Biology Reference
In-Depth Information
(June 2008) consists of 5.8 million comparative protein models for
1.97 million distinct UniProt entries.
7. Future Outlook
With the deluge of new DNA sequence information being generated by
large-scale genomics and metagenomics projects, protein structure mod-
eling will be the only feasible approach for inferring structural informa-
tion for the vast majority of proteins in the foreseeable future. The
development of reliable, accurate, and easy-to-use resources for protein
modeling is mandatory to make best use of the investments in experi-
mental structure determination, and to provide structure information for
biomedical research projects which are not directly supported by experi-
mental structure determination efforts. However, automated protein
structure models still fall short of experimental structures, and significant
research will be required to ensure the reliability and applicability of
models.
One major task will be to develop suitable methods to assess the
applicability of models for specific applications, and to derive appropri-
ate model quality estimates if the accuracy of a given protein model (or
a part thereof) is expected to be sufficient for a given application. Several
tools for estimating the overall relative accuracy of protein models have
been developed recently.
64,123-125
However, significantly more work is
required to correctly estimate the expected absolute local errors within
protein models.
As errors and inaccuracies from template structures are propagated
into the models, further development of the SMTL will be crucial with
respect to the quality of HMM profiles or the correction of experi-
mental errors and inaccuracies.
126,127
A well-curated and annotated tem-
plate library is also a prerequisite to extend our automated modeling
pipeline to automatically model oligomeric proteins in their correct
biologically active state, and to include essential cofactors and ligands
in the models.
In parallel, we will further develop the user interface of the SWISS-
MODEL expert system. Recent advances in web technology will allow us
