Scalable pipeline for computing LD matrix in big sample phenotype
4 modules
- Genodata
- Sumstats
- Liftover
- LDmatrix
Install
pip install cugg
How to use
lf = Liftover('hg38','hg19')
vcf ='/home/yh3455/Github/SEQLinkage/MWE/small_sample_ii_coding.vcf.gz'
lf.vcf_liftover(vcf)
!which python
/home/yh3455/miniconda3/bin/python
region = [5,272741,1213528-900000]
geno_path = 'MWE_region_extraction/ukb23156_c5.merged.filtered.5_272741_1213528.bed'
sumstats_path = 'MWE_region_extraction/090321_UKBB_Hearing_aid_f3393_expandedwhite_6436cases_96601ctrl_PC1_2_f3393.regenie.snp_stats'
pheno_path = None
unr_path = 'MWE_region_extraction/UKB_genotypedatadownloaded083019.090221_sample_variant_qc_final_callrate90.filtered.extracted.white_europeans.filtered.092821_ldprun_unrelated.filtered.prune.txt'
imp_geno_path = 'MWE_region_extraction/ukb_imp_chr5_v3_05_272856_1213643.bgen'
imp_sumstats_path = 'MWE_region_extraction/100521_UKBB_Hearing_aid_f3393_expandedwhite_15601cases_237318ctrl_500k_PC1_PC2_f3393.regenie.snp_stats.gz'
imp_ref = 'hg19'
bgen_sample_path = 'MWE_region_extraction/ukb_imp_chr5_v3_05_272856_1213643.sample'
output_sumstats = 'test.snp_stats.gz'
output_LD = 'test_corr.csv.gz'
def main(region,geno_path,sumstats_path,pheno_path,unr_path,imp_geno_path,imp_sumstats_path,imp_ref,output_sumstats,output_LD,bgen_sample_path):
print('1. Preprocess sumstats (regenie format) and extract it from a region')
if pheno_path is not None:
pheno = pd.read_csv(pheno_path, header=0, delim_whitespace=True, quotechar='"')
if unr_path is not None:
unr = pd.read_csv(unr_path, header=0, delim_whitespace=True, quotechar='"')
exome_sumstats = Sumstat(sumstats_path)
exome_geno = Genodata(geno_path,bgen_sample_path)
print('1.1. Region extraction')
exome_sumstats.extractbyregion(region)
exome_geno.extractbyregion(region)
exome_sumstats.match_ss(exome_geno.bim)
exome_geno.geno_in_stat(exome_sumstats.ss)
if imp_geno_path is not None:
imput_sumstats = Sumstat(imp_sumstats_path)
imput_geno = Genodata(imp_geno_path,bgen_sample_path)
if imp_ref is None:
imput_sumstats.extractbyregion(region)
imput_geno.extractbyregion(region)
imput_sumstats.match_ss(imput_geno.bim)
imput_geno.geno_in_stat(imput_sumstats.ss)
else:
print('1.2. LiftOver the region')
hg38toimpref = Liftover('hg38',imp_ref)
imp_region = hg38toimpref.region_liftover(region)
imput_sumstats.extractbyregion(imp_region)
imput_geno.extractbyregion(imp_region)
imput_sumstats.match_ss(imput_geno.bim)
imput_geno.geno_in_stat(imput_sumstats.ss)
print('1.3. Regional SNPs Liftover')
impreftohg38 = Liftover(imp_ref,'hg38')
imput_geno.bim = impreftohg38.bim_liftover(imput_geno.bim)
imput_sumstats.ss.POS = list(imput_geno.bim.pos)
imput_sumstats.ss.SNP = list(imput_geno.bim.snp)
print('1.1.1 Get exome unique sumstats and geno and Combine sumstats')
snp_match = compare_snps(exome_sumstats.ss,imput_sumstats.ss)
exome_sumstats.ss = exome_sumstats.ss.loc[snp_match.qidx[snp_match.exact==False].drop_duplicates()]
exome_sumstats.extractbyvariants(imput_sumstats.ss.SNP,notin=True)
exome_geno.geno_in_stat(exome_sumstats.ss)
sumstats = pd.concat([exome_sumstats.ss,imput_sumstats.ss])
else:
sumstats = exome_sumstats
print('2. Remove relative samples')
if unr_path is not None:
exome_geno.geno_in_unr(unr)
if imp_geno_path is not None:
imput_geno.geno_in_unr(unr)
else:
print('Warning:There is no file of relative sample. All sample are included in computing LD matrix')
if pheno_path is not None:
print('Warning: This function has been implementd yet.')
pass
print('3. Calculate LD matrix')
if imp_geno_path is None:
cor_da = geno_corr(exome_geno.bed.T)
else:
xx = geno_corr(exome_geno.bed.T)
yy = geno_corr(imput_geno.bed.T,step=500)
imput_fam = imput_geno.fam.copy()
imput_fam.index = list(imput_fam.iid.astype(str))
imput_fam['i'] = list(range(imput_fam.shape[0]))
imput_fam_comm = imput_fam.loc[list(exome_geno.fam.iid.astype(str))]
imput_geno.extractbyidx(list(imput_fam_comm.i),row=False)
xy = geno_corr(exome_geno.bed.T,imput_geno.bed.T,step=500)
cor_da = da.concatenate([da.concatenate([xx,xy],axis=1),da.concatenate([xy.T,yy],axis=1)],axis=0)
print('4. Output sumstats and LD matrix')
index = list(sumstats.SNP.apply(shorten_id))
sumstats.SNP = index
sumstats.index = list(range(sumstats.shape[0]))
sumstats.to_csv(output_sumstats, sep = "\t", header = True, index = False,compression='gzip')
corr = cor_da.compute()
np.fill_diagonal(corr, 1)
corr = pd.DataFrame(corr, columns=index)
corr.to_csv(output_LD, sep = "\t", header = True, index = False,compression='gzip')
main(region,geno_path,sumstats_path,pheno_path,unr_path,imp_geno_path,imp_sumstats_path,imp_ref,output_sumstats,output_LD,bgen_sample_path)
1. Preprocess sumstats (regenie format) and extract it from a region
1.1. Region extraction
this region [5, 272741, 313528] has 498 SNPs
Total SNPs 119 . Flip SNPs 118
/home/yh3455/miniconda3/lib/python3.8/site-packages/pandas/core/generic.py:5516: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead
See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
self[name] = value
1.2. LiftOver the region
this region (5, 272856, 313643) has 1736 SNPs
Total SNPs 385 . Flip SNPs 0
1.3. Regional SNPs Liftover
1.1.1 Get exome unique sumstats and geno and Combine sumstats
keep exact flip reverse both complement
False False False False False False 106
True False True False False False 12
True False False False False 1
dtype: int64
2. Remove relative samples
3. Calculate LD matrix
/mnt/mfs/statgen/yin/Github/LDtools/LDtools/ldmatrix.py:29: RuntimeWarning: invalid value encountered in true_divide
geno_i = (geno_i - np.nanmean(geno_i,axis=0)[None,:])/np.nanstd(geno_i,axis=0)[None,:]
/mnt/mfs/statgen/yin/Github/LDtools/LDtools/genodata.py:74: PerformanceWarning: Slicing with an out-of-order index is generating 22098 times more chunks
geno = geno[:,idx]
/mnt/mfs/statgen/yin/Github/LDtools/LDtools/ldmatrix.py:67: RuntimeWarning: invalid value encountered in true_divide
geno_i = (geno_i - np.nanmean(geno_i,axis=0)[None,:])/np.nanstd(geno_i,axis=0)[None,:]
4. Output sumstats and LD matrix