Skip to main content
Figure 5 | BMC Genomics

Figure 5

From: Autocorrelation analysis reveals widespread spatial biases in microarray experiments

Figure 5

Quantification of the dependence of autocorrelation signals on the size and magnitude of horizontally-shaped biases. Horizontally-shaped biases (similar to those in Figure 4 lanes 5-10) of varying sizes and magnitudes of bias were introduced and the resulting autocorrelation quantified. The colorscale denotes the number of autocorrelation data points with correlation r values >0.05 in the first 200 gene distance runs. Using the same threshold as that used for evaluation of the extent of periodic autocorrelation in the real data (Figure 3F and see text), i.e. >20 signals points complying to the above criterion, it can be concluded that biases that cover more than 15% of the microarray surface and that contain at least a two-fold ratio signal are responsible for the autocorrelations observed in the real data.

Back to article page