sgVAMP-py

Python implementation of gVAMP for summary statistics.

Example

module load python3

sloc={Path to sgVAMP/src folder}

python3 ${sloc}/main.py [Input options]

Input options

Option	Description
--ld-files	Path to LD matrices files, separated by comma
--r-files	Path to XTy files separated by comma
--true-signal-file	Path to true signal file
--out-dir	Output directory
--out-name	Output file name
--N	Number of samples
--M	Number of markers
--K	Number of cohorts
--L	Number of prior mixture components (Including spike)
--iterations	Number of iterations
--prior-vars	Prior mixture variances of different cohorts, separated by comma (e.g. 0,1). First must be 0.
--prior-probs	Prior mixture probabilites of different cohorts, separated by comma (e.g. 0.99,0.01). Must sum up to 1.
--gamw	Initial noise precision
--gam1	Initial signal precision
--lmmse-damp	Use LMMSE damping
--learn-gamw	Learn or fix gamw
--rho	Damping factor rho
--cg-maxit	CG max iterations
--s	Rused = (1-s) * R + s * Id
--mle-prior-update	Updating prior probabilities using MLE ('mle') or EM ('em')
--em-prior-maxit	Maximal number of iterations that prior-learning EM is allowed to perform

Output files

Signal estimates over iterations are stored in binary files: {out_dir}/{out_name}__xhat_it_{it}.bin

Separate CSV output files are created for each cohort: {out_dir}/{out_name}_cohort_{id}.csv

CSV file structure:

Iteration	gamw	gam1	gam2	alpha1	alpha2
...