Protein surface representation and analysis by dimension reduction

Table 2 Performance of dimension reduction methods for mapping a section of the heat shock protein ATPase domain

Methods	Area	Neighbor	Runtime (sec)
PCA	0.69	0.17	0.002
LLE	0.66	0.13	27.1
Laplacian	0.48	0.22	7.57
LLC	0.19	0.07	13.4
AutoEncoderEA	0.52	0.05	73.2
SNE	0.69	0.24	8824
SymSNE	-0.0012	0.002	7733
CFA	0.69	0.08	421.3
GPLVM	0.69	0.17	4477
NPE	0.70	0.13	23.4
LPP	0.69	0.13	20.56
LLTSA	0.69	0.17	13.8
NCA	0.69	0.08	178.8
MCML	0.69	0.16	1337
LDA	0.69	0.14	0.46
FactorAnalysis	0.69	0.16	0.02
tSNE	0.12	0.36	2295
Isomap	0.70	0.23	8396
LandmarkIsomap	0.70	0.23	1739
ProbPCA	0.68	0.01	10.18
KernelPCA	0.03	0.08	679.8
MDS	0.69	0.17	0.002
DiffusionMaps	0.69	0.18	4046
Sammon	0.69	0.18	3294
Sinusoidalcartography	0.10	0.2	0.005

For each performance criteria, the top comparable scores are shown in bold. knn = 3 was used for the neighbor scoring.

ISSN: 1477-5956