Abstract
Understanding how cognitive functions emerge from the physical wiring of the brainis one of the main challenges in modern neuroscience. Recently, the full synaptic
resolution connectome of the adult Drosophila melanogaster (fruit fly) brain has
been made available for computational examination. However, the nature of the
raw connectome is a sparse, cyclic graph without any intrinsic coordinate system,
making systematic analysis at circuit-level spanning across the entire depth of such
brain structures a non-trivial algorithmic challenge.
This thesis provides a computational pipeline that introduces a global hierarchical
structure to the connectome via spectral graph embedding and data-driven level
assignment, grouping all neurons into five discrete functional levels, spanning from
sensory input to motor output. Extending this hierarchy, we propose a scalable
parametric framework for motif discovery that abstracts the search space from in
dividual neurons to functional states and explores circuits within a constrained
search algorithm with respect to three main parameters: (1) depth, (2) horizontal
complexity, and (3) skip connection distance.
Using this pipeline on the full connectome of Drosophila we find several general
architectural principles: the universal prevalence of feedforward-plus-local-feedback
relays at all information processing tiers; a shift from optic-dominated recurrence
in early stages to central-class integration at higher levels; the presence of large
directed cycles in intermediate stages; and the coexistence of qualitatively differ
ent fast-and-slow circuit topologies when allowing long-range skip connections. A
sliding window reconstruction approach illustrates how shorter local motifs build
into complete end-to-end signal transaction pathways that traverse the entire brain
hierarchy.
This framework is therefore general, parametric and directly applicable to other
fully mapped connectomes, laying a reusable foundation for computational connec
tomics that goes beyond the fruit fly.
| Datum der Bewilligung | 2026 |
|---|---|
| Originalsprache | Englisch (Amerika) |
| Gradverleihende Hochschule |
|
| Betreuer/-in | Herwig Mayr (Betreuer*in) & Giorgio Terracina (Betreuer*in) |
Studiengang
- Software Engineering
Zitieren
- Standard