Modern Non-LLM AI

AlphaFold and Biomedical AI

How AI is transforming biology and drug discovery

What it is

AlphaFold2, developed by Google DeepMind, solved the protein structure prediction problem (predicting a protein's 3D shape from its amino acid sequence) earning its creators the 2024 Nobel Prize in Chemistry. The system uses transformer attention applied to multiple sequence alignments and evolutionary co-variation patterns.

The impact was immediate: the entire known protein universe was predicted and published openly, accelerating research across biology and drug discovery. AlphaFold3 extended this to predicting interactions between proteins, DNA, RNA, and small molecules.

Beyond protein folding, biomedical AI includes: drug-target interaction prediction, medical image analysis (pathology, radiology), clinical trial patient matching, and genomic sequence analysis.

Why it matters

AlphaFold is the clearest example of AI solving a previously intractable scientific problem, and represents the potential for domain-specific AI to transform fields beyond software. It also illustrates a key pattern: transformer architectures generalize across domains when the problem can be framed appropriately.

Resources

Artificial Intelligence in Software as a Medical Device

fda.gov· FDA's official overview of AI/ML in medical devices. Covers regulatory frameworks, approved AI medical devices, and the broader landscape of biomedical AI beyond AlphaFold. Authoritative government source. **Confirmed.**

10 min