Atkinson-Lab/Tractor ? reverse-engineered prompt

Build me a simple workflow for running Tractor end to end on an admixed GWAS dataset.

I want to start with a phased VCF and a local ancestry file, then extract the ancestry specific dosage and hapcount files, and finally run the association test to get GWAS style summary results with beta, SE, and t value. It should also handle compressed input files if possible, and I’d like an option to use either a normal MSP file setup or FLARE local ancestry if that’s what the data has.

Please make it easy to run locally with a Conda environment, and include clear scripts or notebooks I can follow without having to piece everything together myself. If there’s a better way to organize the phenotype, sample ID, and covariates, build that in too. I’m okay with you looking up the current Tractor usage docs online if needed.