Missing value imputation improves clustering and interpretation of gene expression microarray data

BMC Bioinformatics

Table 1 Datasets.

Name	N	M	M _C	M _F	MV _SD	MV	Type	PC1
Brauer05	19	6256	3924	3066	4.0	6.7%	MT	54.9%
Ronen05	26	7070	4916	2695	3.2	3.8%	MT	51.1%
Spahira04A	23	4771	2970	2090	3.9	2.7%	TS	62.0%
Spahira04B	14	4771	3340	2898	4.2	3.0%	TS	54.1%
Hirao03	8	6229	5913	259	0.7	0.9%	SS	43.3%
Yoshimoto02	24	6102	4379	2323	1.9	3.2%	MT	64.7%
Wyrick99	7	6180	6169	3600	0.0	0.0%	TS	61.3%
Spellman98E	14	6075	5766	1094	0.4	0.4%	TS	39.9%

N is the number of measurements (columns in the observation matrix), M is the number of genes (rows), M_Cis the number of genes without missing values in the complete dataset, M_Fis the number of genes after applying the filtering, MV_SDis the standard deviation of the missing value distribution over the measurements, MV is the percentage of missing values, Type indicates whether the dataset is a time series (TS), steady state (SS), or mixed type (MT), i.e., multiple time series measured under different experimental conditions, and PC1 gives the proportion of total variance explained by the first principal component (high value indicates high correlation structure between the genes [31]).

ISSN: 1471-2105