Modeling the formation of prediction-error neurons in complex neural networks

• Applicant: Loreen Hertäg, Ph.D.
• Subject Area: Experimental and Theoretical Network Neuroscience
• Term: since 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 460088091

There is accumulating evidence that neural circuits constantly compare the input perceived through our senses with predictions about what we (will) see, hear, smell or feel. The core elements of this comparison are prediction-error neurons that signal mismatches between actual and predicted inputs by changing their activity. It is believed that the activity of prediction-error neurons is the basis for adjusting and refining our inner model of the world (that is, our beliefs and expectations) by governing plasticity in hierarchical neural networks. While neural hallmarks of prediction-errors have been found throughout the brain, only little is known about the underlying circuit-level mechanisms that shape the computation of prediction-errors in excitatory neurons. The central hypothesis of the research proposal is that prediction-error neurons are formed by an intricate interplay of inhibitory interneurons that aims at balancing excitation and inhibition in different compartments of excitatory cells for predicted sensory stimuli. The goal of the present project is, therefore, to decipher the neuronal, synaptic and connectivity conditions under which complex networks with excitatory cells and multiple types of inhibitory interneurons can compute mismatches between actual and predicted sensory inputs. To this end, we will use mathematical analysis to derive interneuron motifs that allow different types of prediction-error neurons to emerge simultaneously. In the next step, we will infer the synaptic plasticity rules that enable networks with excitatory and inhibitory neurons to self-organize into those heterogeneous comparator circuits. Finally, we will study how well-known neural and synaptic properties of different cell classes influence the formation, robustness and specificity of prediction-error neurons. This here proposed work will therefore lead to a better understanding of the network properties and principles underlying the tremendous power of neural circuits to predict their inputs and detect mismatches.

• DFG Programme: Research Grants