def test_gwas_from_source(self):
#full format, OR+SE (which is like beta+se)
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_or":"OR",
"column_se":"SE",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
source = GWASUtilities.gwas_filtered_source("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz")
gwas = GWAS.load_gwas(source, gwas_format)
assert_gwas_zscore_fbse(self, gwas)
source = GWASUtilities.gwas_filtered_source("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", snps={"rs940550", "rs6650104", "rs61770173"}, snp_column_name="SNPID")
gwas = GWAS.load_gwas(source, gwas_format)
numpy.testing.assert_array_equal(gwas[SNP], pandas.Series(["rs940550", "rs6650104", "rs61770173", ], dtype=numpy.str))
numpy.testing.assert_array_equal(gwas[EFFECT_ALLELE], pandas.Series(["C", "T", "A"], dtype=numpy.str))
numpy.testing.assert_array_equal(gwas[NON_EFFECT_ALLELE], pandas.Series(["G", "C", "C"], dtype=numpy.str))
numpy.testing.assert_array_equal(gwas[CHROMOSOME], pandas.Series(["chr1", "chr1", "chr22"], dtype=numpy.str))
numpy.testing.assert_allclose(gwas[ZSCORE], pandas.Series([-1.254557, 0.974874, -0.232505],dtype=numpy.float32), rtol=0.001)
numpy.testing.assert_allclose(gwas[BETA], pandas.Series([-0.0217038334437866, 0.0193025022544974, -0.00369682484428976], dtype=numpy.float32), rtol=0.001)
numpy.testing.assert_allclose(gwas[SE], pandas.Series([0.0173, 0.0198, 0.0159], dtype=numpy.float32), rtol=0.001)
def test_gwas_from_source(self):
#full format, OR+SE (which is like beta+se)
gwas_format = {
"column_snp": "SNPID",
"column_non_effect_allele": "A2",
"column_effect_allele": "A1",
"column_or": "OR",
"column_se": "SE",
"column_chromosome": "HG19CHRC",
"column_position": "BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz",
gwas_format,
force_special_handling=True)
assert_gwas_zscore_fbse(self, gwas)
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz",
gwas_format,
snps={"rs940550", "rs6650104", "rs61770173"})
numpy.testing.assert_array_equal(
gwas[SNP],
pandas.Series([
"rs940550",
"rs6650104",
"rs61770173",
],
dtype=numpy.str))
numpy.testing.assert_array_equal(
gwas[EFFECT_ALLELE], pandas.Series(["C", "T", "A"],
dtype=numpy.str))
numpy.testing.assert_array_equal(
gwas[NON_EFFECT_ALLELE],
pandas.Series(["G", "C", "C"], dtype=numpy.str))
numpy.testing.assert_array_equal(
gwas[CHROMOSOME],
pandas.Series(["chr1", "chr1", "chr22"], dtype=numpy.str))
numpy.testing.assert_allclose(gwas[ZSCORE],
pandas.Series(
[-1.254557, 0.974874, -0.232505],
dtype=numpy.float32),
rtol=0.001)
numpy.testing.assert_allclose(
gwas[BETA],
pandas.Series([
-0.0217038334437866, 0.0193025022544974, -0.00369682484428976
],
dtype=numpy.float32),
rtol=0.001)
numpy.testing.assert_allclose(gwas[SE],
pandas.Series([0.0173, 0.0198, 0.0159],
dtype=numpy.float32),
rtol=0.001)
def build_betas(args, model, gwas_format, name):
logging.info("Building beta for %s and %s", name,
args.model_db_path if args.model_db_path else "no database")
load_from = os.path.join(args.gwas_folder,
name) if args.gwas_folder else name
snps = model.snps() if model else None
b = GWAS.load_gwas(load_from,
gwas_format,
snps=snps,
separator=args.separator,
skip_until_header=args.skip_until_header,
handle_empty_columns=args.handle_empty_columns,
input_pvalue_fix=args.input_pvalue_fix)
if model is not None:
PF = PredictionModel.WDBQF
base = model.weights[[
PF.K_RSID, PF.K_EFFECT_ALLELE, PF.K_NON_EFFECT_ALLELE
]].drop_duplicates()
b = GWASAndModels.align_data_to_alleles(b, base, Constants.SNP,
PF.K_RSID)
b = b.fillna("NA")
keep = [GWAS.SNP, GWAS.ZSCORE]
if GWAS.BETA in b: keep.append(GWAS.BETA)
b = b[keep]
return b
