Materials Informatics and Bioinformatics

1
Materials Informatics and Bioinformatics Scott
Broderick, Xue Li, Yves Sucaet, Krishna Rajan
(krajan_at_iastate.edu)
Data mining SVM applications
Agent Based Modeling
PCA of Interpenetrating Polymer Networks
Principal Component Analysis PCA can be used
to identify important compositions, which visual
inspection does not capture
Scott Broderick
Xue Li
Y ves Sucaet
Support Vector Machines SVMs
Quorum Sensing QS Described first in Vibrio
fischeri (Marine bacterium), which lives in two
states Free-living in the ocean (low gt non
luminescent) Symbiosis with host (high gt
luminescent) Host of V. fischeri is Hawaiian
bobtail squid (Euprymna scolopes) QS helps
organism decide which state it is in
So called optimal hyperplane is a hyperplane that
cannot only separate data of two classes but also
maximizes the margin between the two classes.
Maximize the margin So the target is to
maximize(Lagrange method)
QS applications in nature Biofilm,
polysaccharides, degradative enzymes,
antibiotics, pigments... Temperature sensitive,
motility, cell aggregation Signal is mediated by
Acyl-homoserine lactone AHL
molecules Involved species Pseudomonas
aeruginosa, Burkholderia cepacia, Aeromonas
hydrophila Clinical significance of
QS Orthopaedic biofilm infections, Neurosurgical
device-related infections Dental plaque (1000
different species interacting lab conditions) QS
plays a role in intercellular communication,
therefore by interrupting this process an answer
can be formulated to increased antibiotic
resistance in a variety of disease
vectors Agent-based modelling ABM ABM has
existed since 1940, when the concept was
formulated as cells on a grid. The technique is
ideal to examine complex systems and
interactions, whereby the focus is not on
equilibrium conditions, but rather intermediate
evolution and evaluation. Traditional
applications are in the field of supply chain
optimization and logistics or sociological
applications such as word of mouth and social
network effects.
Application in RNA-protein interface prediction
PCA can quickly identify compositional and
reaction pathway effects that involve timely and
tedious analysis otherwise .
PCA of interpenetrating polymer networks (IPN)
data used is from rheology and tensile
experiments. The samples were created using
combinatorial techniques. IPNs comprised of
acrylate and epoxy. Reaction sequence A denotes
acrylate reacted prior to epoxy, and reaction
sequence E conversely epoxy prior to acrylate.
Two major effects occur in IPNs those related
to composition and those to reaction sequence.
Table 1 Results using SVMs (polykernel)?

• Protocol
• Step 1 Artificially initiate QS by adding signal
  molecule into bacterial colony. This experiment
  is currently under preparation and will be
  conducted off-site under controlled conditions.

Fig. 2 Scores plot for IPNs Labeled percents are
percentage acrylate
Informatics-based screening of FTIR libraries
permits extracting all crucial information from
libraries more efficiently than otherwise possible

Fig. 4 Systematic representation of ABM with
state transitions in a 2D-grid
Reaction sequence
Each point in figure 3 represents an FTIR wave
number of IPN samples of varying composition and
polymerization sequence.From this plot, we can
quickly identify absorbencies related to
compositional effects, reaction sequence effects,
intensity of peaks, peak shifts and change in
intensity between samples.The monomer that the
absorbance is related to can be identified from
this figure with no additional information and
greater understanding of structure effects.
Epoxy-determined absorbancies

• Step 2 analysis mRNA level of thousands of genes
  and identify diff. expressed genes using
  statistical or data mining method (feature
  selection)?

In addition, ABM is now increasingly used to
model and simulate biological conditions. It was
Craig Reynolds who first suggested this technique
to model the reality of lively biological
agents. In the examples below, the following
parameters are set Assume molecules undergo
Brownian movement Naïve approach toward
chemical reactions Molecules bump into each
other and react
Composition
Acrylate-determined absorbancies
Fig. 3 PCA loadings plot based on FTIR
absorbancies.

• Step 3 predict which gene product will be
  outside membrane receptor so as to design
  antibody to block the receptors, hence the
  cell-to-cell communication

From polymer library, using informatics we can
quickly identify if a property is most determined
by reaction sequence or composition. We can
identify interesting compositions that are
otherwise hidden. We can create a new
visualization of many FTIR spectra, and determine
which peaks are related to compositional or
reaction sequence effects. Then the reasoning
for effects can be understood.
Fig. 5 Advanced ABM simulation of a complex
metabolic pathway (glycolisys)
Acknowledgements Joseph Nowers and Balaji
Narasimhan (Department of Chemical and
Biological Engineering, ISU)
Acknowledgements Krishna Rajan (MSE, ISU),
Michael Terribilini, Drena Dobbs (GDCB, ISU) and
Vasant Honavar (Computer Science, ISU)
Acknowledgements BCBLab (ISU), Dr. Rajan (MSE,
ISU), Dr. Proulx (EEOB, ISU), Dr. Wu (Math,
ISU), Garrett Dancik (Statistics, ISU)