PSI Structural Biology Knowledgebase

PSI | Structural Biology Knowledgebase
Header Icons

Related Theme

CASD-NMR: assessing automated structure determination by NMR

PSI-SGKB [doi:10.1038/th_psisgkb.2010.25]
Technical Highlight - June 2010
Short description: Are we close to automated structures being as good as expert-deduced NMR structures?

Caption: Performance of various automated structure calculation methods. The results of fully automated calculations by various programs for one of the masked test data sets of the 2009 Florence workshop compared to the reference structure (bottom right) determined by Aramini, J.M. et al., PSI NESG Consortium (PDB 2KIF).

Assignment of protein NMR resonances, especially those of side-chains, can be difficult and time-consuming. After completing this task, thousands of individual NOESY peaks must still be analyzed to obtain inter-atomic distance information. Software that performs this last stage should speed up structure determination considerably, and might ultimately reduce errors and increase reproducibility.But how close are we to the point at which completely automated structures can be deposited into the Protein Data Bank (PDB) without first having been checked manually? e-NMR (an EC-funded initiative), using data from the PSI NESG, launched a challenge — a sort of competition — in which groups could to test their the automated methods, and the results are encouraging.

CASD-NMR (critical assessment of automated structure determination of proteins) is a challenge open to the whole protein NMR community that allows researchers to blind test the structures derived by their automated methods against manually refined ones. The idea is similar to other critical assessment exercises such as CASP (critical assessment of techniques for protein structure prediction) or GPCR Dock (G-protein-coupled receptor), although CASD-NMR is based on experimental data rather than predictions.

CASD-NMR was launched last year when seven groups took up the challenge. Ten experimental data sets that included two from PSI NESG were made available, and all the teams who submitted an entry got together in Florence, Italy in May 2009 to compare the structures produced by their automated methods. Most of the structures had the correct fold, but some programs could not establish the correct packing and length of secondary structures. Root-mean-squared deviations between the automatically generated backbone and with the manual ones were generally 1–2 Å, but some were as great as 9 Å.

Now, CASD-NMR has evolved into a community-wide project. Any software developer or member of any laboratory can take part and anyone can propose their NMR data as a test case. In addition, PSI NESG will provide at least one data set per month on the CASD-NMR website or within the datasets section of the Structural Genomics Knowledgebase.

Participants have 8 weeks to generate automatic structures and to enter the coordinates and conformational restraints into the CASD-NMR database. The reference structure will then become available for comparison, and the software used must also be made available.

An assessment of the progress made using ten data sets all coming from PSI NESG is set for mid-2010. It will be very interesting to find out how the software is doing.

Related articles

GPCR modelling: any good?

Removing the NMR bottleneck with help from ROSETTA

Maria Hodges


  1. A. Rosato, A. Bagaria, D. Baker, B. Bardiaux, A. Cavall et al. CASD-NMR: critical assessment of automated structure determination by NMR.
    Nature Meth. 6, 625-626 (2009) doi: 10.1038/nmeth0909-625.

Structural Biology Knowledgebase ISSN: 1758-1338
Funded by a grant from the National Institute of General Medical Sciences of the National Institutes of Health