Bench Press

IT Development Director Luke Bracegirdle demonstrates a virtual patient scenario.

After Ben’s post the other day about “the Doctor” from Star Trek Voyager it made me remember a news article I had read at ScienceDaily.com.

While in Star Trek Voyager a computer program was used to play the physician’s role, at Keele University in the UK they’ve developed “virtual patients” to train future pharmacists. These “virtual patients” are used to teach proper techniques in communicating and diagnosing patients. The overall program is called the Virtual Consultancy Program and focuses on providing digital avatar based training that rivals that of one on one interview training without the resource constraints of having faculty work one on one with every student.

The “virtual patients” behave quite believably throughout the example scenario displayed by the Keele University School of Pharmacy site. Most remarkably the program reacted quickly and realistically to questions spoken into a headset. The versatility and seeming ease of use makes this training tool particularly impressive.

What most impressed me most, was the potential that this “virtual patient” program illustrates. Keele University’s Virtual Consultancy Program ultimately uses computers to model certain human interactions in order to provide a learning tool for future pharamacists. Part of the parameters within the program include the patient’s basic physical statistics, symptoms, and even allergies in order to provide effective tests and examples for students. It seems to me that this software if developed properly could eventually be very useful in creating models with which doctors can practice and explore diagnosing difficult diseases/conditions.

It would be amazing to be able to model effects of various diseases within a “virtual patient” to determine the potential outcomes for a real patient. Risks could then be accurately assessed providing doctors with another method for screening potential therapies in various situations as well as potentially developing new diagnostic procedures. While I’m sure a lot of work would need to be done to produce a “virtual patient” of that sophistication not only with regards to software, but medical understanding as well. It’s nice to think about the possibilities.

(Read more – Virtual Consultancy Program)

My favorite character on Star Trek Voyager is “the Doctor” (pictured on the left and portrayed by the very talented Robert Picardo), who despite being “merely” a computer program was able to diagnose and treat nearly any medical ailment. The best part about him was that, because he was “merely” a hologram, he was portable and able to travel through the harshest environments and across any terrain.

While I’m pretty sure we still have years to go before we start being treated by medical holograms with Robert Picardo’s sense of style and humor, the portability of diagnosis is something which we may have just taken one large step closer towards. Christine Keating’s group at Pennsylvania State University has just developed what I’ve dubbed a “Doctor-on-a-Chip” (or DoC, after all system-on-a-chip’s are called SoCs by the semiconductor industry) which have the capability to detect any number of viral pathogens on a single computer chip.

Keating and her colleagues developed a means of coating a chip with nanowires (small wires 8 micrometers long and 300 nanometers in diameter) coated with DNA strands complementary to viral genomes (so that they will bind to viral DNA/RNA if given a chance). But, instead of haphazardly coating the chip, Keating’s group was able to develop a precise, targeted method, employing electrical fields to position the nanowires to exactly where the researchers wanted them (here’s a video showing how the method works).

The picture to the left is almost kind of eerie to me in its precision, much like one would expect on a chip fabbed by Intel or IBM (or Affymetrix) which uses photolithography rather than through electrophoresis.

Impressive to say the least, but the big question remained – does it work? To test this, Keating’s group incubated the chip with suspensions of fluorescently tagged viral DNA fragments complementary to the DNA strands on the nanowires, removed the suspension, and then subjected the chip to fluorescence. What would the result be? Would one see fluorescence organized neatly in the same rows that the nanowires were deposited? Or would there be diffuse or no fluorescence, suggesting nothing at all?

See for yourself:

So, do we have a DoC in the making? Jury’s still out – as we have yet to see if this method can be scaled up, or if its even applicable in a medical setting where time is short, accuracy needs to be very high, and the ability to run controlled samples (e.g. long DNA binding period with perfect fluorescently-labeled viral DNA fragments) is hindered. But, the Keating group is already hard at work creating electrical leads which will enable a faster (and potentially more quantitative) read process for detection.

And who knows, in a few years, this may end up looking like Robert Picardo.

(Image source – Doctor) (remainder of images come from Penn State University public image archive)

Morrow, T. Li, M. Kim, J. Mayer, T. Keating, C; Programmed Assembly of DNA-Coated Nanowire Devices; Science 16 January 2009: Vol. 323. no. 5912, p. 352 DOI: 10.1126/science.1165921

Having read and watched a fair amount of science fiction during my formative years, I’ve always been enamored with the concept of nanobots being used within the human body to help maintain our health. While the reality of microscopic machines moving to normally inaccessible areas of the human body and performing life saving tasks is still decades away, researchers led by Professor James Friend at Australia’s Monash University have demonstrated a proof of concept piezoelectric ultrasonic motor that could be a crucial step in providing locomotion for the nanosurgical robots of tomorrow.

