Reference

Installation

pip install git+https://github.com/shz9/penprs.git

Quickstart

importmagenpyasmgpfrompenprs.model.LassoimportLassofrompenprs.model.SSLimportSSLfrompenprs.model.SSLAlphaimportSSLAlpha# Load a genotype reference and GWAS summary statistics data (chromosome 22)gdl=mgp.GWADataLoader(mgp.tgp_eur_data_path(),sumstats_files=mgp.ukb_height_sumstats_path(),sumstats_format="fastgwa")# Compute block LD matrix:ld_block_url="https://bitbucket.org/nygcresearch/ldetect-data/raw/ac125e47bf7ff3e90be31f278a7b6a61daaba0dc/EUR/fourier_ls-all.bed"gdl.compute_ld('block',ld_blocks_file=ld_block_url,dtype='int16',compute_spectral_properties=True,output_dir='temp/block_ld/')# Initialize respective modelslasso_model=Lasso(gdl,lam=100)ssl_model=SSL(gdl,l0=750,l1=10)# Perform single fit using model initialized argumentslasso_model.fit()# fit using lam=100ssl_model.fit()# fit using l0=750, l1=10

Viewing model effect size estimates (BETA):

# Viewing the effect size estimates (BETA) for LASSOlasso_model.to_table().head()

Pathwise and Warm-Start fashioned fit

# To solve in a path-wise or warm-start fashion# save_intermediate=True to save effect size estimate per ladder steplasso_model_pw=Lasso(gdl)ssl_model_ws=SSL(gdl)# Performs path-wise fit across the default 20 step ladderlasso_model_pw.pathwise_fit(save_intermediate=True)# Performs warm-start fit across the default 20 step ladderssl_model_ws.warm_start_fit(save_intermediate=True)

Command Line Interface (CLI)

penprs_fit -l path_to_train_ld \-s path_to_train_sumstats \--output-dir path_to_output_dir \-m SSL\# model: options = LASSO, SSL, SSLAlpha--hyp-search GS\# hyperparameter search: options = GS (Grid Search), WS (Warm Start)--validation-ld-panel path_to_validation_ld \--validation-sumstats path_to_validation_sumstats \--use-symmetric-ld \--grid-metric pseudo_validation \--threads 4

Citation

@article{10.1093/bioinformatics/btaf578,author ={Song, Junyi and Zabad, Shadi and Yang, Archer and Gravel, Simon and Li, Yue},title ={Sparse Polygenic Risk Score Inference with the Spike-and-Slab LASSO},journal ={Bioinformatics},pages ={btaf578},year ={2025},month ={10},doi ={10.1093/bioinformatics/btaf578},url ={https://doi.org/10.1093/bioinformatics/btaf578},eprint ={https://academic.oup.com/bioinformatics/advance-article-pdf/doi/10.1093/bioinformatics/btaf578/64738987/btaf578.pdf},}