Intro
I'll be showing a prototype of a biomolecular design environment that we've been working on for a couple months. It's based on a new kind of operating system called Realtalk that my lab has been developing.
These are just prototypes. Their purpose is to explore ideas for what Shawn's lab could look like in a couple years.
That said, a lot of this stuff actually does work. We're hand-waving some of the simulation and analysis programs, but the stuff that "looks like magic", that part's all real. It can be pretty hard to follow in a ten-minute talk, so we invite you to come by the lab at UCSF and check it out if you can.
I'll be showing Shawn and me going through tasks from the project that Shawn described. We'll be designing a fusion of two proteins, designing a DNA origami structure for the proteins to bind to, and making the origami in the wet lab.
And this is the space where we'll be working.
Proteins
First, we want to get the BurrH and GFP proteins onto the table so we can start playing with them.
Here, Shawn is typing "BurrH" into the "PDB search" card. We see these are the two results from the Protein Data Bank. [more]
I'm going to pick up the one we want using one of these mock test tubes. [more]
It's kind of like a game piece in a board game.
Now the 3D structure is on the table, and I'm going to use a laser pointer to spin the molecule around and examine it.
I'm now scanning through the sequence to see where those residues are in the structure.
And I'm looking up GFP in the same way. I'm going to grab the one that we want.
And now we're just going to play with them, positioning them relative to each other and thinking about how we want to connect them.
In order to connect them, we'll need to permute the GFP, by closing up its termini with a linker and breaking the loop somewhere else.
Here's a little "permute" program that we made, and I'm pointing its whisker to where we want the termini to be.
And I pick up the permuted GFP using another tube.
Now I'm just comparing them, the new one and the old one, and I'm going to swap our new one into the scene.
And next, here's our "splice" program. I'm going to point its whisker to the spot on BurrH where we want to splice in the GFP.
And I'm going to pick up the protein representing the two of them spliced together.
Here's the new sequence that we generated. I'm scanning through it. The blue text is from BurrH, the green came from GFP, and pink are the linkers that we added.
Here, Shawn is showing me a more complicated idea that uses two BurrHs and a SpyCatcher-SpyTag system to pin the GFP down in two places. He's kind of just doodling on the table.
So that's a future project. But even now with one BurrH, we don't know which splices will actually work.
So here we're trying a different splice point, and saving it off into its own tube.
We have these candidate designs on the table. And one way we like to plan an experiment is by envisioning the gel that we expect to see at the end.
So here we're organizing our mock tubes on a mock rack. We see them as lanes on a simulated gel. [more]
This includes both the new proteins we're making, as well as intermediate products that we'll want as controls.
But before we think about protein expression, we want to design the DNA origami structure that these proteins will bind to.
DNA origami
Here I'm swapping out the scale bar on the table for one that's more origami scale.
The protein is pretty small at the scale. You can start to imagine the goniometer around it.
And we're going to bring out our tub of origami blocks. [more]
Each block here represents a pleated layer of DNA helices.
The blocks snap together using magnets at valid crossover positions to form a honeycomb lattice.
These blocks here are six helices tall and six DNA turns long.
Here we're assembling blocks on the table into a three-dimensional scale model of the nanostructure. As we build the model, we see the routing diagrams above fill in with an auto-routed scaffold strand.
Here we're putting in the blocks for the single helix stage that the proteins will bind to.
This is one goniometer design. Shawn here is going to pick it up into a mock tube.
And I'm playing with positioning one of the proteins that we made onto the stage.
Here Shawn is adding a feature to the bottom of the origami to make a variant. He's going to pick up this variant into another tube.
And now we've got a few designs on the table. We can spin them around, talk about them, point to things.
Here we're organizing the ones that we want to make on a rack, and seeing the simulated gel.
Now, we also want to preview the protocol that we'll be using in the wet lab to make these. So I'm pulling out a couple more mock racks and mock plates. [more]
And the staple strands in our set of designs are automatically assigned to wells in the plates on the left. And we see an auto-designed folding protocol flowing through the tubes to the right.
