Six dodecamers: SP-D – tiny peak height (peak two, in each trimer, located (about half the time) on the down-slope of the N termini junction peak (peak 1) and just before the glycosylation peak (peak 6). This is detected by many of the signal processing programs, at this point it is an unreported peak in SP-D.  Means, SD, median, etc listed below for each individual dodecamer of this set of 6, and as a group of 6. MO is the same as reporting on other peak widths-heights-and valleys. First set of 4 dodecamers, 134.7+/19.5 (grayscale 0-255) and all data for that set  is  here.

Same procedure, N termini junction peak height for a selected data set (n=368 trimer plots), for the mean of the select 6 dodecamers (all plots each of the dodecamer images of SP-D). The grayscale values of the individual images, does make a difference in the peak heights…. whick at some point i will try to normalize. But for now, the N termini peak has always, and continues to be the highest peak in the plots, typically close to center (there are very rare exceptions which can be attributed to overlapping molecules in an image.

With an extensive number of plots of the first four dodecamers, the grayscale peak was 238.6 +/- 3.75, so the fourth and fifth dodecamers added did change this number slightly. Clearly the values are influenced by the contrast of each image.  I think a “limit range” function (which can be found in Gwyddion and likely other programs for image analysis) could be useful. I am more interested in a quick normalizing of peak height and the lowest point in the plot using a graphics program.

(I continue to thank Arroyo et al, and Thomas O’Haver and my two sons, Aaron and Dan for help with image and signal processing apps.

CRD peak is iconic,  present in all dodecamers i have seen except one (not digitized yet) where it obviously just got lopped off somehow during preparation.  Looking like it is going to be about 15nm in width which is close to previously measured peak widths for this domain.

Same MO, neck domain peak, sometimes visible, sometimes hidden under the floppy CRD domain.

Same format, peak 6, with and without the very few undetectable (0 values), width in nm, both as individual values, and means of the 6 different dodecamers.

Peak width for what I have described consistently as peak 5 is given below, with the means found in the current 6 dodecamers (as individual values). The top image shows the peak width even though there are missing values, all values are included in the means and other data. The graphic just below that has the “missing, or undetected” peak 5 values left out of each of the individual measurements.  BOTTOM LINE, this peak 5 is likely to be about 4 nm in width.

It is actually a little disappointing to see how little the widths change whether I delete the undetectable values or not, but it is also a good thing, so that I will continue to measure a molecule many times with different signal processing programs to come up with a stable number.  Each of the lower 6 graphics are the individual surfactant protein D dodecamers, just peak 5 evaluated, with and without the “missing, undetectable” peaks as separate evaluations. Each set of values for the dodecamers is given with (left) and without (right) the missing values.

This tiny peak (peak 2) in the number of estimated peaks in a trimer of SP-D is very close to the N term peak, almost on its downward slope and very thin, at least typically.  This is a peak that I have observed about half of the time so it is easily obscured, if it exists… which i am pretty sure it does.

Six dodecamers: SP-D – peaks numbers for trimers, hexamers, with and without signal processing — this data compare the counts of non-processed peaks along a hexamer, with counts of peaks along hexamers which have been 1) image processed or signal processed, or  2) both.

Each procedure i have maintained with its processing settings (for both image and signal processing). Unprocessed images record more peaks that are less likely to actually be meaningful because it picks up individual pixel values which are not always meaningful.

This seems not to be totally alleviated by the use of signal processing scripts. Some identify peaks that are almost too small to measure, worse than picking up pixels.  Since the difference between what is found with image processing (that would be counts from the original plots in imageJ, and my counts from the actual micrograph before any image processing has been made) and also before the image was subjected to any peak finding programs seems to end up a pretty small difference.

This was confirmed by sorting the two sets of data out into “image” and “signal processed” and I had hoped the peak number found in each plot would be greater in the signal processing apps.  Thus i could rationalize it as a way to avoid processing and recording a huge numbers of grayscale plots with signal processing, and rely more on what my own educated guess says about how many bright peaks there are in the four domains of the structure of surfactant protein D.

But there is little consistent difference between the two approaches much to my dismay, the numbers both ways are pretty close.  Maybe both, though are an over-estimate of peak numbers. My personal feeling is that earlier data (15 peaks per hexamer with the odd peak being central N termini junction peak in the hexamer molecule) was a good number, but these last observations are higher.

It is also, I have found personally true, that the settings (lag threshold, influence, smoothing etc) are all really individually set….  and while i chose to use the same settings on all micrographs of these SP-D dodecamers (out of focus, light, too much or too little contrast, torn and scratched, lying over debris or other molecules) I could have selected options that made them all fit the way I wanted….. this is disconcerting, as i hoped that signal processing would be the perfect unbiased count…

NOPE. This is a way i had hoped to elimate bias….  but if i change setting on signal processing, i feel it is “bias”,

Measures for the 6 individual dodecamers are below( numbers of peaks per each of the two hexamers are listed separately) and from one hexamer ( 80) i removed the hexamer that was suggested to be an extreme outlier using an online statistics calculator for skew. That said, virtually no difference in the mean was seen.

I sense that 15+/- 2  will be the number of consistently found peaks in a hexamer.

Found yesterday, 5 peaks per trimer (or 9 peaks per hexamer) are noted about 100% of the time.  There are three other peaks present at about 40 60  and 70% of the time that are still in question.

My counts from micrograph and my counts from the original grayscale pleak plots.