Fascination with proteins

There’s a moment I remember clearly from early in my biology degree. Someone showed me a protein structure in a computer on Chimera, everything rendered in gray colours against a black background, and that felt new. Very new. You could rotate it, zoom into it, explore the thing in three dimensions.

I remember thinking: this is real. This drives almost everything that happens inside every cell in my body right now, doing something specific, at a scale I could not imagine, faster than I can think.

That was it for me. Proteins all the way.

What still get me of proteins - and i don’t think i will get past it

The part that still gets me, after years of working with them, is how much is encoded in that sequence. The folding instructions are there, or so we think. The function is there, or so we think. The specificity, which molecule this protein will bind, which reaction it will catalyse, which signal it will transmit, all of it compressed into a string of building blocks that, in isolation, tells you almost nothing.

Proteins are also, by any reasonable measure, absurd in their variety. The same twenty amino acids, combined in different sequences and lengths, give you: enzymes that break down food, structural fibres that hold your tendons together, receptors that sense light in your eye, motors that physically walk along other protein fibers inside cells carrying molecular cargo, antibodies that recognize specific shapes they’ve never encountered before. It’s insane.

This breadth is not an accident. Evolution has had a very long time and a very large combinatorial space to explore. What proteins have turned out to be capable of is more than any sane person imagination could have produced.

What we were told - the folding problem

For decades, one of the central challenges in biology was predicting what shape a protein would fold into from its sequence alone. And rightly so, how are you supposed to predict that. You could sequence a gene relatively easily. Understanding what the resulting protein looked like required either crystallizing it (slow, expensive, often impossible), or homology modelling which you needed a reference too. Too bad if your protein is new, different or just… strange.

AlphaFold changed this. DeepMind’s model, released in 2021, predicted protein structures with an accuracy that genuinely surprised the field, including the people working on the problem. The Protein Data Bank had taken decades to accumulate around 170,000 structures. AlphaFold released predictions for over 200 million proteins within a year.

The reaction for me was slightly unsettled. The way any field feels when a problem it had defined itself around gets solved faster than expected. The question shifted almost overnight from what does this protein look like? to what does it do, and can we change it?

Why it stays interesting

I’ve spent most of my career on a narrow part of this, enzymes specifically either for therapeutic use early on and now for catalysis. But the broader fascination hasn’t faded, and I think it won’t.

Proteins are a system where the gap between what we can describe and what we can predict is still enormous. We can name the amino acids in a sequence. We can, now, usually predict the shape. And even what activity they will have, but so much remains unknown. Up until 10 years ago, you’d look at crystal structures and that’s it. Now, Molecular Dynamics is a must. And lab work, obviously.

There’s something intellectually honest about protein science, when the experts are still genuinely surprised by the answers, it means the questions are real and we’re barely getting started.