Conclusion

With the extension of CAMEO to the fully automated assessment of prediction of complexes (including protein-protein, DNA, RNA, peptides, small molecules), we aim to encourage and facilitate the development of automated structure prediction servers going beyond the modeling of single chains of amino acids. In this manuscript, we identified several challenging aspects of modeling which we believe will become more active areas of research in the future, and that are suitable for benchmarking with CAMEO. By assessing prediction targets with the same complexity as experimental structures using an “opt in” mechanism for the diverse modelling tasks, CAMEO will assist the development of new methods tackling these specific modeling challenges. As demonstrated by analysing the PDB releases of the last year, CAMEO will be able to provide a diverse set of challenging blind prediction targets to enable the community to tackle next generation modeling challenges.
We welcome feedback from the community on which of these aspects should be prioritized and how various predictions should be numerically evaluated in CAMEO. We encourage methods developers to register to the beta CAMEO server to help testing and evolving these new features according to the needs of the prediction community.