prstools

prstools is software to create Polygenic Risk Scores (PRS) directly from the commandline
(and optionally from inside python).

It makes PRS generation easier, compared to previous tools, by:

• Super fast reading and matching of sumstats (handles odd formats).
• Rapid creation of the model.
• Automatic generation of the PRS Prediction for target plink files.

All the above, for a real GWAS sumstat, within 30 minutes using only 1 command. Installation and running the demo example should not take more than 10 minutes.

We are actively developing prstools and feedback by mail or our feedback form (with 🏆 lottery) is much appreciated!

Install

To install use the following command.

pip install -U --prefer-binary prstools

For it to work, you should have python3.8 or later installed (pip is included in python3.8+). If the command above does not work directly you can install using conda (or mamba if you have that), by running conda install "python>=3.9". For other install issues please check the install guide or send us a mail.

How to use

Immediately after installing prstools, it should be possible to download & run the demo example (~4mb), by pasting the following into the commandline (if not see install guide):

# Makes 'example' dir with data in current path:
prstools downloadutil --pattern example --destdir ./; cd example

# Run the model with example data:
prstools prscs2 --ref ldref_1kg_pop --target target \
                --sst sumstats.tsv --n_gwas 2565 --out ./result 

This will run PRS-CS2 on the example data, using the new implementation to demonstrate the capabilities of prstools and makes PRS predictions for the example dataset. The best and fastest way to get a PRS for your is to try the Tutorial below.
There is also the prstools documentation, which resides inside of the command-line interface, which you can see by typing prstools.

Usage:
 prstools <command> ...

Convenient and powerfull Polygenic Risk Score creation. 
'prst' is a commandline shorthand for 'prstools'

Models & Utility Commands:
 <command>
  downloadutil  Download and unpack LD reference panels and other data.
  combine       A tool to combine genetics-related text files.
  prscs2        PRS-CS v2: A polygenic prediction method that infers posterior SNP effect sizes
                under continuous shrinkage (CS) priors.

By combining prstools with another <command> a specific model or other functionality can be used.
Forinstance typing prstools prscs2 will output the documentation for the prscs2 subcommand:

Usage:
 prstools prscs2 [-h  --cpus <number-of-cpus>] --ref <dir/refcode> --target <bim-prefix>
                       --sst <file> --out <dir+prefix> [--n_gwas <num>  --chrom <chroms>]
                       [--colmap <alternative_colnames>  --pred  --n_iter <n_iter>]
                       [--n_burnin <n_burnin>  --n_slice <n_slice>  --seed <seed>  --a <a>]
                       [--b <b>  --phi <phi>  --clip <clip>  --sampler <sampler>]

PRS-CS v2: A polygenic prediction method that infers posterior SNP effect sizes under continuous shrinkage (CS) priors.

General Options:
 -h, --help                               Show this help message and exit.
 -c, --cpus <number-of-cpus>              The number of cpus to use. Generally most efficient if
                                          chosen to be between 1 and 5. Functionality can be

... [omitted for readability] ...


 --clip <clip>                            Clip parameter. The default works best in pretty much
                                          all cases. (default: 1.0)
 --sampler <sampler>                      Sampler algorithm. Rue sampling is the original sampler,
                                          which gives good results. (default: Rue)

Examples --> can be directly copy-pasted (:
 prst downloadutil --pattern example --destdir ./; cd example                                                # Makes 'example' dir in current path.
 prstools prscs2 --ref ldref_1kg_pop --target target --sst sumstats.tsv --n_gwas 2565 --out ./result-prscs2 # Run the model with example data.
 prst prscs2 -r ldref_1kg_pop -t target -s sumstats.tsv -n 2565 -o ./result-prscs2 --pred                  # A shorter version of previous that also does the predictions.

As can be seen, there are examples at the end of the help output to illustrate usage, which should work with a simple copy-paste. There is now also an online version of all this documentation (https://prstools.readthedocs.io/).

Tutorial + Video

For more information and a hands on demonstration of what prstools can do have a look at the Getting Started Tutorial. There is also a Video. The tutorial+video is a tiny bit older than the current prstools version, which has more functionality. You can load the tutorial in a free cloud instance by clicking here: .

Contact

For questions and support please send a mail (menno.j.witteveen@gmail.com).