This repository contains the code used to process and analyse the data presented in the "Subtypes of functional connectivity associate robustly with ASD diagnosis" paper.

Requirements

• NIAK
• asdfc (included in this repository)

The analysis scripts included in this repository make use of a number of custom python functions included in the asdfc package. This package does not have to be installed but is locally referenced. The best way to obtain the a working python environment for this project is to pull the project's container image: docker pull surchs/abide_subtype

All data in this project were preprocessed using the NIAK toolkit. The easiest way to obtain and run NIAK is to download the most recent container image here: https://hub.docker.com/repository/docker/simexp/niak-cog

The easiest way to obtain the ABIDE 1 and 2 datasets is to download them directly from their AWS bucket at s3://fcp-indi/data/Projects. The AWS command line interface is an easy way to do so:

e.g. for ABIDE 1

aws s3 sync s3://fcp-indi/data/Projects/ABIDE/RawDataBIDS/ /path/to/ABIDE/RAW/LOCATION --no-sign-request

An alternative way is to directly download the data from the fcon_1000 repository, e.g. for ABIDE 2:

wget -r -np -nd -A tar.gz http://fcon_1000.projects.nitrc.org/indi/abide2/release/imaging_data

The phenotypic information for both datasets are also available at these locations.

Data for the HNU1 dataset can be similarly obtained from the fcon_1000 repository (http://dx.doi.org/10.15387/fcp_indi.corr.hnu1).

The appropriate NIAK preprocessing scripts can be generated with the python scripts under scripts/preprocessing and then be run directly on the raw data.

The raw phenotypic data has to undergo a number of preprocessing steps to harmonize differences in scan site naming and missing value declarations. This can be achieved by running the scripts in scripts/pheno in the following order (for ABIDE 1 and 2 respectively):

1. scripts/pheno/abide_1_prepare_pheno.py to fix issues with the raw phenotypic information.
2. scripts/pheno/abide_1_combine_pheno_and_motion.py to merge the phenotypic data and information on in-scanner head motion generated during preprocessing of the imaging data with NIAK.
3. scripts/pheno/abide_1_pheno_and_qc_rating.py to merge the phenotypic and quality control information (optional).
4. scripts/pheno/abide_1_psm_matching.R to compute the propensity matching scores that are needed in order to balance the clinical and control group across sites.

All subsequent processing steps can be run on the preprocessed data and phenotypic information using the python scripts in the scripts directory.

1. scripts/processing/abide_1_nuisance_regression.py to regress nuisance covariates estimated by NIAK during the preprocessing from the preprocessed fMRI time series data.
2. scripts/processing/abide_1_seed_maps.py to compute seed to voxel functional connectivity estimates based on the MIST_20 atlas.
3. scripts/stability/abide_1_subtype_map_stability.py and abide_1_subtype_stability_core.py to compute the stability of subtype maps and individual assignments respectively.
4. scripts/subtyping/abide_1_subtyping_core.py and abide_1_weights_core.py to compute subtypes and continuous assignments respectively.
5. scripts/association/abide_1_association_categorical.py to compute the association of the previously computed continuous assignment scores with categorical covariates from the phenotypic information.
6. scripts/association/abide_2_association_categorical.py to replicate the association findings on the independent validation sample (ABIDE 2).