--- /dev/null
+\section{Hardware}
+\label{sec:future:hardware}
+
+Very few approach--bind--retract cycles actually pick up a protein and
+produce a clean unfolding curve. This makes it hard to gather
+sufficient unfolding statistics unless you can run experiments
+continuously over a long time span. While our modified
+MultiMode\citep{multimode} gets the job done, some hardware upgrades
+would allow futher automation, increasing throughput through longer
+run times.
+
+MultiModes measure the position of the cantilever my monitoring the
+reflected laser with a four-segment photodiode. While the vertical
+and horizontal signals are accessible in the cable connecting the
+MultiMode to its controlling NanoScope, the total photodiode signal is
+not. The loss of laser signal---which can occur when bubbles in the
+fluid cell obstruct the laser---results in low voltage deflection that
+is independent of piezo position. This flat-line deflection also
+occurs when mechanical drift moves the surface out of range for the
+piezo positioner. In the drift case, we would like to use the stepper
+motor to reduce the tip--surface separation. In the loss-of-signal
+case, we would like to \emph{increase} the tip--surface separation to
+avoid accidentally crashing the tip into a surface we can no longer
+detect. By exposing the total photodiode signal to the control
+software, we could unambiguously distinguish these two cases. This
+would allow for longer runs, aggressively using the stepper motor to
+mitigate mechanical drift.
+
+We could also reduce deflection signal noise---which is especially
+important for accurate cantilever calibration
+(\cref{sec:calibcant})---by automating photodiode positioning. The
+four-segment photodiode has the least signal noise when the deflected
+laser lands near the point between all four sections. However,
+mechanical drift in microscope alignment causes the spot location to
+vary with time. We currently use manual thumbscrews to re-zero the
+photodiode as needed, but unmonitored overnight runs would require
+computer-controlled positioning. Similar automatic positioning would
+be useful for automatically aligning the incoming laser with the
+cantilever. While laser--cantilever alignment seems to be less
+sensitive than cantilever--photodiode alignment, automatic laser
+alignment would also open the door to automatic piezo calibration
+through measurements of laser interference patterns\citep{jaschke95}.
+
+Finally, our current hardware does not address potential piezo
+hysteresis, nonlinearity, or drift. Newer piezos often use capacitive
+feedback, adjusting the driving voltage as needed to maintain the
+target extension. Besides making existing distance measurements more
+accurate, the increased stability opens the door to slower pulling
+speeds needed to monitor proteins with less stable folded positions.
--- /dev/null
+@string{Bruker = "Bruker"}
+
+@misc{ multimode,
+ author = Bruker,
+ title = {{M}ulti{M}ode atomic force microscope with a NanoScope
+ controller},
+ note = {This microscope was originally developed by
+ \href{http://www.di.com/}{Digital Instruments} (DI). DI was
+ aquired by \href{http://www.veeco.com/}{Veeco} in February of
+ 1998, and passed off to
+ \href{http://www.bruker-axs.com/multimode_8_scanning_probe_microscope.html}{Bruker}
+ in August 2010},
+}
\chapter{Conclusions and future work}
\label{sec:future}
+
+\input{future/overview}
+\input{future/salt}
+\input{future/hardware}
+\input{future/software}
--- /dev/null
+Single molecule force spectroscopy (SMFS) provides an experimental
+window the mechanics and kinetics of individual molecules and domains.
+Single molecule measurements extend bulk measurements, by offering
+insight into the variations within a population of molecules in
+addition to insight into the aggregate behavior of bulk solutions.
+They also bridge the gap between chemists experimenting at the bulk
+level and theorists simulating at the amino-acid level. By providing
+data for comparison, SMFS lays the ground work for improving and
+validating all-atom protein simulations. These simulations can then
+be used in predictive biological applications such as high throughput
+drug screening. By developing open source experiment control and
+analysis software, I have made it easier for new labs to get started
+in this field and for existing labs to collaborate on critical tools.
+I validate my experiment and analysis suite by carring out new
+experiments showing that increased \CaCl\ concentration significantly
+decreases the stability of folded I27.
--- /dev/null
+\section{Salt}
+\label{sec:future:salt}
+
+As expected\citep{chauhan97,itkin11,zidar11}, increasing the ionic
+strength of the buffer did significantly decrease the unfolding force
+(folding stability) of I27. For labs with strong gene-splicing
+capability, it would be interesting to replace the glutamic acids
+involved in the major bonding (\cref{fig:I27:H-bonds}) with
+alternative groups to gauge the specificity of the effect.
+
+While the statistics are strong for the two concentrations we tested
+(standard PBS and PBS with an additional $0.5\U{M}$ \CaCl), it would
+be useful to study destabilization scaling over a range of
+concentrations. Carrying out these experiments over a range of
+pulling speeds with additional force clamp experiments would also
+increase confidence in the kinetic models used to summarize the data.
+SMFS is a low-throughput technique, so such an exponential increase in
+assembled data would be much easier with more reliable hardware.
