A Programmable Model for Exploring the Functional Logic of the Drosophila Antennal Lobe

Aurel A. Lazar, Mehmet Kerem Türkcan, Yiyin ZhoubioRxiv 2022 · DOI

TL;DR

This paper introduces a novel, programmable computational model of the Drosophila antennal lobe (AL), the insect's olfactory processing center. By systematically analyzing the impact of local feedback loops (LN motifs) within and between glomeruli, the authors provide a powerful methodology for understanding how complex olfactory computations emerge from simple circuit elements.

What Problem Does This Paper Address?

The advent of connectomics has provided incredibly detailed maps of neural circuits, down to the single synapse. However, understanding the functional logic – how these intricate wiring diagrams translate into computation – remains a significant challenge, especially for complex neuropils with many inputs and outputs. The Drosophila antennal lobe (AL) is a prime example: it's a crucial olfactory processing hub with a highly organized structure of glomeruli, each receiving input from olfactory sensory neurons (OSNs) and projecting to higher brain centers via projection neurons (PNs). The AL also features extensive local interneurons (LNs) that form feedback loops, significantly complicating its input-output (I/O) behavior. Existing models often struggle to capture this complexity and provide a standardized way to analyze the AL's computational capabilities. This paper aims to bridge this gap by developing a programmable model that allows for systematic exploration of how different feedback configurations influence olfactory processing.

Methods

The core of this work is a "programmable model" of the Drosophila antennal lobe. This model is built upon:

  • Glomerulus-Level Modeling: Each glomerulus, the fundamental processing unit of the AL, is modeled as a computational block.
  • Local Feedback Loops (LN Motifs): The model explicitly incorporates different configurations of local interneurons (LNs) that form feedback loops within and between glomeruli. These motifs are systematically categorized and explored (e.g., LN1, LN2, LN3, and their combinations).
  • Composability: A key methodological innovation is the concept of composability. The authors demonstrate how the I/O characteristics of individual glomeruli can be combined to predict the behavior of larger circuits, first by pairing glomeruli and then by considering the entire AL.
  • Parameterization: The model uses parameters derived from the Hemibrain dataset, a comprehensive connectomic reconstruction of the Drosophila brain, to assign synapse numbers and connectivity patterns. This grounds the model in biological reality.
  • Systematic I/O Characterization: The researchers systematically analyze the input-output (I/O) relationships of the AL under various conditions. This involves presenting simulated odorants at specific concentrations and observing the resulting activity patterns of projection neurons (PNs). They quantify these responses using peristimulus time histograms (PSTHs).

Datasets

The study utilizes the Hemibrain dataset, a detailed, single-synapse resolution connectomic reconstruction of the Drosophila adult central brain. This dataset provides the structural basis for parameterizing the model, specifically for determining the number of synapses for each port connectivity pattern within each glomerulus.

Key Insights and Ideas

  • Nonlinear I/O is Crucial: The paper highlights that even simple combinations of local feedback patterns within a single glomerulus can lead to drastically different I/O behaviors, underscoring the inherently nonlinear nature of olfactory processing in the AL.
  • Feedback Loops Shape Computation: Different LN motifs (connectivity patterns of local interneurons) have a profound and distinct impact on the AL's output. This suggests that the specific wiring of these feedback circuits is not arbitrary but actively shapes how olfactory information is processed.
  • Composability Enables Scalability: The ability to build up the AL's functional understanding from smaller units (glomeruli) to the whole circuit is a powerful demonstration of how complex systems can be understood through modular analysis. This approach allows for systematic exploration of the AL's computational repertoire.
  • Cell Type Specificity Matters: The study begins to explore how different types of LNs (e.g., Full, Patchy, Tortuous, Star) contribute to the AL's overall function, suggesting that the diversity of LN cell types is critical for fine-tuning olfactory computations.

Key Results

The paper presents a series of figures illustrating the I/O characteristics of the AL under various conditions:

  • Figure 2: Shows that different LN connectivity patterns within a single glomerulus lead to varied PN responses to a given OSN input.
  • Figure 3 & 4: Demonstrate how pairing glomeruli and applying different feedback motifs between them alters their individual PN responses, revealing how inter-glomerular interactions influence processing.
  • Figure 6 & 7: Illustrate the collective effect of various LN motifs and different LN cell types on the I/O of all glomeruli simultaneously, showing how global feedback configurations can modulate the entire AL's response to odorants like terpinolene and lactic acid.
  • Figure 8: Provides a more granular view by showing the impact of adding individual neurons of specific LN types one by one, highlighting the fine-tuning capabilities of these circuits.

While specific quantitative metrics like accuracy or speedup are not the primary focus, the results qualitatively demonstrate significant modulation of PN responses (ranging from near-zero to substantial changes) based on the implemented feedback motifs and LN types.

Limitations and Future Work

  • Simplifications: The model, while detailed, is still a simplification of the biological reality. The paper acknowledges that the existing local feedback within glomeruli is assumed to be "intact" and its specific configuration depends on the context, implying that further refinement of these internal glomerulus dynamics is possible.
  • Odorant Specificity: The current analysis focuses on a limited set of odorants (terpinolene and lactic acid). Future work could explore a wider range of odorants to understand the AL's response diversity.
  • Dynamic Behavior: While PSTHs are used, a deeper analysis of the temporal dynamics of neural responses could provide further insights.
  • Cell Type Ablation: The paper mentions exploring the effect of "ablating" LNs from the full AL circuit, suggesting future work could systematically remove specific LN types or individual neurons to pinpoint their precise roles.
  • Integration with Higher Centers: This work focuses on the AL. Future research could extend this model to investigate how AL outputs are further processed in downstream brain regions.

Summary updated

Abstract

Abstract Recent progress on connectomics has resulted in huge datasets of brain structures in single-synapse scale. This calls for the modeling of executable circuits to discover the functional logic of these neural circuits in different scales. To provide an approach to standardize analyses of neuropils with multiple-input-multi-output channels, we put forward a programmable model focusing on the antennal lobe (AL), a circuit at the periphery of the olfactory system with nonlinear computation. We present an approach for constructing and analyzing antennal lobe circuits using models of glomeruli and local feedback loops. We establish the composability of the connectivity of glomeruli with local neuron feedback loops by combining pairs of glomeruli, and then by combining all glomeruli together to characterize the I/O of the entire AL. We thus provide a methodology for the quantitative characterization of the I/O of the AL as a function of feedback loop motifs.
BibTeX
@misc{lazar2022programmable,
  author = {Aurel A. Lazar and Mehmet Kerem Türkcan and Yiyin Zhou},
  title = {{A Programmable Model for Exploring the Functional Logic of the <i>Drosophila</i> Antennal Lobe}},
  year = {2022},
  doi = {10.1101/2022.09.10.506216},
  howpublished = {bioRxiv},
}
Figure