This is exactly the same setup we'll be using for real when we get to the wet lab a bit later. But now we're simulating it.
I'm going through the protocol, step by step, to preview each pipette transfer. And we see, at each step, the contents of the tube and their concentrations.
And this is all live. As we rearrange the rack, we get a new protocol.
Here, we're rearranging the origami blocks, and seeing the staples and protocol update immediately.
Cryo EM
Now, our purpose in designing these origami structures is to be able to recognize them under the microscope.
So here, I'm putting down the mock EM grid, putting one of our mock tubes on it, and firing our mock electron beam at the grid. [more]
And this generates a set of simulated micrographs.
Up on the wall, I'm pinning up the "show micrographs" program, and we're looking through our simulated images.
As we put more samples on the grid, we see more species in the images.
Here, we're adjusting the concentration of one of the samples, to change the density.
We can use these simulated images to design our image analysis pipeline.
Here, I'm putting up programs that would pick particles from the image, classify the particles, show class averages, and so on.
Here, Shawn is rearranging the origami blocks, and we're seeing that entire analysis chain update live.
The idea is that we can design our analysis algorithms at the same time, and in the same place, that we're designing the nanostructures to be analyzed.
Because it's simulated data, we already know the right answer. So we can evaluate how well our algorithms are doing, and get everything debugged in simulation before going through real data collection runs.
Wet lab
So we decide on our origami designs. we order oligos, we're ready to fold them in the web lab. [more]
Here, we're putting down our generated protocol on the bench, and it's prompting for the plates and racks that we need.
So we go get those, we get our tubes, get our reagents.
Here, I'm lasering the names of some reagents to see information about them, including where they're physically located in the lab, and where we can order more.
We do a quick preview through the protocol to remind ourselves of what we're doing, and then we begin.
At each step, a projected line shows exactly what to pipette where, and how much.
Next to every tube, we see the expected contents of the tube and their concentrations at that point in time.
Above the bench, we see the protocol as a list of instructions, and we see where we are in that list.
So this is kind of like turn-by-turn navigation for the web lab.
Now, at the end of the project, we might put some of these tubes in the freezer. And years later, maybe we take one out, put it on the bench, and we would expect to see immediately, from that tube, the protocol that generated it. As well as the origami designs, the data we later collected, the paper that was published -- all that could be attached to the physical test tube.
Summary
So physicality is one of the key concepts that we're leveraging here. Using our hands, designing 3D structures using 3D models, seeing what's physically in test tubes.
Molecules and test tubes are physical things, so we'd like our computation out in the physical world where the science is happening.
Another key concept is the social nature of this environment. People work together, side by side, getting their hands on the same thing.
What if this was what a group meeting was like, people getting together to play with each other's molecules?
A third key concept is designing using computational models.
If we can simulate molecules, simulate nanostructures, simulate microscopes in real time, we can get immediate feedback as we design, and have a consistent shared mental model with others.
And a computational model also serves as a precise representation of what we think we understand. If we do a physical experiment and it doesn't match our simulation, then we've actually learned something, and we can update our models.
The last key concept is that this is a completely programmable environment, where programs can be quickly improvised in the moment.
Here, we're adding a different linker to our protein splice by directly editing the splice program.
Here, I'm making a new scale bar by editing an existing one.
And here, we're directly programming an origami block to simulate passivation.
This entire set of tools that you've just seen was made in exactly this way, by editing the pages on those posters there, any of which can be changed live at any time, by anyone, right while we're using them.
What you see on these posters is the entirety of this prototype. There is no code base in GitHub. Everything having to do with biomolecules is literally just these twenty or so sheets of paper.
Of course, we're also using the Realtalk operating system, and that is its own poster gallery. It's a bit larger. [more]
Although lately, we've been keeping our own copies in binders. [more]
So the whole operating system itself is made of a manageable set of physical objects and is completely live editable by anybody.
My lab has been working on Realtalk for several years, and these were the core people involved in that work. [more]
This biomolecule prototype was made by the three of us, just in the last couple of months.
It's been a lot of fun, and we're looking forward to seeing where it goes.
Thank you.