Graduate students at Carnegie Mellon University will spend 24 hours this week trying to “hack” the brain using big data.
As part of CMU’s first Neurohacakathon, students will be looking for patterns in large datasets culled from experiments conducted across campus.
“(The datasets are) quite varied, from gene expression to cell activity to electrical activity to functional anatomy,” said Alison Barth, professor of biological sciences and interim director of the university’s multi-disciplinary brain research center, BrainHub.
Barth said many people think of neuroscience as dependent on imaging and working directly with neurons, but that CMU has a much broader definition of brain science.
“It involves people who are experts in statistical analysis of neural data, robotics, computer science and network analysis, deep learning (and) algorithmic approaches to understanding how we can train machines to learn as well as biological approaches or more traditional psychology approaches to understanding the brain,” she said.
Jay Hennig, a first-year graduate student in neural computation, is interested in finding out if there is a way to determine cell type – sensory or motor neurons, for example – based on recorded signals released from those cells.
“There are so many cells in the brain that you need statistics to study exactly how their properties relate to behavior,” he said. “Cells fire in different ways even given the exact same stimulus, so it’s a really cool statistics problem.”
Those particular data come from mice; researchers in engineering and biology have recorded the neural signals emitted when mice are at rest and then start running.
Barth said Hennig’s question is just one of many ways to analyze the dataset.
“There’s a lot of structure to that signal that we don’t understand,” she said. “How is it initiated? Which cells? What is the transition in the kind of activity when an animal is resting quietly and then the kind of activity when the animal begins this running motion?”
A grand prize of one semester’s tuition will be awarded to one student at the end of the Neurohackathon. Barth said judges will be looking for a novel computational analysis that addresses an important question and could be applicable to other datasets. That student will then be expected to spend the next semester continuing the research begun at the event.
But for Hennig, it’s not about the top prize. He said forgoing sleep to analyze data for 24 hours sounds like a “fun idea.”
“It seems kind of exciting,” he said.
The Neurohackathon begins with at noon on Tuesday with remarks from a representative of the event’s sponsor, Qualcomm. Faculty will then present the datasets and problems and teams will get to work. The top prize will be awarded Wednesday evening.
