AlphaFold Database as a community platform

AlphaFold Database is expanding from a repository into a community platform for highvalue

predicted structure datasets, training, and global access. AFDB now supports

the integration of community-generated model collections, including monomeric and

multimeric predictions across viral, bacterial, archaeal, and divergent parasitic

lineages, within a shared framework for metadata, provenance, validation, and reuse.

These datasets are connected to user-facing features that help researchers interpret

models, including confidence metrics, domain annotations, predicted functional

effects, sequence and structure viewers, and links to related structural and biological

resources. A major example is the new protein complex dataset developed with the

Steinegger lab and NVIDIA, adding 1.81 million high-confidence predicted homo- and

heteromeric complexes across 4,777 proteomes. However, for these resources to

have broad impact, access must go beyond data availability. AFDB is expanding

multilingual training and translation activities, including Portuguese, to support

researchers in assessing model confidence, interpreting predictions, and applying

structure data in local research contexts. Together, these developments position

AFDB as an open, community-driven platform for sharing, understanding, and reusing

predicted structure data.