Make money doing the work you believe in
While Alphafold is often pointed to as the breakthrough, Nobel prize winning example of how AI will transform Science, there has been very little reporting and analysis of how AlphaFold has changed structural biology.
A NBER working paper from economists Ryan Hill at Northwestern and Carolyn Stein at Berkeley have provided some analysis of how AlphaFold has impacted structural biology since its introduction in 2021.
And no, AlphaFold isn’t automating biologists jobs.
Instead, they find:
Structural biologists are still solving and depositing experimental structures to the PDB at the same rate as before 2021 and publishing them in top journals.
Instead of replacing experiments, AlphaFold complements them. Researchers use its predictions as templates for molecular replacement — especially for novel proteins with no close experimentally-solved relative, the structures that were historically hardest to solve.
And basic research has tilted toward proteins that had no experimental structure before AlphaFold: papers on these previously-unsolved proteins (function, expression, disease, and more) rose significantly relative to already-solved ones, and disproportionately cite the core AlphaFold papers. The authors call it a "floodlight" effect, making the broader protein space cheaper to explore.
But downstream drug discovery hasn't followed yet. Looking at bioactivity assays — an early step in testing whether candidate molecules bind a target — they find no comparable shift toward these newly-illuminated proteins. AI slashed the cost of one key input (structure), but the pipeline is still gated by slower-moving complements.