A paper in Science reports a design method that substantially advances the macromolecular technology base for building atomically precise nanosystems.
Background: foldamer engineering
As many readers know, biology shows an effective way build large, intricate, atomically precise systems: Use covalent chemistry to build chains of small building blocks, and design these chains to fold into nanoscale building blocks that undergo spontaneous assembly driven by Brownian motion and selective binding. This is a key step in climbing a ladder of fabrication technologies that leads to broader, more powerful capabilities.
The covalent synthesis of suitable chains of building blocks* was mastered decades ago, using programmable nanoscale machines that operate in biological systems. Designing structures that fold into compact nanoscale objects has become increasingly routine. Designing these building blocks to assemble, however, has lagged.
This highlights the importance of the paper in Science.
The authors (from the Baker lab, and I’m tempted to add “of course”) used RosettaDesign-based protein engineering tools to design proteins with surface structures that bind to a natural protein at a particular location, and with a particular orientation. Finding a protein that binds isn’t too hard — screening and evolutionary methods applied to antibodies (among other proteins) can do this — but achieving high affinity (tight binding) in a specific geometry is new.
They achieved this by designing binders with the correct geometry but mediocre binding, and then using selection (the equivalent of antibody affinity maturation) to refine the interfaces to achieve high affinity. The refinement process retains the initial alignment with good fidelity.
The binding target was a conserved region of the influenza hemagglutinin molecule, hinting at an approach to developing a subtype-independent anti-influenza therapy.
Solving a harder problem than necessary
Note, however, that authors didn’t address the problem of designing building-block interfaces, as an engineer would understand the task: They did something harder. Only side of the interface was designed to bind, while the other was a naturally occurring structure that normally binds nothing.
An engineer designing building-block assemblies, by contrast, would design the interface as a unit, not just one side of it.
It’s easy to see the advantages of being free to tweak both sides to achieve a good fit, to balance solubility and costs of desolvation, and to introduce specific binding interactions (hydrogen bonds, salt bridges, hydrophobic pockets on one side that match hydrophobic side chains on the other, etc.). Freedom to design both sides together also means that protein engineers — when pursuing engineering objectives — can exploit the best-understood motifs, rather than deliberately plunging into the unknown.
In conventional engineering, no one designing a system would freeze the design of one component, and then attempt to mate another to it at a location not designed for the purpose. Interfaces aren’t afterthoughts.
A companion perspective piece for the paper observes that
Although Fleishman et al. have produced a landmark result, it is evident that computational protein interface design is not a solved problem.
For the more symmetric engineering design problem, however, the methods described in the paper can be expected to provide a basis for reliable design tools.
I look forward to seeing the methods and the lab results. This should be low-hanging fruit.
* In other words, peptide foldamers (commonly called “proteins”) which include a range of high-performance engineering polymers.