Recently, a pivotal moment in protein design happened: binding affinities and success rates have improved rapidly.

Protein design is already a very hot area, and the distance from idea to potential solution has shortened dramatically.

Here are some reflections on the importance of this moment and some looking-up thoughts for the area.

Antibodies for the next epidemic

Let’s roll back in time to 2020: the start of COVID-19 pandemic, a mix of chaos and panic, stay-at-home orders, masks, and social distancing.

Sidenote: I liked COVID (that’s uncommon, I know): it was a rare case when multiple governments tried to address real problems with the tools they had. They didn’t look well-equipped, but they weren’t insane. And yet many governments had obvious problems with… being trusted.
But overall, the world made much more sense to me back then. And so many good lectures were recorded and made available online!

Multiple (well-informed) people expected monoclonal antibodies to serve as a quick bridge before vaccines became available. Basically, a shot that would provide immediate, short-term immunity — especially helpful if body’s immune system cannot handle an ongoing infection.

Back then, this plan didn’t quite work — fishing for the right antibody took too long, and in the end, vaccines did most of the work. In 2026, it would be a completely different story — a new antibody could be designed computationally in a day.

Designing a good antibody is only part of the solution — there is still manufacturing and distribution parts to handle, but the science to make it happen is ready.

Basics: why antibodies? How are they used in medicines?

A number of medicines are actually antibodies. Wiki has a list of them. Their names usually end with “-mab” (monoclonal antibody), like Trastuzumab (used against breast cancer) or Infliximab (for bowel disease and rheumatoid arthritis).

“Natural” antibodies generated by the body bind to viruses and other pathogens. Designed antibodies can bind to pretty much anything they can reach.

Pharma likes antibodies because they are highly predictable molecules, naturally abundant in the body, well-studied, and long-lived — lasting about a month, while most other proteins degrade much faster. Furthermore, a month is not a hard limit, their lifespan is engineerable.

Why do we want higher binding affinity?

Broadly, we design proteins to interact with other proteins and molecules, which in most cases starts by binding to them.

High affinity ensures that antibodies do exactly this: they stay bound to their target instead of walking around, and maybe interacting with other molecules. The last thing we want from any molecule is off-target activity, because nobody likes side effects.

So, what has changed for drug development?

It mostly comes down to timelines: developing a molecule for a specific job has become much faster. Now, there is a very high chance of obtaining a molecule that does exactly what is requested and nothing else, satisfies a long list of “developability” requirements (a good molecule should be easy to produce, package, store, and deliver), and is well-tolerated by the body (including the immune system).

Previously, this was a major challenge that could take months or even years, with no guarantee of success. This timeline has been shortened dramatically, and validation is now the bottleneck (which creates a strong incentive to speed up the validation process too! we just aren’t there yet).

This will help reallocate drug development resources to other critical questions: how the disease works, the most efficient way to intervene, and how we can be confident that in vitro findings translate well to the clinic.

Rational design allows one to control binding to multiple targets and exact epitope and binding pose — things not possible with older screening techniques. This in turn unlocks progressively more complex and nuanced drug mechanisms.

Personal hopes: advanced diagnostics

While Chai-3 was in the late stages of cooking, I was barely able to do anything for almost three weeks as my body was fighting some infection. Did I say “some” infection? Yes — I don’t even know what sort of URI I had!

Isn’t this insane? The best available diagnostics right now are tests from the COVID era that were upgraded to detect four viruses (COVID, RSV, Flu A, and Flu B).

There are around 200 common cold viruses, and at this point, it is entirely feasible to detect proteins from each of them and find antibodies (better yet, multiple antibodies) against each one.

I hope this becomes a reality. In a couple of years, I’d love to be able to buy a test, identify the specific viral or bacterial variant, and then get a targeted antibody shot (or even a nasal spray) for that exact strain.

Outside of drug discovery: building blocks of biotech

If you work in a biotech wet lab, you can’t escape antibodies. They are everywhere.

Want to measure the amount of a protein? The most common protocol is ELISA, which requires an antibody that binds to the target. (plot twist: usually two antibodies, and sometimes three.)

Want to label a specific type of immune cell? That’s easy — you just need an antibody for a surface marker. (fooled you! usually you need at least two antibodies. And you might also need secondary antibodies to map them to specific detection channels)

Purification? Antibodies. Flow/CyTOF? Antibodies. Tissue imaging? Antibodies.

Despite their omni-presence, antibodies are notoriously unreliable:
my colleagues have spent (and are still spending) an insane amount of time validating new batches of antibodies,
because a new shipment of previously validated antibody might simply not work.

Even so, molecular biology has already made huge advances using this technology; now, we can make it more efficient, tunable, and reliable.

When a technology becomes fast, cheap, and reliable, it becomes the foundation for the next layer of technology — and that’s what I am waiting for.

We’ve been busy with technicalities, now we can focus on more interesting and valuable problems — ain’t that great news?