18th Dec 2023: Data Exploration
This page describes some exploration of the data- looking at 1D profiles of fluorescence in each well, Fv/Fm for each sample, and finally revisiting the number of empty plates.
- The following plots show a 1D line plot of the initial fluorescence intensity, extracted from a horizontal line centered over the peak of each well. They are grouped together according to the plate, so each line is a line plot of the fluorescence of a single mutant.
- Each plot shows all six treatment conditions for a plate, but since this is the initial fluorescence, the treatment conditions should not have an effect so we would expect all 6 plots to look identical.
- We would also expect all lines on the WT plate (99) to be identical. There does appear to be less variation on this plate compared to others.
- This is also handy to help understand our thresholding method
These measurements were obtained using the original thresholding method (mean + 3 * sigma) with the addition of a morphological opening step to smooth the mask. Fluorescence intensity was then averaged over the mask region. I have generated a large plot for each plate, which will take some explanation.
- The top scatter plot shows the Fv/Fm value for each well in the plate (flattened to a list of 1, 2, ..., 384 wells) colored according to measurement number (M1, M2, ..., M6).
- The middle bar plot shows the standard deviation of the Fv/Fm values for each sample. In other words, how varied this value is for each sample. Ideally it should be zero?
- The lower plot is the least obvious. This is the trend in Fv/Fm for each sample across the six measurements. This is obtained using a linear regression to the six measurements when they are ordered. The height of the bar for a given sample indicates the change in Fv/Fm as we go from MX to M(X+1). Again, in an ideal world these would all be zero, since Fv/Fm should not change across measurements of the same plate with different growth conditions.
Some initial observations:
- In some cases there is a clear trend in Fv/Fm with measurement #. For example, with the WT plate, M6 has a consistently higher value than all other plates. Which correlates with a problem we had with the WT cells (there was a problem whith the cells going through the pipeline, S6 plate was punctured and cells where thrown all around the plate prior to replating M6), so there is a high likeliness of M6 cells being in a different physiological state as the others
This animation loops over every plate, and plots a combined intensity histogram from all 384 wells on that plate. In other words, after splitting the original image into 384 small well images, we then compute the intensity histogram (using all 88/168 time points), and overlay all these histograms on the same axis.
This is good to sanity check the thresholding and check for anomalous data. You can also nicely observe the separation between the blank wells and the non-blank wells. The individual plots are also shown after the animation for completeness.
