Pattern recognition spans several disciplines

Professors take the podium in the ‘Pattern Recognition Across Disciplines’ colloquium. (Photo by Brian Kempf)

By Brian Kempf
Correspondent

What exactly is a pattern? Why do genetically similar organisms respond differently to external stimuli? How can manipulating proteins stem the onslaught of gypsy moths? The School of Science’s colloquium, “Pattern Recognition Across Disciplines,” answered these questions and more, providing an exciting look at projects on which the faculty have been working, as well as possible applications for their research.

Andrea Salgian, assistant professor of computer science, was the first to take the podium. Quoting physicist Satosi Watanabe, she noted that a pattern isn’t merely a regular repeated structure, but “a pattern is the opposite of chaos.” Patterns are sorted into classes and thus recognized, either by computers or humans. Pattern recognition technology is all around us, from Roombas to Facebook photo tagging. To do so, computers must emulate human vision: recognizing objects and tracking their movement over time. A 3-D object must be converted to a two — dimensional matrix of numbers — then an image is produced.

There are two approaches to image recognition: supervised learning (classification), which features information put in by hand, or unsupervised learning (clustering), which would be finding patterns in existing data that can be used to label new data. But there are difficulties of computer vision.

For example, a cake looks different from the top than it does from the side, but it is easy to surmise that it is indeed a cake. A computer can’t pick up on that unless it is told to do so. Or, what if the cake was in a dimly lit room, and then the lights were turned up? The computer may not be able to recognize that it is the same object because the colors are different. Any number of factors conspire to make image recognition no piece of cake, so to speak.

Salgian gave an example of a particularly interesting application of computer recognition. By tracking an orchestra conductor’s hand movements, it can be inferred where beats occurred in the musical piece. Thus, by extracting patterns from the musical score, information can be generated to have an artificial conductor. This set the stage to talk about the central theme of the colloquium — analyzing the construct of Cytochrome P450’s, or in layman’s terms, the very structures of the catalysts that makes life possible.

Leann Thornton, assistant professor of biology, then took the stage to discuss patterns in Cytochrome P450.  Thornton started her lecture anecdotally. “Why is it that when I take caffeine I get all jittery, but my very genetically-similar brother can take a nap?”  Thornton said.
Or why do two of the same kinds of plants grow differently? The answers lie in DNA, which, among other jobs, sends messages to ribosomes about how to encode protein, which are built from chains of patterns of amino acids. How these amino acids — and thus, enzymes — are structured is of the utmost importance.

On the molecular level, even the most minute change in structure can have the widest of implications. Two similarly designed genes would have two different purposes: one is a plant growth hormone, the other is the gene that tells the plant to stand tall. By isolating genes, they can be manipulated in order to modify a plant’s behavior or development. Thornton concluded with the fact that plant growth depends on enzymes that regulate a plethora of chemicals, and that the smallest of changes to these patterns results in different Cytochrome activity.

Stephanie Sen, professor of chemistry, was the last to take the stage, speaking in the context of protein function. After mentioning that the two proteins in question (CYP734A1 and CYP72C1) are part of the same biosynthetic pathway, she asked, “Why are we interested in this?”
Sen’s research is concerned with using these two proteins to block isoprenoid production in insects, in other words, blocking the production of the necessary natural products needed for insect development. By manipulating these genes, the gypsy moth invasions and pests that continually wreck agriculture could be a thing of the past.

According to Sen, if the research is picked up, a tangible application may be only a few years away, and that natural events such as deforestation and crops damaged by pests could be mitigated by halting development in insects altogether. Of course, with 40,000 isoprenoids all being synthesized by the same pathways, this promises to be a Leviathan of a task, yet one that be accomplished using pattern recognition.