Data Miners Excavate Human Brain

Neuroscientists mining data have struck algorithmic gold. The advancement has opened a new realm of possibilities, allowing researchers to model neural properties given a limited amount of data.

The breakthrough happened when scientists at the Ecole Polytechnique Fédérale de Lausanne (EPFL) discovered rules relating to how neurons function in the brain. They thought it was possible to predict neuron functionality from a small set of information, so they with data from a rat’s brain to draw their conclusions.

The researchers “recorded the expression of 26 genes encoding ion channels in different neuronal types from the rat brain. They also had data classifying those types according to a neuron’s morphology (form), its electrophysiological properties, and its position within the six anatomically distinct layers of the cortex.”

Through their study, scientists were able to predict ion channel patterns with 78 percent accuracy using classification data. Their predictive accuracy increased to 87 percent after adding ion channel information into their data mining program. 

This revelation has delivered a scientific shortcut through the use of collected data. Felix Schürmann, an author of the study explained, “Using the methods we have developed, it may not be necessary to measure every single aspect of the behavior you’re interested in.” The EPFLs research means that data generated from a single neuron could potentially lead to predictions about its behavior, morphology and gene expressions.

Although not specifically mentioned, this research may be part of the Blue Brain Project, a program carried out by Henry Makram at the EPFL. The research employs a 16,384 core BlueGene/P supercomputer to model the cortical column from a rat brain. The research requires a lot of compute power as one graphic estimates a petaflops/s system would be needed to model a full rat brain.

Algorithms predicting the functionality of neurons may become an integral part of neuro research. Makram is also behind the Human Brain Project (HBP), which has faced skepticism over his endeavor to simulate the entire human brain. A number of his colleagues believed the research too impractical, requiring exascale compute power and a $1.3 billion price tag. However, his project may seem less far-fetched and become more cost effective through the incorporation of predictive analytics.

As this and a series of other neuroscience-related stories over the last few yeras reveal, algorithms created by researchers at EPFL may result in key tools to understand human behavior or combat degenerative neurological disorders like Parkinson’s and Alzhiemer’s.

Related Stories

SSDs and the New Scientific Revolution

Spinning Data into Super Plants

Business Intelligence Gets a Chemistry Lesson