Ensemble analyses improve signatures of tumour hypoxia and reveal inter-platform differences

BMC Bioinformatics

Table 2 Significant coefficients of linear model for prognostics based on individual gene

Coefficient	Estimate	Standard error	t value	Pr (> \| t \| )
(Intercept)	1995.2	251.9	7.736	1.57×10⁻⁸
Handling, separate	−1305.8	313.1	−4.171	2.51×10⁻⁴
Platform, HGU133 Plus 2.0: Handling, separate	3079.2	236.7	13.010	1.24×10⁻¹³
Platform, HGU133 Plus 2.0: Algorithm, log₂ MAS5	−1844.8	409.9	−4.500	1.02×10⁻⁴
Platform, HGU133 Plus 2.0: Algorithm, MAS5	−1822.2	409.9	−4.445	1.18×10⁻⁴
Handling, separate: Algorithm, log₂ MAS5	−1124.2	409.9	−2.743	1.03×10⁻²
Handling, separate: Algorithm, MAS5	−1132.8	409.9	−1.461	9.83×10⁻³
Handling, separate: Algorithm, RMA	−993.0	409.9	−2.422	2.18×10⁻²

For the linear model, $Y = W + X + \sum_{i = 1}^{5} Z_{i} + W : X + \sum_{i = 1}^{5} (W : Z_{i} + X : Z_{i})$ where Y is the number of genes, W is the platform, X is the data handling and Z₁…Z₅ are specify the 6 options for the pre-processing algorithm, the coefficients that have a p < 0.05 are shown.

ISSN: 1471-2105