Shaping cortical computations via higher-order feedback

Introduction

Our ability to flexibly adapt to changing environments depends on how we perceive, prioritize, and act on stimuli. This involves actively integrating our current sensory experiences with our prior knowledge of the world and the surrounding context.

Stimulus Perception

Stimulus perception is influenced by contextual top-down signals from areas higher up in the processing hierarchy that carry information about internal state, attention, and future actions to early processing stages where they are combined with bottom-up inputs. Commonly referred to as “feedback,” these top-down signals are multi-faceted; they come from diverse brain areas and are integrated at different loci in neural circuits.

Research Gap

What type of information they carry and where is still unclear despite their fundamental role in shaping even the most mundane tasks.

Project Overview

FeedbackCircuits will investigate the mechanistic circuit basis of feedback-driven cortical computations, including:

1. Contextual modulation
2. Amplification of unexpected stimuli
3. Synaptic plasticity mechanisms governing the wiring logic of feedback projections

Methodology

Constrained by experimental data from the mouse visual cortex, I will build a multi-scale theoretical framework that unifies diverse experimental findings and links cellular to circuit-level processing.

Data Strategy

Our strategy leverages new datasets that encompass multiple modalities, including:

• Neural responses in various brain regions
• Detailed synaptic-level wiring diagrams

Expected Outcomes

The proposed mechanistic models will enable the exploration of distinct feedback sources and sites of plasticity. Together with the data, these models will define plausible parameter spaces underlying feedback-driven computations.

Comparison to Other Efforts

In contrast to other efforts training hard-to-interpret artificial neural networks, our models promise to elucidate the mechanistic underpinnings of circuit structure-function dynamics involving feedback. This will help to distinguish between competing mechanistic hypotheses and make numerous experimental predictions.