def generate_geno_batch(mTrait_qtl, mTrait, pTrait, geno, threads, bed_dir,
rs_dir):
if os.path.exists(bed_dir):
shutil.rmtree(bed_dir)
os.mkdir(bed_dir)
if os.path.exists(rs_dir):
shutil.rmtree(rs_dir)
os.mkdir(rs_dir)
plink_extract = 'plink -bfile {} -extract {} --make-bed -out {}'
geno_batch = list()
for mTrait_name in mTrait_qtl.phe_name.unique():
out_name = bed_dir.strip('/') + '/' + mTrait_name
rs = mTrait_qtl.loc[mTrait_qtl.phe_name == mTrait_name, 'SNP']
rs_name = rs_dir.strip('/') + '/' + '_'.join([mTrait_name, 'rs.txt'])
pd.Series(rs).to_frame().to_csv(rs_name, index=False, header=None)
geno_batch.append((plink_extract.format(geno, rs_name, out_name), ))
out_name = bed_dir.strip('/') + '/pTrait'
rs_name = rs_dir.strip('/') + '/pTrait_rs.txt'
mTrait_qtl['SNP'].to_frame().to_csv(rs_name, index=False, header=None)
geno_batch.append((plink_extract.format(geno, rs_name, out_name), ))
mp.parallel(mp.run, geno_batch, threads)
for fn in glob.glob(bed_dir.strip('/') + '/*fam'):
fam = pd.read_csv(fn, sep=' ', header=None)
mTrait_name = fn.split('/')[-1].replace('.fam', '')
if mTrait_name == 'pTrait':
pTrait = pTrait.reindex(fam[0])
fam.index = fam[0]
fam = pd.concat([fam, pTrait], axis=1)
else:
fam.loc[:, 5] = mTrait.loc[:, mTrait_name].reindex(fam[0]).values
fam.to_csv(fn, index=False, header=None, sep=' ', na_rep='NA')
def generate_qtl_batch(omics_phe, phe_sig_qtl, geno_name, threads, bed_dir,
rs_dir):
plink_extract = 'plink -bfile {} --extract {} --make-bed -out {}'
bim = pd.read_csv(geno_name + '.bim', sep='\t', header=None)
qtl_batch = list()
rs = dict()
for index, row in phe_sig_qtl.iterrows():
rs.setdefault(row['phe_name'], []).extend(
bim.loc[(bim[0] == row['chr']) & (bim[3] >= row['start']) &
(bim[3] <= row['end']), 1].values.tolist())
for phe_name in rs:
out_name = bed_dir.strip('/') + '/' + '_'.join(['tmp', phe_name])
rs_name = rs_dir.strip('/') + '/' + '_'.join(
['tmp', phe_name, 'rs.txt'])
pd.Series(rs[phe_name]).to_frame().to_csv(rs_name,
index=False,
header=False)
qtl_batch.append((plink_extract.format(geno_name, rs_name,
out_name), ))
mp.parallel(mp.run, qtl_batch, threads)
for fn in glob.glob(bed_dir.strip('/') + '/*fam'):
fam = pd.read_csv(fn, sep=' ', header=None)
phe_name = '_'.join(fn.split('/')[-1].split('_')[1:]).replace(
'm.z', 'm/z').replace('.fam', '')
fam.loc[:, 5] = omics_phe.loc[:, phe_name].reindex(fam.loc[:,
0]).values
fam.to_csv(fn, index=False, header=None, sep=' ', na_rep='NA')
def gwas(phe, geno, num_threads, phe_fn):
geno_prefix = geno.split('/')[-1]
related_matrix_cmd = 'gemma.linux -bfile {0}.link -gk 1 -o {1}'.format(
geno_prefix, geno_prefix)
gwas_cmd = 'gemma.linux -bfile {0}.link -k output/{0}.cXX.txt -lmm -n {1} -o {2}'
fam = pd.read_csv(geno + '.fam', sep=r'\s+', header=None)
fam[5] = 1
fam = pd.merge(fam, phe, left_on=0, right_index=True, how='left')
fam.to_csv(geno_prefix + '.link.fam',
sep='\t',
na_rep='NA',
header=None,
index=False)
if os.path.exists(geno_prefix + '.link.bed'):
os.remove(geno_prefix + '.link.bed')
if os.path.exists(geno_prefix + '.link.bim'):
os.remove(geno_prefix + '.link.bim')
os.symlink(geno + '.bed', geno_prefix + '.link.bed')
os.symlink(geno + '.bim', geno_prefix + '.link.bim')
values = list()
for _, p in enumerate(phe.columns):
p = p.replace('/', '.')
