. Author manuscript; available in PMC: 2014 Oct 15.

Published in final edited form as: Conf Proc IEEE Eng Med Biol Soc. 2011;2011:6975–6978. doi: 10.1109/IEMBS.2011.6091763

Table 1.

PCA and data quality evaluation (simulation)

Table 2a. PCA-correction
Designed feature	GC-percentage		Date and scanner
Identified Component	1^st		2^nd
P-value	<1E-23		<1E-23
Table 2b. Evaluation of data quality
Data Quality	High noise		Low noise
	σLRR	N_{sub_ex}	σLRR	N_{sub_ex}
Uncorrected	0.30±0.03	76	0.25±0.03	10
Corrected (Comp. 1)	0.28±0.02	46	0.23±0.02	4
Corrected (Comp. 1,2)	0.28±0.02	40	0.22±0.02	4
Table 2c. Detection Accuracy: PCA-correction
Total generated markers with CNVs: 75867
PennCNV results	Overall FPR		Overall FNR
Uncorrected	0.6220		0.1374
Corrected (comp. 1)	0.0389		0.0940
Corrected (comp. 1,2)	0.0351		0.0886
Table 2d. Detection Accuracy: regression-based correction
PennCNV results	Overall FPR		Overall FNR
GC-percentage corrected	0.0389		0.0944

Note: high/low noise: group with high-SD/low-SD Gaussian noise, each containing 100 samples; σ_LRR: overall standard deviation of the simulated LRR data for each sample; N_{sub_ex}: number of bad samples failed by quality control; FPR and FNR are calculated with regard to the total number of markers with CNVs.