In the abstract of their paper they describe their concept as:

A motor for in vivo microbot propulsion is presented with a stator diameter of 250 µm, demonstrating the potential to directly drive a flagellum for swimming at up to 1295 rpm with a torque of 13 nN m. The motor uses coupled axial-torsional vibration at 652–682 kHz in a helically cut structure excited by a thickness-polarized piezoelectric element.

Piezoelectric ultrasonic motors like the one designed by the team are built to harness special materials that exhibit the piezoelectric effect. Materials like lead zirconate titanate, thanks to the piezoelectric effect, are capable of producing electricity when stress is applied as well as the converse, producing stress when an electrical field is applied. This effect is already used in a variety of everyday applications, such as electric guitar pickups as well as auto-focus in reflex cameras. The motor (seen below), called the Proteus, designed by Friend’s team is made up of three main components: the piezoelectric element, the stator, and the rotor.

The team began to design the Proteus by utilizing computer models (seen below) to produce a novel robust stator design. They determined a helically cut stator would serve best in transferring stress and turning the rotor. The group hopes to improve their model in future work by incorporating further criteria and motor components.

After utilizing computer models to design the stator a prototype was fabricated for physical tests. Various tests were run to ascertain the potential of the Proteus. The group’s tests are promising as they found the “output power [is] on the order of what is necessary to navigate small human arteries.” In addition to the promising power output, the stator design is currently “70% smaller than the smallest design produced so far”.

While the design of a small motor is still just a small first step towards in vivo swimming surgical nanobot I can’t help but think about the possibilities this innovation will lead to. Maybe Professor Friend can contact Dr. Gracias at John Hopkins about working on a swimming microgripper next.

All graphics and quotes from:

B Watson, J Friend and L Yeo 2009. Piezoelectric ultrasonic resonant motor with stator diameter less than 250 µm: the Proteus motor. Journal of Micromechanics and Microengineering.

We’ve argued before in favor of increasing our use of computer modeling to enhance our ability to understand complex scientific problems, design new technologies, and make smart decisions. But, the subject of this post is a whole different ball game. We’re not talking about systems biology or design. We’re talking about armageddon.

In December 2004, astronomers discovered an asteroid designated 2004 MN4. Now, I’m fairly certain that, under normal circumstances, this would have just been recorded and noted, and the world would have happily moved on, never thinking twice about 2004 MN4.

Of course, the fact that I’m talking about it now, almost 5 years later, suggests that there was nothing normal about these circumstances. The basic story is that computer modeling of the trajectory of 2004 MN4, later called 99942 Apophis (after the Egyptian serpent god known as “The Uncreator”) revealed that this near-earth asteroid had a score of 2 on the Torino scale, a number which calculates how much we should worry about a particular asteroid/comet, the highest score of any near-earth object ever (as far as I know). And what does a 2 mean? According to the Torino scale, a 2 is “merely”:

A discovery, which may become routine with expanded searches, of an object making a somewhat close but not highly unusual pass near the Earth. While meriting attention by astronomers, there is no cause for public attention or public concern as an actual collision is very unlikely. New telescopic observations very likely will lead to re-assignment to Level 0 [no risk]

What the computer modeling showed was that there was a range of trajectories that the asteroid could take (the range stemming from a number of uncertainties), and a few of them could hit Earth (at a probability of about 3%) in 2029.

The result of the impact? As Apophis is roughly the size of the Rose Bowl, it has been estimated that upon impacting the Pacific Ocean (where its currently estimated to be if it does hit), Apophis would cause a hole in the ocean roughly 3 miles deep and 3 miles wide, which would then follow with a series of tidal waves so destructive that it would eradicate any cities and regions unfortunate enough to be on the Pacific.

Thankfully, subsequent modeling involving more data and more calculations showed that the probability of hitting the Earth dwindled to near 0% (1 in ~45000 to be more precise) in 2029, although interestingly enough, when the asteroid passes by the Earth in 2029, it has been predicted to pass close enough that it actually dips below the altitude at which many satellites orbit (causing a bizarre NASA-German student debate over the probability of impact increasing because of a collision with a satellite).

However, we are not out of the woods yet. Just because 2029 is relatively safe, doesn’t mean that 2036 is. It turns out that Apophis has a troublesome gravitational keyhole. If a near-earth asteroid happens to pass through this gravitational keyhole, a very narrow region of space which in the case of Apophis is an area roughly 2000 feet in diameter, then the Earth’s gravitational field will actually deflect the asteroid’s orbit almost guaranteeing that Apophis will hit the earth in April, 2036.

