Project Overview
Biophysics is a relatively
new discipline of physics that seeks to understand living systems. Dr. Jeremy Schmit
is a theoretical Biophysist who is working toward
modeling various aspects of living systems. This summer’s project stems from
his interest in self assembling systems. Self-assembling systems are promenade
in biological organisms. Understanding how these work
will be beneficial to understanding how biological and neurodegenerative
diseases organize/emerge, improve drug development, and may prove to be a novel
technique for nonorganic assembly.
My project was to develop a
self-assembling system and then begin to look at how it works. Initially I
began by learning techniques in C++ to build a simulation. Then I replicated a
section of a paper by Dr. Schmit on Amyloid fibril
growth using direct templating. Direct templating develops systems based on a dichotomy that two
molecules can bind either correctly or incorrectly. The next step was to build
a 2-D environment for this system to grow in. Once developed, data was
collected about its growth rate in relation to concentration. The data supports
a general notion among two of Dr. Schmit’s papers
that these systems do not increase linearly with concentration. Rather, they
grow until a critical point where if pushed further the molecules do not bind
efficiently and growth levels off or decreases. The
environment was then converted to 3-D. Simultaneously,
it was of interest to develop methods to probe the geometry of the system
during growth. These methods were developed but only a few were implemented and
tested. This project warrants further investigation into the geometry of this
system.
Research Log:
Growth by direct templating:
This was the heart of the
project for the REU. I was tasked to develop and explore growth based on a
two-body interaction between molecules. I started in 1-D by replicating part of
Dr. Schmits paper referenced in the overview. The
next step was to develop a 2-D environment. Working in 2-D
produces a host of ergonomic problems that I did not expect. For example
in two dimensions it’s possible for an overlap of sites (figure 1). It required
developing code so that each site could talk to its nearest neighbors so that
an overlap of binding was avoided. Another example is dealing with boundary
conditions. The end of the grid is not the boundary on the growing amyloid. So
the simulation had to ignore this boundary and wrap around to the other side of
the grid. These are just a few examples of the problems that came up in
developing the environment of the simulation. It took a while to ensure that
the program was doing everything properly but this turned out to beneficial in
the end. After weeks of working in 2-D and collecting data and making upgrades
to the code I moved to 3-D. Quite a few of the same problems arose from the
conversion between 1-D and 2-D. Fortunately, because I spent a long time really
making sure that any major issues were fixed, these issues were trivial.
Putting windows on the Blackbox:
Once the system was developed
the next step was to figure out how to look inside of it while it works. This
is perhaps the most exciting part of the project. The task was to gather data
on how each site was binding and unbinding during growth. I was able to develop
a method of looking at how these site are binding and
unbinding. Furthermore, it was of interest to look at how kinked vs. nonkinked binding (pictured below) progressed through the
simulation. This is where the project concluded. However, the initial data
implies that there may be a difference in geometry between the most productive
concentrations and less productive concentrations. This should be researched further, perhaps next year’s REU could explore this further.
Visualization:
A big part of science
research is presentation of data. In simulation based research this usually
involves some sort of movie or gif. In
the beginning of the project the main concern was writing code and
upgrading/debugging it. However, as the code took shape and I became confident
in its accuracy, presenting and visualizing the simulation became very
important. In the first few weeks I would make pictures out of the data
outputs. Then I found Virtual Molecular Dyanmics
(VMD) which is a nice software for visualizing
simulations like this one. It took a while to write code that would not
interrupt the simulation and interface with VMD to produce simulations.
Eventually interfacing between the two became reality and simulations could now
be presented visually.
Background photo: http://martin-protean.com/protein-structure.html