Finding a Neuron in a Haystack: Ramanathan lab develops new strategy for assessing causality in complex networks

December 20, 2018
Needle in a haystack

How do cells give rise to behavior? This fundamental question has been remarkably difficult to answer over the last century. Why?

“Not all cells in the brain are created equal,” says Dr. B.N. Queenan, Executive Director of Research of Harvard’s Quantitative Biology Initiative. “Different neurons control different aspects of behavior.” Considering the large number of neurons in most organisms – from flies (250,000) to mice (71 million) to crows (2 billion) to humans (86 billion) – “finding the causal cell or cells is like looking for a needle in a haystack.”

“Even if you found the ‘right’ neuron,” continues Queenan, “it’s remarkably hard to pin Behavior A on Neuron A v. Neuron B because, generally speaking, a neuron does not control a complex behavior by itself. There isn’t a walking neuron, or a talking neuron, or an eating neuron. Instead, each neuron controls a smaller component – maybe a blink, a turn, a step, a squeeze. Collectively, the activity of many neurons can be sequenced into a coherent action. So you need to find not just one neuron in a haystack but the right cluster acting in the right sequence sequence.”

Recently, the Ramanathan group at Harvard University developed a new strategy to efficiently determine which cells cause which behavior. The team took advantage of a method called ‘compressed sensing,’ a strategy which may bring back memories of high school statistics class.

“Consider this puzzle,” says lead author Abdullah Yonar, a Ph.D. student in Applied Physics who led the study. “You have a pile of 64 coins. 63 coins are identical, 1 is different (in this case heavier). What is the quickest way to find the heavier coin? The most tedious way would be to measure all the coins one at a time. The quickest way would be to split the pile of coins in half, weigh both, then split the heavier pile. By repeating this, you can find the heaviest coin in 5 (or log (64)) measurements.”

“Now consider a much harder problem,” continued Yonar. “You have 64 coins again. Most of them have the same weight, but a few are different (some heavier, some lighter). What is the fastest way of identifying the different coins? With compressive sensing, you can use math to split and compare the piles, again finding the coins using remarkably few measurements.”

Following the principles of compressed sensing, Yonar and colleagues expressed a light-sensitive protein in arbitrary groups of neurons within the worm, C. elgans.  While the worm was slithering around freely, they turned on light which turned off the neuron groups. From these limited number of measurements (recording the behavioral effects of light-induced inhibition), the researchers efficiently identified three key neuronal types which control the worm’s speed.

Using higher magnification imaging and more sensitive neural recordings, the authors then discovered that the three neuron subtypes identified by compressive sensing control three different aspects of the worm’s movement. One type acts as a switch, determining whether the worm moves or not. Another acts as a rectifier, determining whether the animal moves forward or not. The third modulates the speed of locomotion continuously over slower time scales.

“Compressed sensing has been remarkably successful as a tool in signal processing, but has not previously been applied to neural systems,” says Dr. Sharad Ramanathan, Llura and Gordon Gund Professor of Neurosciences and of Molecular and Cellular Biology and Professor of Applied Physics and Stem Cell and Regenerative Biology at Harvard University. “This study shows the importance of bringing ideas from engineering and the quantitative sciences into biology. There are so many rich tools we can use to answer the most fundamental questions about living things. It’s a great time to be working at the interface of biology, engineering, and mathematics.”

Article: https://www.nature.com/articles/s41592-018-0233-6?WT.feed_name=subjects_physics

Lee et al 2019 Nature Methods A compressed sensing framework for efficient dissection of neural circuits