values.append((gwas_cmd.format(*[
geno_prefix, _ +
2, '.'.join(phe_fn.split('/')[-1].split('.')[:-1]) + '_' + str(p)
]), ))
s = mp.run(related_matrix_cmd)
if s != 0:
return None
else:
s = mp.parallel(mp.run, values, num_threads)
os.remove(geno_prefix + '.link.bed')
os.remove(geno_prefix + '.link.bim')
os.remove(geno_prefix + '.link.fam')
return s
def region_gwas_parallel(bed_dir, threads, geno):
local_gwas_args = list()
geno_prefix = geno.split('/')[-1]
fam = pd.read_csv(geno + '.fam', sep=r'\s+', header=None)
fam[5] = 1
fam.to_csv(geno_prefix + '.link.fam',
sep='\t',
na_rep='NA',
header=None,
index=False)
if os.path.exists(geno_prefix + '.link.bed'):
os.remove(geno_prefix + '.link.bed')
if os.path.exists(geno_prefix + '.link.bim'):
os.remove(geno_prefix + '.link.bim')
os.symlink(geno + '.bed', geno_prefix + '.link.bed')
os.symlink(geno + '.bim', geno_prefix + '.link.bim')
related_matrix_cmd = 'gemma.linux -bfile {0}.link -gk 1 -o {1}'.format(
geno_prefix, geno_prefix)
s = mp.run(related_matrix_cmd)
if s != 0:
return None
gemma_cmd = 'gemma.linux -bfile {0} -k ./output/{1}.cXX.txt -lmm -n 1 -o {2}'
for i in glob.glob(bed_dir + '/*.bed'):
i = i.replace('.bed', '')
i = i.replace('m/z', 'm.z')
prefix = i.split('/')[-1]
local_gwas_args.append((gemma_cmd.format(i, geno_prefix,
prefix + '_plink'), ))
s = mp.parallel(mp.run, local_gwas_args, threads)
os.remove(geno_prefix + '.link.bed')
os.remove(geno_prefix + '.link.bim')
os.remove(geno_prefix + '.link.fam')
return s
def gwas_plot_parallel(phe, p, threads, t, phe_fn):
values = list()
for i in phe.columns:
i = i.replace('/', '.')
values.append(
('output/' + '.'.join(phe_fn.split('/')[-1].split('.')[:-1]) +
'_' + str(i) + '.assoc.txt', p,
'.'.join(phe_fn.split('/')[-1].split('.')[:-1]) + '_' + str(i),
t))
s = mp.parallel(gwas_plot, values, threads)
return s
def MR_parallel(mTrait_qtl, mTrait, pTrait, geno, threads, pvalue_cutoff):
args = list()
for index, row in mTrait_qtl.iterrows():
rs = row['SNP']
mTrait_name = row['phe_name']
args.append(
(mTrait.loc[:, mTrait_name], pTrait, geno.loc[:,
rs], pvalue_cutoff))
res = mp.parallel(MR, args, threads)
res = pd.concat([i for i in res])
return res
def MR_MLM_parallel(mTrait_qtl, mTrait_effect, pTrait_effect, pTrait_se,
threads, pvalue_cutoff):
args = []
for index, row in mTrait_qtl.iterrows():
mTrait_name = row['phe_name']
rs = row['SNP']
args.append(
(mTrait_effect.loc[';'.join([mTrait_name, rs]), :],
pTrait_effect.loc[rs, :], pTrait_se.loc[rs, :], pvalue_cutoff))
res = mp.parallel(MR_MLM, args, threads)
res = pd.concat([i for i in res])
return res
def genome_cluster(G, window, step, threads):
paras = list()
if threads > np.unique(G.chrom).shape[0]:
threads = np.unique(G.chrom).shape[0]
for chrom in np.unique(G.chrom):
G_chr = G.where(G.chrom == chrom, drop=True)
paras.append((G_chr, chrom, window, step))
res = mp.parallel(chr_cluster_pca, paras, threads)
#res_pc = pd.concat([i[0] for i in res], axis=1)
res_pc = pd.concat(res, axis=1)
res_pc.loc[:, :] = np.around(
MinMaxScaler(feature_range=(0, 2)).fit_transform(res_pc.values),
decimals=3)
#res_variant = pd.concat([i[1] for i in res])
