Predicting Anatomically Realistic Cortical Connectomes using Statistical Inference

The overall goal of the project is to develop methods to reveal the principles of cortical organization from the wealth of data generated by modern connectomics approaches, and then to use these principles to build models of brain function. This shall be achieved by developing and utilizing a statistical framework for formulating, discovering and testing laws of synaptic organization on both sparse and dense connectomics data. It will make possible to implement hypotheses of synaptic organization in terms of mathematically formulated rules. Given distributions of pre- and postsynaptic structures from all neurons located within a cortical volume of interest, each rule predicts dense synaptic wiring diagrams, whose properties can be tested directly against results from state-of-the-art connectivity measurement techniques. Bayesian inference techniques will make it possible to identify which local and global structural features are predictive of synaptic connections, to quantify how well a connectivity rule is constrained by data, and to determine which set of rules is most consistent with empirical data.

The framework’s usability will be tested by comparing its predictions of connectivity in layer 5 of rat primary somatosensory cortex against novel in vivo-based connectivity measurements. Preliminary results let us expect that the envisioned approach has potential to reveal local rules that underlie global synaptic organization in neocortical circuits. If this applies, the project will provide a foundation for studying the relationships between structural properties of neuronal networks, their underlying principles of synaptic organization and cortical functions. In addition, the framework will serve to investigate developmental mechanisms leading to these relationships and to analyze the structural origin or correlates of cortical malfunctions during pathological conditions.
We will combine complementary expertise in data science, Bayesian statistics and in vivo-based neuroanatomy to build, test and utilize the framework.

Visit at GEPRIS

Principal Investigators

Professor Hans-Christian Hege
Konrad-Zuse-Zentrum für Informationstechnik Berlin (ZIB)
Department Visual Data Analysis

Dr. Jakob Macke
Forschungszentrum caesar

Dr. Marcel Oberlaender
Forschungszentrum caesar