Future Reflections         Summer 2011

(back) (contents) (next)

Drawing from Experience: Picturing Molecules with Sound

by Tim Cordes, MD, PhD

Psychiatry resident Dr. Tim Cordes, walking with his guide dog in a hospital corridor, talks with a physician and a medical student as the group does a morning round of patients. [Photo courtesy of photographer Jeff Miller/UW-Madison]From the Editor: Dr. Tim Cordes is a practicing psychiatrist in Madison, Wisconsin. As a graduate student he developed a unique, nonvisual method for creating and understanding images of complex molecular structures. In this article he describes the method he devised.

While I was attending graduate school at the University of Wisconsin, I studied biochemistry in the laboratory of Katrina Forest. Biochemistry is the study of the atoms and molecules that make up living systems, whether they are plants, animals, or bacteria. In living organisms, proteins are the molecular structures that do the heavy lifting. They perform a host of functions, from making our fingernails hard to digesting the meals we eat. Each protein has a unique three-dimensional shape that places each specific atom where it can do the job it needs to do. The shape of a protein determines its function, and by figuring out the structure, scientists can learn how the protein works.

The standard way to show the structures of molecules is through pictures. Since I am blind, however, visual representations were not available to me. As I thought about alternative methods I might use, I reflected on the fact that no one can actually see a molecule. All but the most gigantic molecules are smaller than the wavelength of light. For all practical purposes, a single molecule is invisible. A picture is simply a visual way to convey information about a molecule's structure. Other ways of presenting this information, such as tactile diagrams or three-dimensional models, are just as valid as pictures if they enable someone to learn.

My lab used a technique called X-ray crystallography to compute the structure of the protein I was studying, which was named Virulence Factor Regulator. When I began to examine the molecule's structure, I started with off-the-shelf software. Using a screen reader, I wrote short scripts, or sets of instructions, to list where particular atoms were positioned in space. I saved these lists to textfiles and read them using the computer's editor function. In this way I began to learn about my protein, puzzling over which atoms were near each other and how they might interact. When I taught my lab mates, I did the reverse. I wrote scripts that selected certain atoms and controlled their graphical presentation. The process was cumbersome, and I clearly needed a simpler, faster method.

How could I explore which atoms were near each other in space and learn about the bonds and relationships between them? How could I learn all of this without the tedium of writing and reading dozens of textfiles? Part of the solution came when I realized that I was dealing with a database problem. I began by writing a program that made a database of where the atoms were in space. I realized that I needed to move a virtual "box" within that space, figure out which atoms fell into it, and display its contents in a text format. I built in the ability to use the keyboard to scroll through three-dimensional space, much as I might use a screen reader's review cursor.

My program was helpful for a local neighborhood of the molecule, but I needed a better sense of the overall structure. I had heard about a blind chemist who developed a system that turned the two-dimensional line graphs of infrared spectra into audible tones. Perhaps I could develop a way to listen to molecules.

As a teenager, I had learned to play keyboards and compose electronic music. Until now, the high point of my musical career was a Guns n' Roses song that my friends and I performed in our high school talent show, featuring my keyboard playing and screeching vocals. Now I reached back to my keyboard skills for a means to produce audio tones. I wanted to avoid the complexity of surround sound programming or complicated speaker arrangements. I needed a system that could be accessed using simple headphones. I decided that MIDI music would be perfect.

Dr. Tim Cordes shares his thoughts about a particular case while meeting with a team of physicians and medical staff at the UW Hospital and Clinics. [Photo courtesy of photographer Jeff Miller/UW-Madison] Three-dimensional space has three axes--one running left and right, another going up and down, and the third going forward and back. I decided to represent the left to right dimension by sending sounds to the left and right ears. I made the pitches rise as objects went upward, and made the tones quieter as they moved in the forward direction. I used different instruments to represent each type of atom--a piano for carbon, an organ for oxygen, etc. A colleague suggested that I build in the option to step through the backbone of the protein, playing notes along the way.  This was a nice addition that gave each protein its signature song. The same colleague also gave my program its signature name, cheerfully calling it TIMMol (pronounced tim-mole) from early in its development.

With TIMMol I could better conceive of my protein, but I still couldn't show others what I was listening to. Text output and audio tones worked well for me, but my sighted colleagues were used to the standard graphical displays. I joined forces with Britt Carlson, a fellow graduate student in biochemistry, who had an interest in education. Routines to display molecules graphically were freely available on the Web. I grafted these onto TIMMol and, with Britt's input, tweaked the visual presentation step by step. I would make a change and send her the new program. She would examine my work and give me feedback. For example, to show where the audio cursor was, we settled on a sphere that looked like it was made of spiderwebs. Now I could show people visually what I was looking at with my audio program.

Britt and I believed in the program, but could students use it effectively? We gathered nine volunteers from our laboratories to find out. We gave them a tutorial, let them practice, and then set them loose. We disabled the graphical display on their computers and asked them to perform a variety of tasks such as identifying the general shape of a piece of a molecule and figuring out which atoms held a metal ion in place. The students showed success in many of these areas, and it is likely that their abilities would improve with practice. Interestingly, we noticed that people who had the least experience using the standard graphical models had the most success with our program. I speculate that they weren't yet locked into thinking about structure in conventional ways. This discovery was very encouraging. Beginning students, and especially students who are blind or visually impaired, would probably have little experience with the conventional tools. Given our success, we published a paper describing TIMMol in a journal called Biochemistry and Molecular Biology Education. To my delight I learned that the editor who ultimately accepted our piece had written the textbook I used when I first learned about biochemistry at the University of Notre Dame.

TIMMol helped me better understand the properties of the particular protein I was studying. However, the ideas behind TIMMol reach far beyond protein structures. Because the source code can be easily modified, the framework of TIMMol could be used to convert almost any three-dimensional data into sound. Uses for this system could range from helping blind people learn the layout of a multilevel airport to letting them inspect MRI scans. Beyond that, TIMMol shows that, when given the chance, a person can meet a challenge by mobilizing tools from his or her life experience. The solution may come in a surprising form, one that can be shared for the benefit of others.

(back) (contents) (next)