pokemon is a structure based variance component test for studying rare variants

Use through docker image
dockerhub repo: docker.io/bushlab/pokemon:latest

dir=/home/myhomespace
gene=ENSG00000183134
docker run -it --rm -v ${dir}:${dir} -w=/POKEMON docker.io/bushlab/pokemon \
python run_pokemon.py --gene_name ${gene} --genotype ${gene}.raw --phenotype test.pheno --annotation ${gene}.csq \
                     --cov_file test.cov --cov_list APOE4_dose,APOE2_dose,PC1,PC2 --alpha 0.0 --use_blosum  \
                     --out_file ${dir}/results --figures --out_fig_dir=${dir}  

Use through pip installation

pip install --user PokemonAnalysis  

gene=ENSG00000183134     
run_pokemon --gene_name ${gene} --genotype ${gene}.raw --phenotype test.pheno --annotation ${gene}.csq \
                     --cov_file test.cov --cov_list APOE4_dose,APOE2_dose,PC1,PC2 --alpha 0.0 --use_blosum  \
                     --out_file results 

python > 3.7

gcc > 6.3.0

Required package:(can be installed via pip)

• fastlmmclib see the instruction guide here: https://pypi.org/project/fastlmmclib/
• biopython see the instruction here: https://biopython.org/wiki/Download
• sklearn see the instruction here: https://scikit-learn.org/stable/install.html

Optional pacakges if visualization is on:

• pymol
  pymol must be installed to use the flag --figures, see instruction here: https://pymol.org/2/

--gene_name: required
A label used as gene name in the output result file

--genotype: required
allele count matrix note1: snp must be named as chr:pos:alt:ref (e.g., 6:41129275:G:C)
note2: snps must be unique

A example for genotype file is

A typical command to generate the genotype file:

plink2 --vcf <vcf file with genotypes> --set-all-var-ids @:#:\$r:\$a --snps-only --mac 1 --export A --out test_gene

--phenotype: required
phenotype file
note: phenotypes must be unique

format1 for testing with single phenotype: column1 for individuals, column2 for phenotype, seperated by space.

format2 for testing with multiple phenotypes: row for phenotypes, columns for individuals, seperated by space.

--annotation: required
Consequence annotations from Ensembl VEP with vcf format
INFO columns must contains CANONICAL|SWISSPROT|Amino_acids|Protein_position(can be easily achieved when run vep with outputing everything)
A typical script to generate the annotation file:

${dir_to_vep}/vep -i <vcf file with genotype> --cache --offline --dir <dir to cache> \
--check_ref --symbol --protein --uniprot --domains --canonical --biotype --coding_only --assembly GRCh37 \
--vcf -o test_gene.csq --no_stats

--alpha: required
alpha = 0: using structural kernel only
alpha = 0.5: using combined kernel of frequency and structure
alpha = 1: using frequency kernel only(equivalent to a standard SKAT test)

--out_file: required
output file where the POKEMON will write

--cov_file: optional covariate file.

--cov_list: optional, but required if --cov_file is flagged
covariates to be used, seperate by comma
covariate must be present in the columns for covariate file

--pdb: optional
e.g., --pdb 5eli, POKEMON will run on the specificed protein:5eli rather the optimal one

--maf: optional, default as 0.05
e.g., --maf 0.01, POKEMON will only run on variants with minor allele frequency < 0.01

--database: optional, default as pdb
can only be pdb or alphafold

--use_aa: optional
if explicitly flagged, the kernel will be further scaled by AA change weight from BLOSUM62 matrix

--figures: optional
if explicitly flagged, POKEMON will save pymol figures
pymol must be installed to use this flag

--out_fig_dir: optional but required if --figures is flagged
Directory where POKEMON will write figures to

Jin B, Capra JA, Benchek P, Wheeler N, Naj AC, Hamilton-Nelson KL, Farrell JJ, Leung YY, Kunkle B, Vadarajan B, et al. 2022. An association test of the spatial distribution of rare missense variants within protein structures identifies Alzheimer’s disease–related patterns. Genome Res 32: 778–790.