#return res_pc,res_variant
return res_pc
def qtl_pc_lmm_gwas_parallel(omics_phe, bimbam_dir, threads, geno, sample_id):
qtl_pc_lmm_args = list()
#g = read_plink1_bin(geno+'.bed', geno+'.bim', geno+'.fam', verbose=False)
#g = g.sel(sample=sample_id)
geno_prefix = geno.split('/')[-1]
#if os.path.exists(geno_prefix+'.link.bed'):
# os.remove(geno_prefix+'.link.bed')
#if os.path.exists(geno_prefix+'.link.bim'):
# os.remove(geno_prefix+'.link.bim')
#write_plink1_bin(g,geno_prefix+'.link.bed', geno_prefix+'.link.bim,', geno_prefix+'.link.fam',verbose=False)
fam = pd.read_csv(geno + '.fam', sep=r'\s+', header=None)
fam[5] = 1
fam.to_csv(geno_prefix + '.link.fam',
sep='\t',
na_rep='NA',
header=None,
index=False)
omics_phe = omics_phe.reindex(fam[0].values)
omics_phe.to_csv('bimbam_phe.txt',
sep='\t',
index=False,
header=None,
na_rep='NA')
if os.path.exists(geno_prefix + '.link.bed'):
os.remove(geno_prefix + '.link.bed')
if os.path.exists(geno_prefix + '.link.bim'):
os.remove(geno_prefix + '.link.bim')
os.symlink(geno + '.bed', geno_prefix + '.link.bed')
os.symlink(geno + '.bim', geno_prefix + '.link.bim')
related_matrix_cmd = 'gemma.linux -bfile {0}.link -gk 1 -o {1}'.format(
geno_prefix, geno_prefix)
s = mp.run(related_matrix_cmd)
if s != 0:
return None
gemma_cmd = 'gemma.linux -g {0} -a {1} -p bimbam_phe.txt -k ./output/{2}.cXX.txt -lmm -n {3} -o {4}'
for _, m in enumerate(omics_phe.columns):
m = m.replace('m/z', 'm.z')
qtl_pc_lmm_args.append((gemma_cmd.format(
bimbam_dir.strip('/') + '/tmp_' + m + '.geno.txt',
bimbam_dir.strip('/') + '/tmp_' + m + '.anno.txt', geno_prefix,
_ + 1, m + '_bimbam'), ))
s = mp.parallel(mp.run, qtl_pc_lmm_args, threads)
os.remove(geno_prefix + '.link.bed')
os.remove(geno_prefix + '.link.bim')
os.remove(geno_prefix + '.link.fam')
return s
def qtl_pc_lm_gwas_parallel(omics_phe, bimbam_dir, threads, geno):
qtl_pc_lm_args = list()
geno_prefix = geno.split('/')[-1]
gemma_cmd = 'gemma.linux -g {0} -a {1} -p {2} -lm -o {3}'
for m in omics_phe.columns:
phe = omics_phe[m].to_frame()
m = m.replace('m/z', 'm.z')
phe.to_csv(bimbam_dir.strip('/') + '/' + m + '_phe.txt',
index=False,
header=None,
na_rep='NA')
qtl_pc_lm_args.append((gemma_cmd.format(
bimbam_dir.strip('/') + '/' + geno_prefix + '_qtl_pc.geno.txt',
bimbam_dir.strip('/') + '/' + geno_prefix + '_qtl_pc.anno.txt',
bimbam_dir.strip('/') + '/' + m + '_phe.txt', m + '_bimbam_lm'), ))
s = mp.parallel(mp.run, qtl_pc_lm_args, threads)
return s
def plink_clump(geno_path, p1, p2, num_threads):
if os.path.exists('./clump_result'):
shutil.rmtree('./clump_result')
os.mkdir('./clump_result')
cmd = 'plink --bfile {0} --clump {1} --clump-p1 {2} --clump-p2 {3} --clump-kb {4} --clump-r2 0.2 --out {5}'
cmds = list()
ms = list()
for fn in glob.glob('./clump_input/*'):
phe_name = fn.split('/')[-1].replace('.assoc', '')
cmds.append(
(cmd.format(geno_path + '/' + phe_name, fn, p1, p2, str(500),
'./clump_result/' + phe_name + '_' + str(500)), ))
ms.append(phe_name)
s = mp.parallel(mp.run, cmds, num_threads)
if sum(s) != 0:
print(','.join(list(np.array(ms)[s])) +
' do not successfully generated clumped file.')
