ContentContextNeurons

This repository contains code and data accompanying the following paper:

Distinct neuronal populations in the human brain combine content and context

Marcel Bausch¹, Johannes Niediek¹, Thomas P. Reber^1,2*, Sina Mackay¹, Jan Boström³, Christian E. Elger¹, Florian Mormann¹

¹ Department of Epileptology, University of Bonn Medical Centre, Bonn, Germany

² Faculty of Psychology, UniDistance Suisse, Brig, Switzerland

³ Department of Neurosurgery, University of Bonn Medical Centre, Bonn, Germany

If you have any questions, please contact marcel.bausch.ukb@gmail.com.

Requirements

• MATLAB R2021a or later
• MATLAB Statistics and Machine Learning Toolbox

Code Usage

To reproduce the figures in the manuscript, start MATLAB, navigate to the repository folder and add it to the MATLAB path:

cd('/Path/to/this/repo');
addpath(genpath(pwd));

% To generate Figure 2E
plot_figure_2E

% To derive neuron counts with specific significance threshold
alpha_level = 10^-3; 
load all_unitinfo
load rmANOVA
valid_neurons = find(get_unitinfo('sitenums', all_unitinfo) & get_unitinfo('session', all_unitinfo) ~= 26);
stimulus_neurons = sum(all_ps(valid_neurons, 1) < alpha_level);
context_neurons = sum(all_ps(valid_neurons, 2) < alpha_level);
stimulus_context_neurons = sum(all_ps(valid_neurons, 3) < alpha_level);

% Display results
fprintf('Neurons responding to stimulus: %d\n', stimulus_neurons);
fprintf('Neurons responding to context: %d\n', context_neurons);
fprintf('Neurons responding to stimulus-context interaction: %d\n', stimulus_context_neurons);

% To generate Figure 3
plot_fig3

% To generate Figure 4
plot_fig4

% To generate Figure 5
plot_fig5

Data Files

• all_unitinfo.mat: Metadata for each neuron (structure with fields 'data' and 'labels')

  • Contains information about recording sites, sessions, and patient IDs
  • Access metadata using get_unitinfo function, e.g., site_info = get_unitinfo('sitenums', all_unitinfo)

• rmANOVA.mat: Results from repeated-measures ANOVA for each neuron

  • all_etas (Nunits × 3): Partial eta squared values for stimulus, context, and stimulus-context effects
  • all_etas_imperm_boot (Nunits × 3): Eta values from permuting image labels PER question (control)
  • all_etas_qperm_boot (Nunits × 3): Eta values from permuting question labels PER image (control)
  • all_ps (Nunits × 3): P-values for stimulus, context, and stimulus-context effects

Main Functions

• plot_2E.m: Generates boxplots comparing partial eta squared values across effect types
• boxplot2.m: Enhanced boxplot with better formatting than MATLAB's built-in function
• get_seaborn.m: Creates color palettes similar to Python's Seaborn
• get_sitenames.m: Returns site name identifiers
• get_unitinfo.m: Retrieves metadata about specific units
• sigstar.m: Adds significance stars to plots

Analysis Details

Neuron Classification

Neurons are classified as responsive to different effects based on significance thresholds:

% Example classification with p < 0.001
alpha_level = 10^-3;
valid_neurons = find(get_unitinfo('sitenums', all_unitinfo) & get_unitinfo('session', all_unitinfo) ~= 26);

% Count neurons responding to each effect
stimulus_neurons = sum(all_ps(valid_neurons, 1) < alpha_level);
context_neurons = sum(all_ps(valid_neurons, 2) < alpha_level);
stimulus_context_neurons = sum(all_ps(valid_neurons, 3) < alpha_level);

% Count neurons responding to combinations of effects
stimulus_only = sum(all_ps(valid_neurons, 1) < alpha_level & all_ps(valid_neurons, 2) >= alpha_level & all_ps(valid_neurons, 3) >= alpha_level);
context_only = sum(all_ps(valid_neurons, 1) >= alpha_level & all_ps(valid_neurons, 2) < alpha_level & all_ps(valid_neurons, 3) >= alpha_level);
interaction_only = sum(all_ps(valid_neurons, 1) >= alpha_level & all_ps(valid_neurons, 2) >= alpha_level & all_ps(valid_neurons, 3) < alpha_level);

Statistical Analysis

The analysis uses repeated-measures ANOVA to calculate partial eta squared values and p-values for:

1. Main effect of stimulus (image identity)
2. Main effect of context (question type)
3. Interaction effect between stimulus and context

Statistical significance in Figure 2E is assessed using Wilcoxon signed-rank tests with Bonferroni correction, comparing actual data against permutation test controls.