Beyond single proteins: complexes join the AlphaFold database

What the research found

EMBL reported the addition of approximately 1.7 million high-confidence homodimer predictions in March 2026 and around 80,000 high-confidence heterodimers in a May update. Larger lower-confidence collections were made available separately for bulk use. These are complex predictions, extending the database beyond individual protein models.

Where the evidence stops

These counts should not be added to the over-200-million single-protein headline as if they were all unique newly decoded proteins. Confidence categories, complexes and individual protein entries describe different things.

The private-client perspective

For research planning, the distinction matters because a protein’s interaction partners may be as important as its isolated structure. Good computational practice records the provenance and confidence of each model and asks whether its context matches the biological question. Scale can widen the search space, but it also increases the importance of prioritisation and experimental validation.

Read the source

1. EMBL: first complexes in the AlphaFold database; updated 19 May 2026 (2026-03-16)

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.