In Silico Brain Sciences

Max Planck Research Group

We reconstruct neural networks and elucidate mechanistic principles of how the brain integrates sensory information.

Research Focus

We reconstruct neural networks and elucidate mechanistic principles of how the brain integrates sensory information.

One key challenge in neuroscience is to elucidate mechanistic principles of how the brain integrates sensory information from its environment to generate behavior. At present, experimental methods to directly monitor sensory-evoked streams of excitation throughout the brain, at cellular and millisecond resolution are lacking. To overcome these limitations, the ‘Max Planck Research Group: In Silico Brain Sciences’ seeks to develop an alternative reverse engineering approach. The novel approach comprises reconstructing the detailed 3D structure of neural circuits, quantifying local and long-range synaptic connectivity and simulating sensory-evoked signal flow within the resultant anatomically realistic network models.


Mechanistic Principles of Perception and Decision-Making

Long before the dawn of modern neuroscience, René Descartes pointed out that the problem of perception is at the core of understanding decision-making. In this respect, perception may be defined as understanding the environment by organizing and interpreting sensory information. These rather abstract definitions became much clearer in a beautiful behavioral experiment, performed by Celikel and Sakmann in 2007. In this study it was demonstrated that rodents decided reliably to cross a gap after they were able to reach and detect the other side of a platform using only a single facial whisker. This finding poses the following two questions, to which we devote our own research agenda:

1. What are the structural organization principles that underlie whisker-evoked signal flow?

2. What are the mechanistic principles that allow interpreting whisker-evoked stimuli?


Reverse Engineering the Vibrissal Sensory-Motor-System in Rodents


A cortical column is thought to represent the elementary functional unit of the mammalian cortex. In rodent vibrissal cortex, an anatomical equivalent, designated as a “barrel column”, has been described. Barrel columns are somatotopically arranged, resembling the organization of the facial whiskers on the animals’ snout. In addition to the fact that sensory input from a single facial whisker is sufficient to trigger decision-making, the well-defined anatomical layout and the one-to-one correspondence between a single facial whisker and a barrel column render the vibrissal system as the ideal starting point to reverse engineer the structure and function of neural microcircuits. To do so, our research focuses on (i) obtaining the 3D anatomical data that is necessary to reconstruct neural circuits, (ii) assembling average, anatomically realistic neural networks and (iii) simulating signal flow within the resulting large-scale, full-compartmental models of thousands of neurons and millions of synaptic contacts. First, we developed a semi-automated imaging and tracing pipeline, called NeuroMorph, which allows reconstructing the complete 3D dendrite and axon morphology of individual neurons, labeled in vivo.

Compared to state-of-the-art manual tracing tools it reduces the reconstruction time from several months to approximately three days. More importantly, the results are independent of the experience and performance of a human tracer. Using NeuroMorph, we will reconstruct morphologies from all cell types involved in whisker-evoked information processing in rat vibrissal sensory-motor cortex and thalamus. Second, we developed an image processing pipeline, called NeuroCount, which allows automated detecting of neuron somata within large brain volumes. Using NeuroCount, we will determined the number and 3D distribution of all neurons within rat vibrissal sensory-motor cortex and thalamus. Third, we developed an interactive, visual-computing tool, called NeuroNet, which allows assembling average 3D network models that are based on anatomical data obtained by NeuroMorph and NeuroCount. NeuroNet will allow reconstructing the cell type-specific excitatory network of rat vibrissal sensory-motor cortex and estimating thalamocortical and intracortical synaptic innervation of each neuron.

Finally, we developed a concept to investigate signal flow within average network models using Monte Carlo simulations. Using this framework, we will simulate the whisker-evoked activation of individual network-embedded neurons to reproduce their functional responses measured in vivo previously. In summary, the described concepts and tools open the possibility to (i) reconstruct the average 3D structure and synaptic wiring of the entire vibrissal sensory-motor-system, (ii) simulate whisker-evoked signal flow within the resultant anatomically realistic network models, (iii) compare simulations with functional data in vivo and finally (iv) elucidate mechanistic principles underlying perception and ultimately behavior.