Biology Reference
In-Depth Information
The SWISS-MODEL expert system for comparative protein structure
modeling and its related resources
10-20
have been developed to facilitate
the usage of protein structure information (both experimental and
model-based) for the broad biomedical research community.
2.1. SWISS-MODEL Pipeline
Fully automated, large-scale protein structure modeling requires a sta-
ble and reliable modeling pipeline. Starting from no other input than a
protein sequence, the system should (a) identify possible template struc-
tures related to the target protein from the template library and select
the most suitable ones; (b) generate reliable alignments of the target and
template sequences; (c) build a 3D model of the target, including side
chains, structurally variable regions, and segments corresponding to
gaps in the alignment; and (d) evaluate the resulting models in order to
identify errors and inaccuracies, and select the most reliable model. The
SWISS-MODEL pipeline combines these individual steps into a work-
flow,
12,16,21
which powers both the automated modeling steps in
the Workspace and the large-scale modeling efforts for the Repository
(see below).
2.2. SWISS-MODEL Template Library
Comparative protein structure modeling requires high-quality experi-
mental protein structures as templates. The SWISS-MODEL template
library (SMTL)
19
is derived from the remediated Protein Data Bank.
22
Each PDB entry is split into individual chains to allow sequence-based
template searches. Each template chain is annotated with information
about the experimental method, resolution (if applicable), ANOLEA
mean force potential,
23
force field energy, and quaternary state assign-
ment to allow for rapid retrieval of the relevant structural information
during template selection. Low-quality structures consisting only of C
α
atoms and short peptide fragments are removed. Searchable sequence
indices
24
and a library of hidden Markov models (HMMs)
25
at several
levels of sequence redundancy are provided.