import trans
if args.ncores is not None:
import torch
torch.set_num_threads(args.ncores)
logging.info('Loading tables.')
df_pheno = load_phenotype(args.phenotype_table)
if args.covariate_table is not None:
df_covar = load_covariate(args.covariate_table, args.covariate_yaml)
indiv_lists = [df_covar.indiv.to_list(), df_pheno.indiv.to_list()]
else:
df_covar = None
indiv_lists = [df_pheno.indiv.to_list()]
if args.individual_list is not None:
indiv_lists.append(load_list(args.individual_list))
indiv_list = take_intersect(indiv_lists)
if args.individual_list_exclude is not None:
indiv_list = exclude_b_from_a(a=indiv_list,
b=load_list(
args.individual_list_exclude))
indiv_list = sorted(indiv_list)
df_pheno = rearrange_rows(df_pheno, indiv_list)
df_pheno = transpose_df(df_pheno, col='indiv')
if df_covar is not None:
df_covar = rearrange_rows(df_covar, indiv_list)
df_covar.set_index('indiv', inplace=True)
logging.info('There are {} individauls being included.'.format(
logging.basicConfig(level=logging.INFO,
stream=sys.stderr,
format='%(asctime)s %(message)s',
datefmt='%Y-%m-%d %I:%M:%S %p')
from tqdm import tqdm
from pyutil import load_list, intersection
if args.rename_yaml is not None:
from pyutil import read_yaml
rename_dict = read_yaml(args.rename_yaml)
else:
rename_dict = None
out = []
model_files = load_list(args.model_list)
logging.info('There are {} model files to load.'.format(len(model_files)))
for model_path in tqdm(model_files):
model = load_perf(model_path)
out.append(model)
out = pd.concat(out, axis=0)
# clean rename dict so that it only contains keys occurring in out
rename_dict_new = {}
for k in rename_dict.keys():
if k in list(out.phenotype):
rename_dict_new[k] = rename_dict[k]
rename_dict = rename_dict_new
new_pheno = []
for i in range(out.shape[0]):
input h5
''')
parser.add_argument('--gene_list', help='''
path to gene list
''')
parser.add_argument('--output', default=0, help='''
output csv
''')
args = parser.parse_args()
import h5py
import numpy as np
import pandas as pd
from pyutil import load_list
gene_list = load_list(args.gene_list)
with h5py.File(args.input, 'r') as f:
genes = f['genes'][:].astype(str)
gene_idxs = []
gene_list_new = []
for gene_id in gene_list:
gene_idx = np.where(np.isin(genes, gene_id))[0]
if len(gene_idx) == 0:
continue
else:
gene_idxs.append(gene_idx)
gene_list_new.append(gene_id)
gene_idxs = np.concatenate(gene_idxs, axis=0)
gene_list = gene_list_new
df_tmp = pd.DataFrame({'idx': gene_idxs, 'gene': gene_list})
A plink BIM file.
''')
parser.add_argument('--output', help='''
Output SNP list.
''')
args = parser.parse_args()
import logging, time, sys, os
# configing util
logging.basicConfig(level=logging.INFO,
stream=sys.stderr,
format='%(asctime)s %(message)s',
datefmt='%Y-%m-%d %I:%M:%S %p')
import pandas as pd
from pyutil import load_list, intersection, write_list
snp_list = load_list(args.input)
df_bim = pd.read_csv(args.input_bim, sep='\s+', header=None)
df_bim.columns = ['chr', 'rsid', 'placeholder', 'pos', 'a1', 'a2']
df_bim = filter_out_ambiguious_snps(df_bim)
snp_list = intersection(snp_list, list(df_bim.rsid))
# before output
# I found that there are some SNPs with ID like: AFFX-SNP_XXX__rsYYY or SNP_A-ZZZ
# I cleaned them up by removing all SNPs with SNP_A-ZZZ as ID and
# keep rsYYY from AFFX-SNP_XXX__rsYYY
snp_list = clean_up_rsid(snp_list)
write_list(snp_list, args.output)
import pyemma
if args.grm_cache is not None:
if args.reml is True:
grm_cache = args.grm_cache + '.reml.pkl.gz'
else:
grm_cache = args.grm_cache + '.mle.pkl.gz'
else:
grm_cache = None
logging.info('Loading phenotype table.')
df_y = read_table(args.y_table[0], indiv_col=args.y_table[1])
pheno_indiv = df_y.indiv.to_list()
if args.y_list is not None:
y_list = load_list(args.y_list)
df_y = df_y[['indiv'] + y_list]
pheno_list = df_y.columns[1:].to_list()
if grm_cache is None or not file_exists(grm_cache):
logging.info('Loading GRM from scratch.')
grm, grm_indiv = pyemma.load_grm(args.grm + '.grm.gz',
args.grm + '.grm.id')
common_indiv = intersection(grm_indiv, pheno_indiv)
grm, indiv = subset_grm(grm, grm_indiv, common_indiv)
ymat = subset_y(df_y, indiv)
logging.info('Starting GRM EVD.')
eig_val, eig_vec = pyemma.pyemma_mle_mat_fac(grm)
to_cache = {'vec': eig_vec, 'val': eig_val, 'indiv': indiv}