AlphaFold

Inspiring video that dives deep into how artificial intelligence is reshaping protein science—a field that once relied on decades-long experimental methods like X‑ray crystallography. The video recounts the evolution from early computer models in the CASP competitions to the revolutionary breakthroughs we see today.

DeepMind’s entry into this space changed everything. Their first version, AlphaFold 1, used a deep neural network and evolutionary data to predict a 2D pair representation of protein structures—a major win at CASP, yet one that still had room for improvement.

Enter John Jumper and the team behind AlphaFold 2. By harnessing more powerful computing, refining their algorithm, and leveraging the transformer model of attention, they introduced the innovative EVO Former. This breakthrough architecture separated biological and geometric data, using triangular attention to refine protein structures before assembling them in a novel way—no longer simply a chain of amino acids.

AlphaFold 2 didn’t just win awards; it revolutionized the field by predicting millions of protein structures and ultimately earned its creators a Nobel Prize. On a parallel track, David Baker’s development of RF Diffusion is pushing the boundaries even further—designing entirely new proteins with specific functions like human-compatible anti-venom.

These advancements are not only accelerating discoveries in vaccine development and materials science but also opening up solutions to some of our most pressing global challenges. It’s a fundamental shift in scientific methodology—making what was once impossible, possible.