Previously counted plots included all that were available at the time (found in this link; which had about equal numbers of plots from each dodecamer (hexamer – trimer). The data on this page show just about every plot that I have ever made….. these are divided as to trimer…. so this is mean trimer peak number. The latter counts the N term peak with each trimer (even though it is shared between all the trimers…. each trimer it gets counted once.   Therefore a peak number of 8 per trimer would become peak number of 15 per hexamer owing to the N term peak being shared.  In terms of progress…. its best to assume that the more recent posts are the better data.  Every possible image and signal processing filters and algorithms are summed here. Some people counts, all counts….  this represents a huge N, in my estimation.

I think my current favorite image processing filter is still the gaussian blur, and my favorite peak counting program is PeakValleyDetectionTemplate.xlsx by Thomas O’Haver.

Peaks per hexamer were calculated four ways.  As every plot made for each of these four dodecamers. This includes hundreds of counts for one dodecamer, and between 26 and 50 hexamer plots for the other three. Certainly one carries more weight, or one would think, but the data using each of the plots separately is not different than using each of the methodologies (at an absolute minimum there were 2 image processing filters, and at least 5 signal processing algorithms for several of these image processed pictures.

Data are also given with each of the four dodecamers individually: (41_ak45; 42a_aka_44; 43; 97-1).  In addition. n, mean, sd, and other parameters were calculated from my original peak counts from just the “image”, as well as from my original peak counts from the plot generated in ImageJ.  This is in addition to the whole lot of plots subjected to signal processing.    Bottom line is that signal processing appears to increase the peak count in a significant way. Whether the signal processing is “better” (which i dont think it true” or counts from images is “better” remains to be sorted out.   Below is a comparison of the various “sorting” that I used to determine mean number of peaks per hexamer of SP-D.

Two left columns are: 1) Every plot of a hexamer separately, 2) Plots divided into each of the four dodecamers separated into groups; Two right columns are counts separated into  “image with signal processing” per dodecamer ( and separated again into, my counts of image processing plots only plus my counts of the peaks in plots made in ImageJ (that is… NO SIGNAL processing)  18+ peaks with signal processed plots, and 15+ peaks using my hand counts.  15 peaks per hexamer is in my bet for the best number.  See previous posts here.

As separate trimers, and as hexamers peak width summaries, regardless of how they are sorted seem to be consistent for these first four image and signal processed dodecamers of SP-D.  A model for further peak counting using these parameters for an intelligent signal processing approach might be interesting. The 8 peaks per trimer (15 peaks per hexamer as N is counted as one peak in each trimer and becomes the odd center peak in hexamers) widths are given below.

AFM images (from rhSP-D at pH 7.4 Arroyo et al) of SP-D are informative and numerous and during a careful analysis of them it seems pretty obvious that there are a significant number of occasions where there is a close “sticking” together of the strands of the trimer from the N termini junction, through the tiny peak on either side of the N peak and up to the glycosylation (often including it) peak(s)(plural here because there are a significant number of imaging and signal processing applications that count more than one peak in the area considered the glycosylation peak).
Here are two images (labels on each show the number of nm diameter, the length in nm of each of the hexamers, and arrows that show where the trimers from two hexamers are in close association. I scanned 83 images from the images of Arroyo et al, and found that such an association (which in order to be visualized with any confidence requires that the dodecamer to be lying such that the rest of the trimer-arms are separated, thus not just overlapping — overlapping trimer arms were not included in the count of closely associated N, tiny peak, and glycosylation peak associations.
I have also given a number (my reference number for the set of thumbnail images, and the set of dodecamers to which some measurements have been applied.
Any reason for such a close association between trimers in that specific location is not known by me, comments welcome. 100 nm bar is given for your “enjoyment”,  green nm measurements correspond to green segmented tracing through the center of the hexamer to which it refers, same for red segmented line and nm values.  Diameter values (orange) are made in ImageJ, and requires that three of the four carbohydrate recognition domains be touched.  It is easy to see that once past the glycosylation peaks, the arms of each trimer are separate. On the bottom image, the trimers are close on the top half of the micrograph, but not the bottom half, as the glycosylation peaks are separate. (42 dodecamers, of 86 total, show N, tiny peak, glycosylation peak closeness on one or both sides of the dodecamer.)

