Using Data Mining Techniques to Probe the Role of Hydrophobic Residues in Protein Folding and Unfolding Simulations - Evolving Application Domains of Data Warehousing and Mining - page 260

Figure 1. Overview of a “data mining pipeline” applied to the analysis of data arising from multiple
molecular dynamics simulations

Thus, the study of SASA variation profiles of
different amino-acid residues along multiple MD
unfolding trajectories may shed light on potential
coordinated behaviour of residues far apart in the
protein primary structure, or even far apart in the
three-dimensional structure.
In recent years, we have made a significant
effort in developing a series of comparative-based
approaches to analyze SASA variation profiles of
individual amino-acid residues in protein unfold-
ing simulations (Azevedo, 2005; Ferreira, 2006,
2007). In this chapter, we show the application of
two different methods - clustering and association
rules - to study the solvent accessible surface area
profiles of individual amino-acid residues across
multiple MD unfolding simulations of the amy-
loidogenic protein transthyretin in search of groups
of amino-acid residues playing a coordinated role
in protein unfolding. The identification of clusters
of residues that change solvent accessible surface
area (SASA) concurrently during one unfolding
simulation or across several unfolding simulations
might prove important in shedding new light in
the highly complex problem of protein folding,
unfolding and misfolding (Brito, 2004).

In the following sections, we first introduce a
general overview on molecular dynamics unfold-
ing simulations of the protein transthyretin (TTR).
Then, we describe the main ideas associated with
the application of two data mining techniques to the
study of solvent accessible surface area variation
of individual amino-acid residues upon protein
unfolding. Next, we present and analyse the results
obtained by the application of these methods, and
show their usefulness in the interpretation of this
type of data. In closing, we discuss the need and
advantages of using data mining tools to analyse
data arising from protein folding and unfolding
computer simulations.

MoleculAr dynAMIcS
unfoldIng SIMulAtIonS
of trAnSthyretIn

In many fatal neurodegenerative diseases, includ-
ing Alzheimer´s, Parkinson´s and spongiform
encephalopathies, proteins aggregate into fibrillar
structures to form insoluble plaques known as
amyloid. More than 40 human diseases have been