return s
def get_MLM_effect_parallell(assoc_dir, threads):
mTrait_effect = pd.DataFrame()
args = []
pTrait_name = []
for fn in glob.glob(assoc_dir.strip('/') + '/mTrait*.assoc.txt'):
mTrait_name = fn.split('/')[-1].split('_')[-1].replace(
'.assoc.txt', '')
assoc = pd.read_csv(fn, sep='\t')
assoc.index = mTrait_name + ';' + assoc['rs']
mTrait_effect = pd.concat([mTrait_effect, assoc[['beta', 'se']]])
for fn in glob.glob(assoc_dir.strip('/') + '/pTrait*assoc.txt'):
pTrait_name.append(
fn.split('/')[-1].split('_')[-1].replace('.assoc.txt', ''))
args.append((fn, ))
pTrait_res = mp.parallel(get_MLM_effect, args, threads)
pTrait_effect = pd.concat([i['beta'] for i in pTrait_res], axis=1)
pTrait_effect.columns = pTrait_name
pTrait_se = pd.concat([i['se'] for i in pTrait_res], axis=1)
pTrait_se.columns = pTrait_name
return mTrait_effect, pTrait_effect, pTrait_se
def calc_MLM_effect(bed_dir, pTrait, threads, geno):
args = list()
geno_prefix = geno.split('/')[-1]
fam = pd.read_csv(geno + '.fam', sep=r'\s+', header=None)
fam[5] = 1
fam.to_csv(geno_prefix + '.link.fam',
sep='\t',
na_rep='NA',
header=None,
index=False)
if os.path.exists(geno_prefix + '.link.bed'):
os.remove(geno_prefix + '.link.bed')
if os.path.exists(geno_prefix + '.link.bim'):
os.remove(geno_prefix + '.link.bim')
os.symlink(geno + '.bed', geno_prefix + '.link.bed')
os.symlink(geno + '.bim', geno_prefix + '.link.bim')
related_matrix_cmd = 'gemma.linux -bfile {0}.link -gk 1 -o {1}'.format(
geno_prefix, geno_prefix)
s = mp.run(related_matrix_cmd)
if s != 0:
return None
gemma_cmd_mTrait = 'gemma.linux -bfile {0} -k ./output/{1}.cXX.txt -lmm -n 1 -o {2}'
gemma_cmd_pTrait = 'gemma.linux -bfile {0} -k ./output/{1}.cXX.txt -lmm -n {2} -o {3}'
for i in glob.glob(bed_dir + '/*.bed'):
i = i.replace('.bed', '')
if i.split('/')[-1] != 'pTrait':
prefix = i.split('/')[-1]
args.append((gemma_cmd_mTrait.format(i, geno_prefix,
'mTrait_' + prefix), ))
else:
for _, pTrait_name in enumerate(pTrait.columns):
args.append(
(gemma_cmd_pTrait.format(i, geno_prefix, _ + 2,
'pTrait_' + pTrait_name), ))
s = mp.parallel(mp.run, args, threads)
os.remove(geno_prefix + '.link.bed')
os.remove(geno_prefix + '.link.bim')
os.remove(geno_prefix + '.link.fam')
return s