--- /dev/null
+\section{Software}
+\label{sec:future:software}
+
+Open source experiment control is possible! Even for a small lab,
+with a single novice developer\footnote{
+ I started this project a bit of LabVIEW and Matlab experience, but
+ only a few days of Python from the physics department's ``Welcome to
+ Drexel'' boot camp. I stumbled across version control on my own,
+ after a year of maintaining a directory full of version-stampted
+ tarballs.
+
+}, building reasonable software on top of existing pieces is possible.
+After a significant investment in developing
+\sawsim\ (\cref{sec:sawsim}), the \pyafm\ stack
+(\cref{sec:pyafm,sec:calibcant}), and \Hooke\ (\cref{sec:hooke}), we
+have a complete experiment control and analysis suite for single
+molecule force spectroscopy. All of the software in the
+stack---including the existing libraries and systems layers that I've
+built on---is open source, so other labs are free to use, improve, and
+republish it as they see fit.
+
+As the body of existing science increases, new researchers must become
+at the same time more specialized and more interdisciplinary than
+their fore-bearers. With a relatively fixed undergraduate curriculum,
+new researchers cannot afford to spend time becoming experts in every
+field that bears on their research project. By pooling resources
+between labs, individual researchers can reduce time spent on generic
+tooling and increase time spent on their particular project.
+Experiment control, analysis, and simulation software is particularly
+amenable to community development, because the cost of sharing
+software between labs is minimal.
+
+Besides the low cost of transferring the data itself, the rise of
+distributed version control systems such as \citetalias{git} have
+reduced the administrative overhead of maintaining a project with many
+far-flung contributors. Researchers can automatically fetch and merge
+changes made by other groups, incorperating remote improvements. They
+can also commit and push local improvements, which are then available
+for remote researchers to incorperate. The version control systems
+and workflows that facilitate this cooperation scale well, from small
+projects with a single user, to huge projects like the Linux kernel
+with thousands of developers contributing to each release.
+
+Once the software used in a lab has been published, it is also easier
+to audit by others who may be skeptical of the summary published in a
+journal article. For example, resolving the confusion about the
+``Lorentzian'' (\cref{sec:calibcant:lorentzian}) would be trivial if
+\citet{florin95} had also published their explicit procudure for
+generating their figure. Do you think I'm not calibrating my
+cantilevers correctly? Feel free to dig through my code. Let me know
+if you find something wrong (or fix it and send me a patch!). Science
+is built on reproducible experiments and analysis, and open sourcce
+software allows you to explicitly specify your methods. With well
+organized code, the specification should be clear from a high-level,
+experiment-design choices down through low-level bit manipulation.
+
+Many researchers have not received formal training in software
+development best practices, so how do we bootstrap this transition to
+open source science? There is a wealth of documentation available
+online for self-teaching, and scientists have lots of experience
+reading technical writing in their own field. For those who are
+overwhelmed by the amount of available resources, organizations such
+as \href{http://http://software-carpentry.org/}{Software Carpentry}
+are actively reaching out to scientists with short boot camps to lay
+the ground work. Mastery of any subject takes a significant
+investment, but gaining a working level of knowledge in distributed
+version control should only take a few days\footnote{
+ Software Carpentry allocates half a day to take students from ``What
+ is version control?'' to being functioning \citetalias{git} users.
+}. The difficulty for the uninitiated is often not mastering the new
+tool or workflow, but learning that it exists at all. There are a
+number of papers highlighting best practices and tools that are good
+surveys for guiding future
+learing\citep{wilson06a,wilson06b,vandewalle09,aruliah12}.
project = "Cantilever Calibration"
}
+@article{ jaschke95,
+ author = MJaschke #" and "# HJButt,
+ title = {Height calibration of optical lever atomic force
+ microscopes by simple laser interferometry},
+ journal = RSI,
+ year = 1995,
+ volume = 66,
+ number = 2,
+ pages = {1258--1259},
+ publisher = AIP,
+ url = {http://rsi.aip.org/resource/1/rsinak/v66/i2/p1258_s1},
+ doi = {10.1063/1.1146018},
+ issn = {0034-6748},
+ keywords = {atomic force microscopy;calibration;interferometry;laser
+ beam applications;mirrors;spatial resolution},
+ abstract = {A new and simple interferometric method for height
+ calibration of AFM piezo scanners is presented. Except for a small
+ mirror no additional equipment is required since the fixed
+ wavelength of the laser diode is used as a calibration
+ standard. The calibration is appliable in the range between
+ several ten nm and several μm. Besides vertical calibration many
+ problems of piezo elements like hysteresis, nonlinearity, creep,
+ derating, etc. and their dependence on scan parameters or
+ temperature can be investigated.},
+}
+
@article { cao07,
author = YCao #" and "# MBalamurali #" and "# DSharma #" and "# HLi,
title = "A functional single-molecule binding assay via force spectroscopy",
calibcant/main,%
hooke/main,%
salt/main,%
+ future/main,%
packaging/main,%
figures/main,%
root}
projects / thesis.git / commitdiff