Understanding how the brain is able to transform sensory input into behavior is one of the major challenges of systems neuroscience. While recording and imaging during sensory-motor tasks have identified neural substrates of sensation and action in various cortical areas, the crucial questions of 1) how these correlates are implemented within the underlying neural networks and 2) how their output triggers behavior, may only be answered when the individual functional measurements can be integrated into a coherent digital model of all task-related circuits. The goal of my research is to use the rodent whisker system for building such a model in the context of how a tactile-mediated percept is encoded by the interplay between cellular and circuit mechanisms. For this, my group has developed a ‘bottom-up’ approach that allows bridging across the different levels brain function analysis. Inspired by the Tri-Level Hypothesis of David Marr, we combine mathematical and computational approaches with experiments in the living animal. At the “implementational” level, we build realistic models of the rat whisker system, whose parameters are constrained by empirical anatomical and physiological data that my group collects systematically at synaptic, cellular and network scales. At the “algorithmic” level, we use these models to perform multi-scale simulations that mimic the specific experimental and behavioral conditions of our in vivo recording experiments. Based on these simulations, we explore which cellular and circuit mechanisms could in principle account for the observed activity patterns, and test these in silico predictions via pharmacological and optogenetic manipulations in vivo. At the “computational” level, we reduce the in silico predicted mechanisms to analytically tractable mathematical models. Ultimately, we seek to translate these mathematical models of sensation and action into design principles for artificial neural networks, which could reveal how higher brain functions – such as robust sensory perception – can emerge from their neurobiological implementations.
Discover our homepage here.
- In vivo electrophysiology and optogenetics
- Computational Modelling of neurons and neural networks
- Multi-scale simulations of cortical information processing
- Statistical inference on high-dimensional parameter spaces
- Development of machine-learning and artificial intelligence approaches
5 selected publications
- Guest JM, Bast A, Narayanan RT, Oberlaender M. (2021) Thalamus gates active dendritic computations in cortex during sensory processing. (in revision, bioRxiv doi: 10.1101/2021.10.21.465325)
- Egger R, Narayanan RT, Guest JM, Bast A, Udvary D, Messore LF, Das S, De Kock CPJ, Oberlaender M. (2020) Cortical Output Is Gated by Horizontally Projecting Neurons in the Deep Layers. Neuron 105 (1):122-137.
- Rojas-Piloni G, Guest JM, Egger R, Johnson AS, Sakmann B, Oberlaender M. (2017) Relationships between structure, in vivo function and long-range axonal target of cortical pyramidal tract neurons. Nat Commun. 8(1):870.
- Landau ID, Egger R, Dercksen VJ, Oberlaender M, Sompolinsky H. (2016) The Impact of Structural Heterogeneity on Excitation-Inhibition Balance in Cortical Networks. Neuron 92(5):1106-1121.
- Oberlaender M, Ramirez A, Bruno RM. (2012) Sensory experience restructures thalamocortical axons during adulthood. Neuron 74(4):648-55.