Normalization of cluster fluorescence intensities
This page describes what happens during the normalizeSeries script. See Fitting a binding series for a description on how this script fits into the overall pipeline and an example of how to run it.
This script will normalize cluster fluorescence by an all-cluster image (also called a red-channel image). Data from a CPseries is normalized by dividing cluster fluorescence by the corresponding fluorescence in another CPseries file of only one column, here referred to as the "red fluorescence values". To prevent values from being divided by really small numbers (and inflating the normalized signal towards infinity), bounds are applied to the fluorescence values. If the red fluorescence of a cluster is below the lower bound, it is set to the lower bound, and an analogous procedure is followed when the fluorescence exceeds the upper bound.
By default:
- lower bound = 1st percentile of the distribution
- upper bound = the median of the distribution + five times the standard deviation of the distribution
You may also set the flag --no_bounds to prevent bounds from being applied, or set the bounds with the --bounds input, i.e. --bounds 10 inf would set the lower bound at 10, but let the upper bound float to infinity.
Note: in practice I've never changed any of these parameters.
The distribution of fluorescence values per cluster in the red channel. The dashed vertical lines represent the lower and upperbounds applied to the distribution before normalizing. The solid green line is the KDE estimate of the probability density of the distribution, and isn't really important.