connectome_diffusion

Code to reproduce all analysis in Henderson et al. 2019 ("Spread of α-synuclein pathology through the brain connectome is modulated by selective vulnerability and predicted by network analysis", https://www.nature.com/articles/s41593-019-0457-5). In these analyses, we use a linear network diffusion model to explain the spread of α-synuclein pathology through the brain over time after injection of misfolded α-synuclein into the caudoputamen.

Requirements:

• MATLAB R2017a or later
• R 3.3.3 or later. Requisite packages are listed in code/packages.R

Directory structure

Master branch contains 2 major folders:

• code/ contains folders, each of which contain scripts specific to certain analyses, i.e. ‘code/diffmodel’ contains code that uses linear diffusion models to predict spread of protein through structural connectome, 'code/G20vsNTG' deals with comparisons between G20 mice and NTG mice.
• Data83018/ contains csv and xlsx files with 1) experimentally obtained pathology data and 2) parcellated Snca expression data and connectome data from Allen Brain Institute.

Using gene expression and linear dynamics of spread along the connectome, we attempt to predict the spatial distribution of experimentally observed pathology at 1, 3, and 6 months post injection.

Input specification

The file ‘pipeline.R’ is located in the main directory. This file will coordinate the sequential execution of all scripts within the code/ folder, generating all the figures in the paper and more. Custom specification of the following inputs at the top of ‘pipeline.R’ is required:

• basedir: path to the main directory containing the 'code' and 'Data83018' folders
• matlab.path: path to MATLAB binary
• opdir: name of output directory that contains all results, which will be housed in basedir
• grps: character vector containing the name of groups in data file to test. For our data set, these were 'NTG' and 'G20'.

Questions, suggestions, comments?

Please contact Eli Cornblath (Eli DOT Cornblath AT pennmedicine.upenn.edu) with any questions regarding network analysis and code, and contact Mike Henderson (hendm AT pennmedicine.upenn.edu) with any questions regarding experiments and data.