The below figure is created from two manual build PDB models: the large membrane floor and the “dinosaurs” or rather the MBL oligomer of the lectin pathway. Instead of using the standard PyMol surface representation i opted to use the Isosurface function which i find easier to customize to my likings and obtaining the look I wanted. The cell membrane is based on an small atomic model of the core LPS layer from Pseudomonas Aeruginosa, which is duplicated a bunch of times before translating the copies in the X and Y directions, creating a seemingly single combined membrane.
The MBL models with their extended MASP molecules somewhat resembles some of the giant dinosaurs which is why I call the figure “The dinosaures of the complement system“.
The below figure is basically a top-down view of a combined model of two of the MBL molecules from the top figure, showing how they may interact with each other. The membrane is the same as in the top figure, only smaller.
The below figure is a recreation of a figure published by another researcher. The emphasis was meant to be on the two MASP molecules which are bend into the cone defined by the legs of MBL molecule, therefore they were given vibrant colors compared to the faint gray color of the MBL molecule.
Figures from my PhD thesis
While writing my PhD thesis I produced a ton of different figures, many of which was never used, because I was constantly pushing the boundaries for the quality and style I wanted in my thesis. I also had a very specific aim with my figures, as I wanted all of the models within a figure, to be to scale relative to each other. I was sick and tired of seeing high as well as low impact scientific papers, getting the scale completely wrong. Many reviews would e.g. completely misrepresent the scale of interacting proteins in their introductory overview figures either due to space issues or because they never bothered to look up the actual size of the proteins in question. So one misrepresentation could propagate down the following generation of papers because no one bothered to check or think about it etc.
Thus I was highly motivated to not commit the same mistake in my thesis, although the number of ppl who would actually read my thesis could probably be tallied on two hands. I created atomic models of all the known proteins in the human complement system, carefully mirroring the actual size and domain architecture (if known) of each. For many models, where no crystal structure was known, I was forced to create composite domain models using similar and sizable domains from known unrelated structures and combine them based on published characterizations and structural studies. Sometimes I simply just had to guess a bit, as was the case with my favourite of the models, the C4 binding protein:
When it came to combining the single models within a figure, I was very alert as to not introduce any deliberate scale bias at this point in the process.
Unfortunately I had to compromise slightly with the scale in the classical pathway figure concerning the antibodies, due to time constraints. I discovered the mistake at the day of my PhD defence.
As is I am sure is evident from the above figures, I was (and still very much am) quite inspired by David Goodsells “Molecule of the Month” look. The Goodsell look can be obtained quite easily using the program QuteMol (however you cant do anything else with the program) or by using the Goodsell-like PyMol script from the PymolWiki gallery page. I used the latter option, although in a slightly modified version. IMO the Goodsell look is at its best when you move a bit closer than in the above figures. Below I compare the domain and super-domain architecture of some of the related complement proteins using the Goodsell look.
In every proteins crystallographic PhD thesis it is mandatory to show some electron density, and I of course also made quite a few electron density figures. I guess what sets my figures apart from the more classical electron density figures you see, is primarily that I took the time to use the map double setting in PyMol twice per rendering. This setting resamples your map at twice the resolution each time it is used, and you can use it as many times you want! Unfortunately it comes at an eightfold increase in memory use each time it is used, which logically slows down the process quite significantly and naturally the upper limit for the number of times it can be used is reached quite fast. However I think the end result is worth the extra processing time.
I also did a lot of electrostatic surface representations to visualize the electrostatic interaction properties of different interacting domains and proteins. This was done using the brilliant APBS plugin for PyMol as opposed to using the “generate vacuum electrostatic” function, which basically only performs a charge smoothing. APBS takes a bit of tinkering to use efficiently, but it is definitely worth it.
I only did a couple of schematic figures for my thesis but I actually enjoyed making them as simple as possible while still containing the information I wanted to convey . The next two figures were made in something as simple as powerpoint. They is nothing fancy about them, just simplicity. They are also the two biggest errors in my thesis concerning my figures. I had from the start planed to color code the schematic figures with the structural figures, but for some reason it completely slipped my mind. Instead I color coded them to show which gene each chain originates from, which only makes sense in the MASP figure. But no one noticed so fuck it.
The below figure took forever to make because I first had to learn how to use a new (for me that is) program called ChemSketch. The program is pretty good at drawing chemical structures in a very fast and correct way. However it has a bit weird learning curve, but I got there. I used it for showing the catalytic mechanism of a serine proteinase.
My first attempt of a scientific cover
The color palette was chosen based on the color scheme PNAS used this “season” (green) exemplified by this mockup of a final cover I did to test how the final product could look like (NB this is not the final version of the actual design):
Without going into boring details about the project, the major finding of the paper could be summarized with the word clustering. So I tried to incorporate that into the design (all the light spheres clustering together, plus the molecules clustering together on bacteria surfaces etc.).
Also I didnt want to just use a classic PyMol/Chimera/Qutemol rendered ray-trace image of the molecule in question, as that had been done soooo many times before (and it rarely works). Instead I wanted a conceptual piece with a minimalistic and silhouette-like aesthetic containing no highly detailed three dimensional models, a bit like the visual design from the game limbo.