Average:
Averaging across the biological replicates
After normalisation, we first calculate 'M' (similar to a log-ratio) for each spot on each slide. This allows partial self-normalization of spatial effects, e.g., variations in hybridisation efficiency across the slide surface. We then calculate an uncertainty of 'M' from the 'pixel intensity fluctuations' reported by dapple. This assumes a flat disc spot model. Then, seperarately for 'normal' and 'dye-swapped' slides, we calculate a weghted average, using the certainties of 'M' as weights. Lastly, we form a simple average of 'M' for 'normal' and 'dye-swapped' slides. The average is unweighted to allow partial compensation of dye-swap effects.
Normalisation and averaging of an entire replica group
Key
M = channel 2 - channel 1
A = (channel 2 + channel 1) / 2
