k-pops (version 1.0.0)

K-PoPS is based on and follows the general framework of PoPS . It is designed for gene prioritization based on functional genomic data, and additionally allows for explanation of predictions.

K-PoPS has 3 steps,

In Step 0, use the src/munge_feature_files.py from PoPS to process the functional genomic data in a tabular format:

python munge_feature_directory.py \
--gene_annot_path data/gene_annot_jun10.txt \
--feature_dir test/features_raw/ \
--save_prefix test/pops_features \
--max_cols 500

K-PoPS doesn't directly use these genetic features, but instead construct a kernel matrix from these genetic features. We have prepared a script to create kernel. Users may try other kernels such as "linear", "rbf", and "polynomial".

python prepare_kernel.py \
--input_prefix test/pops_features \
--kernel_type linear \
--standardize \
--output_prefix test/kernel_linear

In Step 1, run MAGMA. You may use data/Ensembl.hg19.gene.loc as the --gene-annot. This annotation file uses Ensembl ID, which better aligns with other scripts.

./magma \
--bfile {PATH_TO_REFERENCE_PANEL_PLINK} \
--gene-annot {PATH_TO_MAGMA_ANNOT}.genes.annot \
--pval {PATH_TO_SUMSTATS}.sumstats ncol=N \
--gene-model snp-wise=mean \
--out {OUTPUT_PREFIX}

In Step 2, run K-PoPS:

K-PoPS produce two types of explanation evidence:

1. It reports top n contributor genes for each prediction. If top contributor genes are functionally related to the trait, then the prediction is more likely to be biologically plausible.
2. If users specify --anchor_genes, then it will calculate the anchor score. A biologically plausible prediction should have high anchor score (compare to nearby genes).

python k-pops.py \
--gene_annot_path data/gene_annot_jun10.txt \
--kernel_mat_prefix test/kernel_linear \
--magma_prefix test/ApoB \
--use_magma_covariates \
--training_chromosomes loco \
--device cuda \
--top_n_contributor_gene 5 \
--anchor_genes PCSK9,ANGPTL3,APOB,ABCG5,ALB,NPC1L1,GIGYF1,JAK2,A1CF,PDE3B,APOC3,CETP,ASGR1,LDLR,ZNF234,ZNF229,NECTIN2,RRBP1 \
--anchor_genes_type NAME \
--out_prefix test/ApoB \
--verbose