A complete, carefully written guide to the AI that cracked one of biology's greatest puzzles — built for curious students who want to really understand it, not just hear the buzzwords.
Each chapter builds on the last. Start at the beginning if you're new to biology, or jump in wherever curiosity strikes.
What proteins actually are, how they fold from a string of amino acids into precise 3-D machines, and what goes catastrophically wrong when they don't fold correctly.
Levinthal's paradox, energy landscapes, and why the world's sharpest biophysicists couldn't crack the protein folding problem for half a century.
From CASP competitions and DeepMind's moonshot to the historic 2020 breakthrough that made structural biologists drop their jaws worldwide.
The real technical meat: MSA, the Evoformer, attention mechanisms, triangular updates, and the structure module — all explained with intuition first.
pLDDT colours, PAE heatmaps, and how to distinguish trustworthy predictions from model hallucinations — the skill that separates savvy users from naive ones.
Drug discovery, neglected diseases, the AlphaMissense variant classifier, antibiotic resistance, heat-tolerant crops, and plastic-eating enzymes — real impacts.
Five critical limitations that even expert users get wrong — static snapshots, disordered regions, hallucination, and what AlphaFold fundamentally cannot tell you.
Find a sequence on UniProt, check the AlphaFold Database, run ColabFold, and visualise your result in Mol* — with common mistakes clearly flagged.
AlphaFold-Multimer for complexes, AlphaFold 3's expanded biomolecule universe, molecular dynamics simulations, and open-source alternatives like ESMFold.
A curated reading list, essential papers with plain-English summaries, the best video tutorials, and communities where you can ask questions without fear.
Most resources about AlphaFold either oversimplify ("AI predicted protein shapes!") or assume a PhD in structural biology. This guide does neither.
Every technical term is explained from first principles — including the ones other guides just drop without explanation, like missense variants, intrinsically disordered regions, and co-evolutionary signals.