ModelArchive, an open repository for sharing computationally determined protein structure models, has been awarded a grant by swissuniversities to expand open research data (ORD) principles. Several SIB group leaders are collaborating with international experts to develop this data resource, which complements the Protein Data Bank (PDB) for experimentally determined protein structures.
Harnessing the explosion of predicted protein structures
Novel modelling methods powered by deep learning, such as AlphaFold, now allow prediction of protein structures at a high quality, reaching near-experimental accuracy. This has resulted in an explosion of computer-predicted models, which complement the otherwise limited range of experimentally determined structures. Furthermore, advances are already being observed in the modelling of protein complexes such as of their interactions with small molecules (e.g. drugs), RNA and DNA. The modelling of proteins in different conformational states is also progressing as well as in protein engineering and design studies. All these models need a dedicated space to be stored and made available to life scientists for use in their respective fields of research, from drug discovery to molecular biology.
This is where ModelArchive comes in. Developed in the group of SIB's Torsten Schwede, it enables protein structures determined by computational methods to be deposited and shared. It thus complements the Protein Data Bank (PDB) and PDB-Dev which store structures derived and partially derived from experimental data respectively. By building and expanding on current open research data best practices such as the ModelCIF data format, ModelArchive aims to become the reference repository for sharing computationally modelled protein structures. This initiative to foster open research data (ORD) has been awarded funding by swissuniversities (See box).