So, the two big questions for humanity (and a third about science):

1. Will Apophis pass through its gravitational keyhole when it swings by the Earth in 2029?
2. If yes, what hope does humanity have of deflecting said asteroid?

The answer to both will require extensive computer modeling. The challenge of the first question is particularly intensive as the gravitational keyhole (which we have a good sense of) is so small, and yet the number of potential influences on Apophis’s trajectory is so large. Not only does one need to deal with the gravitational pull of the Earth, Moon, and Sun, one has to factor in things like the spin of the asteroid, the asteroid’s ability to absorb and reflect sunlight, and even the gravitational pull of other near-earth asteroids!

The challenge of the second question is not only technological (do we use nuclear weapons? rockets? lasers? can we coat the asteroid with a reflective material to change its absorption of sunlight?), but also one of modeling. This one is especially challenging, as even assuming that mankind is able to deflect the asteroid (and not just shatter it, leaving many many little asteroids to hit the Earth), various space agencies will likely have to continue tracking Apophis to make sure that the deflection did not cause the asteroid to change course to potentially hit the Earth again.

Do we have the computer technology and the mathematical wisdom to solve these questions? I sure hope so. 2029’s not all that far off…

(Image credit)

While reading through VentureBeat.com, I chanced upon an article written by Dean Takahashi which caught my interest. Apparently, a team of IBM researchers have developed a microscope with 100 million times the resolution of a conventional MRI. The secret to this amazing success? Magnetic resonance force microscopy (MRFM). Combining the ideas of magnetic resonance imaging and atomic force microscopy, MRFM is sensitive enough to detect the magnetic spin of a single electron and thus, provides a substantially greater resolving power.

From IBM’s abstract:

We have combined ultrasensitive magnetic resonance force microscopy (MRFM) with 3D image reconstruction to achieve magnetic resonance imaging (MRI) with resolution <10 nm. The image reconstruction converts measured magnetic force data into a 3D map of nuclear spin density, taking advantage of the unique characteristics of the “resonant slice” that is projected outward from a nanoscale magnetic tip. The basic principles are demonstrated by imaging the 1H spin density within individual tobacco mosaic virus particles sitting on a nanometer-thick layer of adsorbed hydrocarbons.

While we have seen other exciting developments in the MR industry as well as the imaging industry, this breakthrough is especially revolutionary because of how much it influences the scientific community. If IBM’s new microscope is as good as advertised, we will be able to produce three dimensional images of viruses, view the structure and interactions of proteins, and study the physical nature of certain chemical reactions, all while evading the disadvantages which plague electron microscopy. The benefits of understanding how things work at a molecular level can lead to better modeling, better drugs, smaller chips, and maybe even better detection mediums for cancer.

Here’s a video describing the technique:

One thing great about basic biology research is that so much cool and inventive technology is built around trying to peer into the depths of how organs, tissues, and cells work their mojo. A lot of times, the best stuff is built around microscopy and imaging, because seeing is pretty much believing. And beautiful pictures make scientists just as happy as it does everyone else.

There’s an imaging technique that’s been emerging over the last ten or so years called imaging mass spectrometry, which can tell you where a whole bunch of different molecules are in tissue all in one go! Basically, you take a slice of the tissue you want to image, apply a chemical resin to it (which is called a “matrix”), and then you shine a fairly strong laser at the tissue. This causes a physical reaction in which the resin absorbs the laser, ionizes, and causes molecules from the tissue to get ejected into the air. The mass spectrometer then sucks those ejected molecules into its cavernous depths and weighs them, breaks them apart, and then weighs their fragments to try to identify what the molecules are. For proteins, this identification process is actually quite easily done.

What you end up with is a really remarkable picture, a 2D map of where a range of different molecules are in the slice of tissue. You can look at the distribution of molecules of different specific weights across the entire slice. By stacking images of adjacent slices, you can then reconstruct an approximate 3D map of where molecules are in a tissue, like a brain or a tumor. Pretty neat!

A section of a mouse brain imaged for molecules of various weight (from Stoeckli et al. (2001) Nature Medicine 7, 493–496.)

Up until now, there have been two main ways to see macromolecules like proteins and RNA in biological tissue. First, you can look at where a specific macromolecule is by sticking a fluorescent dye to it and looking for that “tag.” It’s a great technique, but it’s also pretty laborious and limited in scope, because you can only examine at most three or four things at a time or the colors will start overlapping together a little too much to reliably distinguish what you’re looking at. In addition, you have to know what molecules you’re interested in first in order to tag them.

