l - People

structure-based hinge predictors (TLSMD, StoneHinge, and FlexOracle), plus HingeSeq,
a sequence-based hinge predictor. Each of these methods predicts hinges using very
different sources of information -- normal modes, experimental thermal factors, bond
constraint networks, energetics, and sequence, respectively. Thus it is logical that using
these algorithms together would improve predictions. We integrated all the methods into
a combined predictor using a weighted voting scheme and showed that this was superior
to any of the individual ones. Finally, we encapsulated all our results in a web tool which
can be used to run all the predictors on submitted proteins and visualize the results.
.
Introduction
The structural information provided by the atomic coordinates of a protein tells only part
of the story of protein function. Much of the remainder is told by the trajectory of
motion. Motions can be classified according to the size of the mobile units, which may
be fragments, domains, or subunits, and according to packing as hinge, shear, or other.
Hinge bending motions are the largest single class of motions, comprising 45% of the
motions in a representative set compiled by Krebs et al.[3, 9] This class is further
subdivided into domain hinge motions (31% of the total)[3] and fragment hinge motions
(13%). A logical first step towards the goal of motion prediction for the case of domain
hinge bending motions is to predict the hinge location. In this work we compare several
existing algorithms, present new ones, and combine all of these into HingeMaster, a
composite predictor.
160