Biology Reference
In-Depth Information
we cannot, from the structure alone, define its biological role and the
processes and pathway(s) in which it is involved. In any case, predicting
the molecular function of a protein on the sole basis of its 3D structure
is in itself a very challenging task. If the active site has been observed pre-
viously
71-73
or if the protein has been cocrystallized with a substrate ana-
log, we have a better chance of succeeding; however, correctly predicting
specificity remains a relatively rare event.
74-76
As protein models inherit
many of the features of their templates, but with reduced accuracy, it is
not surprising that the detection and functional assignment of a known
active or binding site requires models of good quality.
77-79
The relationships between sequence identity, model correctness, and
model accuracy have been described above. One should not overlook the
fact that these sequence identity levels are an average value computed
over the whole of the protein or, at least, of the core elements common
to the target and modeling template. In most cases, however, enzyme
active sites or ligand-binding regions are more conserved than the rest of
the structure. Therefore, one can often build high-quality models for
active sites, while other regions of the protein model are less accurate or
even display irresolvable correctness issues in distal loops.
Nonetheless, low-accuracy models can provide additional hints to
protein function and can help confirm sequence similarity searches and
the results of fold recognition. Indeed, it is an advantage to build mod-
els
75
in such cases, as structural insights can confirm hypotheses derived
from homology detection. In our hands, we were able to verify and
confirm the assignment of several
C. elegans
insulin-like genes using
low-accuracy models.
80
Similarly, the trimeric nature of the CD40 ligand
was first proposed based on a low-accuracy model in which the target and
the TNF
α
template share less than 26% sequence identity.
81
5.1.1. Studying the impact of mutations and SNPs
on protein function
Functional analysis of proteins greatly benefits from the observation of
naturally occurring mutations. Indeed, diseases, or less severe phenotypic
variations, which can be univocally assigned to single-point mutations
provide a good framework for understanding the molecular function and
