2021 · Computational research
A structural atlas for the human proteome
Broad coverage transformed access to protein models while preserving important uncertainty.
Research publication: 2021-07-22 · English briefing: 18 September 2026
What the research found
Tunyasuvunakool and colleagues applied AlphaFold to almost the entire human proteome. Their paper reported coverage of 98.5% of human proteins, but confidence was not uniform: only a subset of residues had very-high-confidence predictions. The atlas made many new biological questions easier to investigate and illustrated why coverage and certainty are different measures.
Where the evidence stops
A predicted model is not an experimentally solved structure. Some regions may be disordered or difficult to model, and a static structure does not describe every functional state.
The private-client perspective
An effective research dashboard should show both the result and the uncertainty attached to it. For an individual’s variant, researchers must consider whether the relevant region is reliably modelled, whether the variant affects function and whether that effect has clinical significance. The ability to display a protein in three dimensions is useful communication; the evidence supporting a decision is the more important deliverable.
Read the source
An independent editorial synthesis of published research, not an original NoahThera study or a personal medical recommendation. Evidence stages are identified above; cited researchers and institutions are not represented as NoahThera partners.