project_efc_afterglow

Code and resources supporting the manuscript:

Neurocomputational evidence of sustained Self–Other mergence after psychedelics
Mallaroni, P., Mason, N. L., Preller, K. H., Razi, A., Ereira, S., & Ramaekers, J. G.
medRxiv preprint (doi: 10.1101/2025.10.07.25337510).

Status

This repository accompanies a manuscript that is under peer review.

The analytical outputs are intended to reproduce the reported findings, but the codebase is being progressively updated for legibility and structure.

Practical implications:

• Script names, function boundaries, and directory layout may change as refactoring proceeds
• Hard-coded paths are being removed over time; for now, several scripts require editing a main_path or BASE variable
• Documentation is being expanded incrementally

What is in this repository

The project integrates:

• Behavioural modelling of a probabilistic false-belief task (pFBT)
• Effective connectivity (spectral DCM, PEB) on peak-effect 7T resting-state fMRI
• Permutation-based multivariate statistics linking computational and neuroimaging predictors to (sub)acute psychosocial outcomes
• NeuroSynth-derived region definitions and peak MNI coordinates for the Theory-of-Mind network used in the DCM analyses

Repository layout

project_efc_afterglow/
  neurosynth_maps/
    pub_getneuro.py
    neurosynth_coord.csv
    vmpfc/ dmpfc/ precuneus/ tpj/     # NeuroSynth association-test maps used to derive coordinates

  scripts/
    pFBT/
      model_fitting/                  # Probabilistic false-belief task model fitting utilities (MATLAB)
      synthetic_data/                 # Synthetic pFBT sessions used for demonstrations
      Parameter Estimates.mat         # Winning-model parameter estimates used by demo scripts
      TaskAccuracy.m                  # Demonstration of task-accuracy estimation (MATLAB)
      VisualiseParameters.m           # Reproduces key behavioural inference/plots (MATLAB)

    dcm/
      pub_dcmfirst.m                  # First-level spectral DCM specification/estimation (MATLAB/SPM)
      pub_dcmpeb.m                    # Within-subject PEB + third-level PEB-of-PEBs (MATLAB/SPM)
      plot_eFC_delta.m                # Plotting helper for effective connectivity contrasts (MATLAB)
      plot_eFC_mat_supplement.m       # Supplementary matrix plotting helper (MATLAB)

    wellbeing/
      pub_manova.py                   # λ predictor: perm MANOVA + canonical analysis + CV (Python)
      pub_manova_efc.py               # eFC predictor variant of the same pipeline (Python)

Getting started

1) Behavioural modelling (pFBT)

The scripts/pFBT/ folder contains self-contained demonstration scripts.

Full model fitting:

• The subfolder scripts/pFBT/model_fitting/ contains the functions used to fit the nested model family described in the manuscript.

2) Effective connectivity (spectral DCM)

The DCM scripts assume an SPM-based workflow and a local project directory with:

• Preprocessed resting-state NIfTI files available per subject/session
• Corresponding BIDS JSON sidecars for acquisition metadata
• ROI definitions derived from the NeuroSynth coordinate file

Key scripts:

• scripts/dcm/pub_dcmfirst.m runs first-level GLMs, extracts ROI time series, and estimates a spectral DCM
• scripts/dcm/pub_dcmpeb.m aggregates session-level DCMs into within-subject PEBs and runs a third-level PEB-of-PEBs for group effects and associations with behavioural predictors

Important: these scripts currently contain hard-coded paths (e.g., main_path, paths.spm, ROI directories). You will need to edit them to match your local environment.

3) Multivariate wellbeing association (Python)

The Python pipelines in scripts/wellbeing/ implement:

• Drug-aware within-subject residualisation
• Permutation MANOVA with within-subject shuffling
• Univariate follow-ups with Benjamini–Hochberg FDR
• Canonical analysis with structure coefficients
• Repeated-measures correlation
• K-fold cross-validation with permutation p-values

pub_manova.py uses lambda_val as the predictor.

pub_manova_efc.py uses an effective connectivity predictor specified by PREDICTOR_COL (default: rtpjdmpfc).

Important: these scripts currently assume local subject-level data excel inputs and paths:

• significant_behaviour.xlsx
• significant_outcomes_all.xlsx

You will need to update BASE, EXCEL_PATH, and OUTDIR to run them.

Dependencies

MATLAB

• MATLAB (tested in a modern MATLAB distribution)
• Statistics and Machine Learning Toolbox (for fitlme)

For DCM:

• SPM12 (or a compatible SPM build providing spectral DCM and PEB utilities)

Python

The wellbeing scripts use:

• numpy, pandas
• scipy
• statsmodels
• scikit-learn
• matplotlib
• pingouin

The NeuroSynth utility uses:

• nibabel
• nilearn
• scipy
• seaborn

Data availability

This repository includes:

• NeuroSynth maps and derived peak coordinates used for ROI definition
• Synthetic pFBT sessions used for demonstrations
• A .mat file containing winning-model parameter estimates used by the figure-generation demo scripts

Raw behavioural and neuroimaging data are not included here.

Contact

For questions, issues, or requests related to analysis details, please use contact the corresponding authors listed in the manuscript.