Skip to main content

Table 1 Dinucleotide model performance on different datasets

From: Background correction using dinucleotide affinities improves the performance of GCRMA

Data set

nc a

np b

 

Single nucleotide model (eq. 1)

Dinucleotide model (eq. 4)c

Latin Square [20]

42

248152

PM

0.17 ± 0.01

0.22 ± 0.01

  

248152

MM

0.40 ± 0.01

0.50 ± 0.01

Golden spikein [16]

6

195994

PM

0.20 ± 0.02

0.22 ± 0.02

  

195994

MM

0.46 ± 0.02

0.51 ± 0.02

Leukemia [35]

72

201800

PM

0.49 ± 0.06

0.55 ± 0.07

  

201800

MM

0.60 ± 0.04

0.69 ± 0.04

Etoposide response [34]

60

496468

PM

0.05 ± 0.04

0.08 ± 0.06

  

496468

MM

0.11 ± 0.06

0.16 ± 0.08

BK knockout [36, 37]

20

496468

PM

0.09 ± 0.04

0.13 ± 0.04

  

496468

MM

0.29 ± 0.050

0.36 ± 0.06

  1. R2 of Naef and Magnasco [15] model (Single nucleotide) and the dinucleotide model for the five data sets used in this study. Results presented as average R2 ± SD.
  2. anc: number of chips.
  3. bnp: number of probes.
  4. c The differences in R2 between single nucleotide model and dinucleotide model are all statistically significant (p < 10-3) using paired one-sided Wilcoxon and t tests.