Second, you can look at cells as a whole using high energy techniques like transmission electron microscopy (TEM), which uses beams of electrons to “X-ray” cells. Since electrons have a very short wavelength, that means that they can be blocked and scattered even by very small structures, including groups of proteins, which means that a TEM image can look at a lot of things with an insanely high amount of detail. Unfortunately, electron microscopy won’t really tell you what you’re looking at, though there are a few techniques (like sticking gold beads to stuff you’re interested in) that try to get around that. Nonetheless, it doesn’t tell you much besides the overall “visual” picture.

Imaging mass spectrometry gets over a lot of these hurdles of traditional microscopy by directly associating molecular weights with specific coordinates in a tissue sample. And because the mass spectrum measures a wide range of weights, a computer algorithm can potentially tease out patterns that might not have occurred to a researcher or doctor looking for a few things at a time. The amount of information that can be obtained from a simple mass spectrum image is tremendous. With newer, cheaper super-accuracy mass spectrometers coming out, such as the orbitrap, this technology is growing in popularity outside of specialized physics labs, and it promises to revolutionize the way a lot of science and medicine is done. In the clinic, for example, a ton more information could come out of a biopsy all at once using this technology, from identifying cancer markers to positively identifying infectious agents and inflammation. I look forward to what’s coming in the future!

The sentiment expressed in the comic above isn’t just common when you’re absorbing your grandparents’ wisdom; but is something you often run across when talking with grizzled research veterans as well. Often times you’ll here reminiscent stories about the good ‘ol days, without complicated equipment and how scientists nowadays are soft and spoiled by kits and electronic pipettes. Today, Derek Lowe has a great post talking about the dangers of romanticizing ancient equipment and techniques.

Derek writes:

If you’re in a resource-limited situation, then, you’ll probably try to carefully pick out problems that can actually be well addressed with what you have. That’s a good strategy, but it’s not always a possible one. Huge areas of research can be marked off-limits by the lack of key pieces of equipment, and by the time you’ve worked out what’s possible, there may not be anything interesting or important left inside your fence. Medawar’s point was that being stuck inside such a perimeter would not only hurt the way that you did your work, but could eventually do damage to the way that you thought.

While learning about old techniques and how they contributed to previous breakthroughs can provide tremendous value by illustrating thought processes and helping to build a solid scientific foundation, Derek accurately points out that clinging to old methods just for the sake of doing it the “old fashioned way” can sometimes be detrimental to not only one’s work, but one’s thought process as well. In the end, make sure you don’t handicap your most precious resource by continuing the tradition of “walking uphill both ways”.

(Image Source – Bobwama’s Wallpaper of the Day)

At Bench Press, we tend to cover new technologies used to solve challenging scientific problems or enhance the quality of science — whether it be electronic lab notebooks, the use of computational modeling, distributed supercomputing, or new nano-sensor technology.

And, while these are all interesting and powerful applications of technology, we oftentimes forget the beautiful simple solutions. Roughly 158 years ago (plus or minus a day or two), Jean Benard Leon Foucault demonstrated a very simple and elegant proof that the Earth rotated on its axis. Of course, by then, I’m sure most scientists, having seen the movement of the stars and the sun in the sky, accepted this, but Foucault gave the first non-celestial visual proof of the earth’s rotation.

He used a pendulum.

A more rigorous explanation of how this experiment works is on Wikipedia, but in short, if you were to hang a pendulum from the North Pole (or the South Pole), you’d expect to see the oscillations gradually move as the planet rotated. It’s not that the pendulum’s rotation is changing (relative to someone from outer space), it’s that the earth is rotating underneath the pendulum. This effect becomes weaker the further away from the poles you move (and Foucault derived a mathematical rule which explains this, as described in the Wikipedia entry).

This brilliant insight led Foucault to invite the scientific community in February 1851 “to see the Earth turn” — hoisting up a massive 62-pound brass sphere with a marker on the end to trace the pendulum’s path as it went back and forth. 

So, yes, we may beam with pride today at our amazing, cutting-edge technology, but I personally am far more impressed (and hope the scientific community is too) by the scientist who needs only an elegant demo/prop to make the same conclusion.

(Image source)

Although digital modeling may eventually lead to the creation of the Matrix, I think it's safe to take our chances.

The digital age is upon us! SLR cameras are now DSLRs, VHS tapes are now Blu-ray discs, CRT monitors are now OLED screens, and every two years, our chips double in computational power (as predicted by Moore’s Law). With all these rapid advancements in technology, I believe the next step will be a dramatic increase in the use of computer modeling.