Same process  for determining peak width, and sub peak number…. as before. peak 6 i expected to show up a little wider (here it is about 7.5nm+0.12).  in previous plots that are colored one can see it as that peak(s) which are white.

summary plot from months ago,  below to remind one what the scheme of the bilateral symmetry was/is.

PEAKS PER TRIMER (trimer=about 74 nm including the full N term peak with each plot)
Peak 1=N term peak (full) (peach) @20nm
Peak 2=Tiny peak (purple) @2.5nm
Peak 3=glycosylation peak(s) light green @8nm
Peak 4= unnamed (darker green) @12nm
Peak 5= unnamed (pink) @4nm
Peak 6- unnamed (white) @7.5nm
Peak 7= Neck (yellow) @4.5nm
Peak 8=carbohydrate recognition domain peaks (orange) @ 17nm

This is a small peak by comparison to the glycosylation peak(s) and to the fourth peak just lateral (in a bilaterally symmetrical hexamer) which is @12nm in width, this very narrow and lower peak is only 4nm wide (at the valleys).  In previous ridge (Joy) plots it is colored in a bright pink just in case you want to look up its position and relative height and width.

After finding width and number of sub-peaks for each of the 7 peaks on either side of the N, i will make a diagram with the dimensions (width, height, and sub-peaks) found for each of the four dodecamers in this short list of test molecules.
There is a peak 5 in 74% of the trimers, less than 1 per trimer, and no sub-peaks noted.

Here is a peak series that I think occurrs with a frequency that legitimizes it as being thought of as a distinct entity.  The peak number in grayscale plots for SP-D hexamers has been calculated by numerous signal and image programs but the pattern itself is the result of observing countless images and the patterns within each hexamer mirrored as two trimers.  (N being central (peak 1), tiny peak lateral to N (peak 2), glycosylation peak being (3), and this peak (4) with unknown function or quality is consistently as wide, with a lower peak height than the glycosylation peak.  The divisions are mine.  It appears (looking at four dodecamers of SP-D (as individual trimers) to be about 11 nm in width.  It does show subpeaks just like the N and the glycosylation peak.  It would be amazing if the subpeaks were indicators of the “twisting” together at a subtle distance the three individual SP-D molecules.  Thats not too far fetched.  This conjecture might hold up until the CRD are reached which becomes rather random positioning of the globular part of the protein.

It helps to visualize this peak (deep green in the ridge plots link to one here). The “tiny peak” (purple) can be viewer towards the top of this ridge plot, but not at the base of the plot.

The number of sub-peaks (tabulated without bias within the five signal processing peak finding algorithms, and the hand counts are consistent with this peak being one for sure but maybe with a smaller element or subpeak. I am defining sub-peaks as smaller peaks which visually appear to be part of a single wider taller peak.

n=4 dodecamers, with many different signal and image processing filters, giving rise to about 114 different plots and peak numbers; the peak number summary of each dodecamer provides: n=4, 17.25 total peaks per hexamer+2.48. This is still just four molecules.

1) N term peak = @20nm which is divided into two sub peaks about 50% of the time.
2) Tiny peak of @3nm either side of the N term peak
3) Glycosylation peak @9nm, either side of N term peak, and consistently composed of two parts (two peaks)

Using just four SP-D molecules, with arms plotted as hexamers, and recorded as trimers (reversing the mirror symmetry and using the N term value as the whole N term peak width), the glycosylation peak (now called the 3rd set of peak(s) in the trimer) sometimes has a lumpy appearance, and signal processing programs routinely divide that single (sort of) largest of peaks (second to the N term and the CRD peaks) into separate adjacent but clearly contiguous peaks, sometimes as many as 6.  When i see the brightness areas i dont often agree on what the signal processing programs call “separate” peaks as they really come together in one single significant elevation, clumped as separate but connected peaks.  The number of separate peaks within what is typically pointed out as the glycosylation peak in AFM micrographs actually has more than one peak… these data sort out using 4 dodecamers (16 trimer plots – assessed by 5 signal processing programs, several imaging processing filters, and one unprocessed image collectively) that about 2 peaks show up in the glycosylation “bright” spot.   1.92+0.204.  This will likely change with the addition of more molecules added to the set.