Here are a few notes:
Additional research context for Protein Kit
One of the lab's goals is to build a second generation of cryo-EM molecular goniometers and apply them to visualize small proteins of unknown structure. Initially, we'll focus on natural DNA-binding proteins (e.g., nuclear receptors), but eventually, we want to tackle the more challenging problem of studying non-DNA binding proteins. We could jump into directly studying proteins of unknown structure, but it may be hard to determine the cause(s) if that approach fails since we don't know quite what we're looking for. Therefore, we want to engineer some control proteins based on known structures, including our old favorite, BurrH. These custom proteins will serve as proxies for the natural targets as we test out new goniometer designs and learn how to, for example, stabilize flexible regions.
Mockups and prototypes
Our paper mockups and the first Realtalk prototype have been remarkably useful for thinking, discussing, and improvising, both in the context of tool-building and in the protein engineering work described above, despite lacking most of their planned functionality! I think this is a great sign for things to come. (It also reminds me of cooking recipes where it's fun to eat the intermediate ingredients as well as the final dish.)
Emergent seamlessness
I, too, have been struck by how the distinct project headings seem to be merging into something that feels unified. For me, a big turning point was the Realtalk-enabled prototype that included the simulated gel, which was suggestive of several other things that we might represent and simulate. It is indeed hard to describe, but I'll give it a shot:
As Bret and I have talked through molecular biology concepts, or imagined the potential arcs of new research projects, we've frequently touched on aspects from separate headings that are also deeply connected. For example, our protein designs must take into account not only its 3D molecular structure, but the process of synthesizing the gene that encodes it, the expression strain, the purification strategy, and so on. Typically, those connections are not represented anywhere, let alone with interactive simulations in a shared space. Instead, each lab member gradually assembles a fragmented collection of physical and digital assets (lab notebooks, physical samples, powerpoint slides, software tools, image data, etc.). We manually shuttle assets from one app (or tube) to the next, and track most of the details in our heads. We spend a lot of time in powerpoint-driven meetings attempting to synchronize our understanding and plan the next experiments. Hopefully, we accumulate enough materials to assemble a convincing narrative in pdf format after a few years.
In contrast, the latest prototype offered a glimpse of future scientific conversations where speaking, gesturing, and programming dynamic simulations by manipulating physical objects will help us quickly update our (visible, shared!) models, make clear and efficient plans, see our entire inventory and how it was made, and so on:
"We should include an extra control here [picking up construct 5 in a Realtalk "test tube"] because it will strengthen that [gesturing at a draft figure in the manuscript area] section of the paper. And this condition should probably be in a separate gel [moving the test tube; the adjacent simulated gel refreshes]."
"Oh, these types of constructs typically take four to six weeks to synthesize, so we should move them over here so we can order them first. You can find the relevant vector in that –80° freezer box..."
"Hmm, this simulated gel band is running a bit faster than this one in the actual data. We saw something similar a few years ago in this gel; let's see what model we used at the time..."