We see the signs everywhere. More and more, people are relying on computers to help them gather and process information. But why stop there? Why not have computers be the primary engine of design and testing? Obviously it’d be remiss to ignore the human factor in completing any project, at least until the apocalypse foretold by the Matrix or the Terminator series comes true. However, I’m convinced that the more we incorporate computer programs into our decisions and our design flows, the better, for three basic reasons:

1. Objectivity: To quote Dr. Gregory House, “people always lie.” I don’t mean this in a cynical way, but humans “lie” even when they don’t intend to. We’ve all heard of the placebo affect and how manipulable and inconsistent human memory is. Computers, on the other hand, aren’t designed to lie or to be irrational. Give a computer program an input, and you can rest assured that it will faithfully return the output it calculates is best. No matter what the circumstance, you can rely on computer programs to provide unbiased, objective results.
2. Consistency: Try shooting 1000 free throws in a row. Impossible? Maybe, for a human (unless you’re Steve Nash). However, program a robot to shoot a basketball the same way, with the same stroke, and the same force, and you’re looking at a free-throw shooting machine (literally). Computers only know how to do what they’re programmed to do. Given an algorithm, a computer will duplicate it consistently, no matter how many times the user asks.
3. Speed: How fast can you calculate the roots of the polynomial 23x³ + 17x² – 31x + 71? With advancements in transistor speed and computing algorithm, the complexity of problems a computer can solve (and the speed at which they can be solved) is astounding — and growing. Some calculations that are considered intractable for humans are done in seconds on a computer.

These three points underscore my main argument: we shouldn’t use computers only when it’s necessary, but whenever it’s possible. Computers are fast, efficient, and reliable. With careful ingenuity and good programming, there’s no limit to what computers can accomplish. That’s why, to me, the ability to model things digitally is such an appealing prospect. The benefits include:

1. Thorough testing: As a computer scientist myself, I can attest to the fact that there is no such thing as too much testing, and especially when it comes to safety concerns. Using computers, builders of things ranging from automobiles to buildings will be able to run many tests on a product or a feature before release and, potentially, before a single tool is touched. This enhances our ability to design effective and safe products.
2. Lower costs: Instead of spending millions on supplies and materials for multiple physical prototypes, virtual trials will enable cheap testing of new designs and features. Obviously, physical trials cannot (and should not) be omitted, but digital models which are intrinsically easily re-tested and modified, are clearly the cost-effective approach to designing/testing/tweaking. Ask yourself what’s cheaper: designing four prototypes to test four possible features, or running four computer models of the different features and building out only two prototypes with the two best features?
3. Better understanding of design: Some scientists joke that engineers can only understand things that they can take apart and put it back together. While that subject is worth a post on its own, it’s hard to argue that one level of understanding comes from being able to model. Comparing these computer models to reality and seeing the differences, then, allows the engineer/designer to test their understanding. A product which succeeds on the computer monitor but fails in reality suggests that there is some aspect of the design that isn’t being modeled correctly — and that insight not only furthers scientific understanding, but can also help designers quickly design the next product or feature.

This is the power of virtual modeling. Using our computers to do the dirty work while we reap the benefits of inexpensive and efficient computing. We’ve already seen glimpses of this happening, and I believe that these trends will not only continue, but are good things for us to embrace. Imagine the FDA requiring drug trials to be first run in a computer model: testing its efficacy on a simulation of a human body. Imagine regulations requiring architects and contractors to digitally test the earthquake safety of a building before they even lay the first beam. The possibilities, and possible benefits, are endless.

To reach that end stage, I see several things needing to happen:

1. Greater emphasis on developing good mathematical models: In order to model things rigorously on a computer, we need a solid understanding of how things work, and we need to ingrain the “art” of computer modeling into a new generation of designers and engineers. Computer models should not be a luxury that only specialty shops engage in, they should be the norm, and something every architect, engineer, and product designer is aware of.
2. Developing better computing technologies: Although processors today are operating at unprecedented speeds, in order to realize large-scale digital modeling, we still need our computers to be faster and more efficient. This may mean designing faster computer chips (or using GPUs and/or gaming processors) or looking into innovative new technologies such as distributed computing.
3. Developing/sharing numerical computing algorithms: While faster computers means being able to physically process more data, this is only half of the equation. On the software side, computer scientists still need to find and deploy innovative solutions to solving these challenging modeling problems or else no amount of computing power will ever be enough. Hopefully, the scientific computing community will move to develop and share new (ideally open source) innovations oriented around solving these computing challenges, and an open environment of collaboration will drive the greatest innovation and adoption of these techniques.

If you have any thoughts on this subject, would love to hear them in the comments!