pavanchaggar / local-fkpp

Code for "Personalised Regional Modelling Predicts Tau Progression in the Human Brain", Chaggar et al 2024

This repository contains the code used in the paper Chaggar et al., 2024 for the analysis of ADNI data.

Code installation

We use the Julia Progamming Language for this project. The full package environment is uploaded in the Project.toml and Manifest.toml and can be used to instantiate the dependencies used by this project.

Data analysis

ADNI data were downloaded from the LONI portal https://adni.loni.usc.edu/. Amyloid and tau PET data were last accessed 18/12/23. We classify subjects as into three cohorts depending on amyloid and tau status. Amyloid status is determined provided by ADNI and the script for determining individual amyloid status is adni/data/data-processing.jl. The script for calculating the demographics of each cohort used in the paper is in adni/data/demographics.jl.

Inference

We perform inference over four models, the local FKPP, global FKPP, logistic and diffusion models.

The code used for loading the connectome and ADNI data is in adni/inference/inference-preamble.jl and is imported into the separate analysis scripts for each model.

For the local FKPP model, we run inference for three cohorts, A+T+, A+T- and A-T-, for the remaining models, we only run inference on the A+T+ groups. For each of these models, inference scripts can be found in adni/inference directory. For example, inference for the A+T+ group using the local FKPP model is found in adni/inference/local-fkpp/taupos.jl.

Additionally, for all models, we run inference using a test/train data split, using only the first three scans for training. The scripts for these are found in the model inference directories and suffixed with -three.

Visualisation

The visualisation the posterior analysis can be found in visualisation.

The panels in Figure 1 of the manuscript are created using the scripts in visualisation/models.

The panels in Figure 2 of the manuscript are created using visualisation/inference/model-selection/model-selection-pos.jl.

The panels in Figure 3 of the manuscript for the ADNI posterior distributions are created using `visualisation/posteriors/posteriors.jl.

The panels in Figure 4 of the manscripts are generated using 'visualisation/inference/model-selection/out-sample-three-pos.jl.

Revisions

Code used for analysis stemming from the peer-review process is currently in the revisions branch and will be merged upon acceptance.

Citation

Chaggar, Pavanjit, Jacob W. Vogel, Alexa Pichet-Binette, Travis B. Thompson, Olof B. Strandberg, Erik Stomrud, Niklas Mattsson-Carlgren et al. "Personalised Regional Modelling Predicts Tau Progression in the Human Brain." bioRxiv (2023): 2023-09.

@article{chaggar2023personalised,
  title={Personalised Regional Modelling Predicts Tau Progression in the Human Brain},
  author={Chaggar, Pavanjit and Vogel, Jacob W and Pichet-Binette, Alexa and Thompson, Travis B and Strandberg, Olof B and Stomrud, Erik and Mattsson-Carlgren, Niklas and Karlsson, Linda and Jbabdi, Saad and Magon, Stefano and others},
  journal={bioRxiv},
  pages={2023--09},
  year={2023},
  publisher={Cold Spring Harbor Laboratory}
}