"Programmed disassembly of an Interplanar Amida Network coordinates 3D epithelial morphogenesis"
Artistic renditions of scientific data
This was a fascinatingly multi-faceted project that lasted over a year. The tasks definitely put me out of my comfort zone, requiring to learn many new tools and develop processes for generating very specific visuals. Also I had never worked in such a culturally diverse team of proficient and accomplished scientists.
I worked part-time in University of Tartu, Institute of Molecular and Cell Biology, official position was engineer. I worked with the core team in Tartu, but also many other people form different universities were involved in the making of the research paper.
The challenge was two-fold:
Working with highly skilled scientists that are nevertheless specialised in a very narrow field, so they had little knowledge about how and what I could do;
Trying to figure out which tools to use and develop a workflow with them that would meet all the requirements (that were often changing)
From the start I was given many different descriptions for graphics that had to be generated. Based on those assignments I determined that the ideal system would be able to generate visualisations of cells that have to smoothly merge and separate, not just clip through but still have the ability to precisely determine their shape and also simulate them at large scales while interacting with particle systems. Also there was no budget to buy additional tools so I had to work with what I got already.
For ideation, meeting notes, illustrations, compiling important info and project management we extensively used Miro boards.
I started my experiments with Maya. To get smoothly merging and dividing geometry my first idea was to use morph target/blend shape approach, but it required lots of manual work and was not scalable. Second idea that seemed more promising was to use various pieces of base geometry and curves as guides for particles and when converting particles to polygons then they are able to procedurally merge and divide.
In the end I ran agains many show-stoppers, for example it turned out to be very difficult to get different particle systems to merge together. I concluded that the tools in Maya work very well within their bounds but when trying to combine new tools with Maya's more native functions then troubles are abundant, quite often they produced either unexpected results or just crashed. Maya does have an extremely flexible and powerful node-based system called Bifrost, perhaps even comparable to Houdini, but my previous experience with it was none and I had to start producing results, so learning that from scratch seemed the wrong strategy.
Secondly, there had to be simulated particles that would interact with that dynamic geometry and also follow the pre-existing shapes.
And finally, the whole system would have to be animated, probably partly procedurally but still in an art directable way. For that I tried using Mayas MASH system, it is quick to set up, it has many different functions and produces high-quality simulations. It has a pretty approachable block-logic system that initially seemed promising. At the same time I also started developing the shaders.
I was however able to produce some of the first visualisations with the previously described method and with lots of manual labor it theoretically would have been animatable. The visuals were perhaps a little underwhelming but for the sake of common understanding between all the team members it was a huge help, finally we had something visible that could be discussed over. Previously it was just words on paper and some rough sketches.
After abandoning Maya with a good depiction of what a cluster migraines would feel like, I went over to Cinema 4D, it is after all the go-to tool for procedural animations.
The quest was after dynamic simulation at scale in order to represent cell division. MoGraph tools produced high quality cool-looking results very quickly and combining them with Volume builder almost gave the dividing look I was after. Overall the Volume builder seemed very promising in combining any number of objects with any shape in order to get the exact result I needed. However I struggled to get the single cell visuals, especially when things involved particles and various curve animations. In the end Cinema 4d tools proved to be too inflexible, they work really well in the bounds that are set by the parameters of each tool but for more specific custom setups it proved quite limiting. The next version of Cinema promised to have a more customisable Scene Nodes system, but I did not have that version.
I was starting to get the theme of it, that I am demanding a bit too much from one single setup, perhaps I can't do everything in a single animation/simulation but need to make many different systems for various illustrations and just blend them together in the video editing step.
Finally I turned to Blender. The metaball system seemed to be the quickest and most flexible way to smoothly merge and separate different shapes so far. Metaballs could be instanced to particles or vertices that follow a curve and they would all automatically combine when close enough to each other.
So again I started creating a single cell with this system, the "tail" extrusion is guided by a curve and could be animated to move. Initial material system was also drafted.
For example when instancing the metaballs on animated vertices that were in turn in an array and followed a curve, then in the viewport they randomly expanded exponentially, but when toggling Edit mode they turned back to normal and rendered view didn't show any faults, with that I had to just deal with.
Fortunately Blenders bug reporting system got some quick replies, unfortunately for many of those bug that were really counted as bugs, not just that specific systems peculiarity, the solution by the developments team was to just use the upcoming Geometry Nodes system, but that was still under development and in beta form, so using that in production did not seem very good either. In the end I just figured out some workarounds for the showstoppers and learned to just live with others.
The final setup had extendable and contractible extrusions in many directions, the shape and location of curve points could be animated and retractable tubes were inside.
While developing this setup, I started running into some arcane bugs that were highly specific when combining so many different systems and trying to squeeze out all their functionality.
The teamwork was quite challenging but also thereby interesting. The nature of the tasks was varied and changed quite a bit during the whole project, it was a process of constant experimentation for everyone involved. The details for every task evolved alongside with the observations and experiments done by the scientific team. I often updated visualisations that were previously considered done many times just as new data came in. We tried many different types of visualisations for the process described in the paper at different scales and at different levels of stylisation, many were left out from the published materials.
Another intricacy of the process was the varied cultural and linguistic background of many team members but it was also interesting to explore and overcome those communication barriers.
Overall this was one of the most intriguing project that I have ever participated in and I am very appreciative that I was given the chance to contribute to such an endeavour.
After all the main tasks were complete, I did some concepts for the paper cover image, they were supposed to represent the main concepts in a surrealistic or metaphorical way, also water-colour was chosen as the general aesthetic.