Julijana Gjorgjieva
Department of Biology
Brandeis University
Theoretical Principles Underlying Sensory Pathways Diversification
The human brain consists of a dizzying number of different types of cells. Computers store and process information with simple switches — why do we need diversity in neuronal populations? Dr. Gjorgjieva’s research uses computational methods to assess how diversity in the types of signaling neurons could provide advantages for information transfer in the brain.
There are many distinct types of neurons in the brain, which are distributed in a highly organized fashion and interconnected with remarkable specificity. Such diversity of cell types is seen in different sensory modalities: vision, audition, smell, and also across different species from worms to flies to humans. This suggests an evolutionary fitness benefit of a very general nature.
What is the computational role of cell type diversity in a large population of neurons? One prominent hypothesis is termed efficient coding. It suggests that neurons in different sensory organs have evolved so that they transmit the maximal amount of information to downstream brain areas for further processing. I use this hypothesis to study the diversity of cell types in the retina of the eye. We have known for some time that there are about 20 different type of cells in the retina; some process light intensity, others process color or motion. I examine the benefits for the existence of ON and OFF cell types, where ON cells respond to increases in light intensity, while OFF cells respond to decreases in light intensity in the visual scene. Using a mathematical analysis, I show that more information about what we see is transmitted to our visual cortex if we have ON and OFF cells than if we have only ON, or only OFF, cells.
To do the analysis, I computed the statistics of bright and dark contrasts in a large bank of natural images collected by fellow visual neuroscientists. I also used empirical measurements from colleagues in the lab on the realistic constraints in the cells that we study, i.e. the maximal or mean amount of activity in the cells. In addition to showing that having both ON and OFF cells is better than just having cells of one kind, I also derived predictions for what the properties of these cells should be. For instance, given a number of neurons in a population that all code for the same visual stimuli, at what light intensity should each cell become active, and at what firing rate?
My collaborators are now testing these predictions in the real experimental system using retina recording in the mouse. Revealing principles for cell type diversification in the retina will aid in understanding the benefits of cell type diversity in subsequent stages of the visual system and in other sensory systems.