estimates

Figure 4

Comparisons of Metagenes and MIPS Functional Classes. Evaluation of metagenes for LS-NMF, NMF, and NMF on scaled data, as well as expression profiles for unreduced data, was done by determining the ability of each approach to find functionally related genes based on the MIPS functional classification. In a, the fraction of correct classifications at level 2 in the MIPS ontology are shown in terms of increasing numbers of gene pairs (i.e., decreasing threshold for correlation). In b, the same information is shown for level 3 in the MIPS ontology. The performance of NMF can be improved by utilizing a divergence based update rule, however performance is still significantly below LS-NMF, with recovery ranging from 12 to 30 times greater with LS-NMF.

Back to article page