def build_betas(args, model, gwas_format, name):
logging.info("Building beta for %s and %s", name,
args.model_db_path if args.model_db_path else "no database")
load_from = os.path.join(args.gwas_folder, name)
if model or args.skip_until_header:
snps = model.snps() if model else None
snp_column_name = args.snp_column if model else None
load_from = GWASUtilities.gwas_filtered_source(
load_from,
snps=snps,
snp_column_name=snp_column_name,
skip_until_header=args.skip_until_header,
separator=args.separator)
sep = '\s+' if args.separator is None else args.separator
b = GWAS.load_gwas(load_from, gwas_format, sep=sep)
if model is not None:
PF = PredictionModel.WDBQF
base = model.weights[[
PF.K_RSID, PF.K_EFFECT_ALLELE, PF.K_NON_EFFECT_ALLELE
]].drop_duplicates()
b = align_data_to_alleles(b, base, Constants.SNP, PF.K_RSID)
b = b.fillna("NA")
keep = [GWAS.SNP, GWAS.ZSCORE]
if GWAS.BETA in b: keep.append(GWAS.BETA)
b = b[keep]
return b
def build_betas(args, model, gwas_format, name):
logging.info("Building beta for %s and %s", name, args.model_db_path if args.model_db_path else "no database")
load_from = os.path.join(args.gwas_folder, name) if args.gwas_folder else name
snps = model.snps() if model else None
b = GWAS.load_gwas(load_from, gwas_format, snps=snps, separator=args.separator,
skip_until_header=args.skip_until_header, handle_empty_columns=args.handle_empty_columns, input_pvalue_fix=args.input_pvalue_fix, keep_non_rsid=args.keep_non_rsid)
if model is not None:
PF = PredictionModel.WDBQF
base = model.weights[[PF.K_RSID, PF.K_EFFECT_ALLELE, PF.K_NON_EFFECT_ALLELE]].drop_duplicates()
b = GWASAndModels.align_data_to_alleles(b, base, Constants.SNP, PF.K_RSID)
b = b.fillna("NA")
keep = [GWAS.SNP, GWAS.ZSCORE]
if GWAS.BETA in b: keep.append(GWAS.BETA)
b = b[keep]
return b
def test_load_gwas(self):
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format, strict=False)
assert_basic_gwas(self, gwas)
#full format, OR+SE (which is like beta+se)
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_or":"OR",
"column_se":"SE",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_fbse(self, gwas)
# full format, beta+SE
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_beta":"BETA",
"column_se":"SE",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_fbse(self, gwas)
# full format, pvalue+beta
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_beta":"BETA",
"column_pvalue":"P",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_pb(self, gwas)
# full format, pvalue+beta_sign
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_beta_sign":"BETA_SIGN",
"column_pvalue":"P",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_pb(self, gwas)
# full format, pvalue+or
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_or":"OR",
"column_pvalue":"P",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_pb(self, gwas)
def build_betas(args, model, gwas_format, name, model_snp_map):
logging.info("Building beta for %s and %s", name,
args.model_db_path if args.model_db_path else "no database")
load_from = os.path.join(args.gwas_folder,
name) if args.gwas_folder else name
snps = model.snps() if model else None
b = GWAS.load_gwas(load_from,
gwas_format,
snps=snps,
separator=args.separator,
skip_until_header=args.skip_until_header,
handle_empty_columns=args.handle_empty_columns,
input_pvalue_fix=args.input_pvalue_fix,
keep_non_rsid=args.keep_non_rsid)
if model_snp_map:
logging.info("Loading mapping")
PF = PredictionModel.WDBQF
snp_map = pandas.read_table(model_snp_map)
snp_map_ = snp_map.rename(columns={
"a0": PF.K_NON_EFFECT_ALLELE,
"a1": PF.K_EFFECT_ALLELE
})[[
PF.K_RSID, PF.K_EFFECT_ALLELE, PF.K_NON_EFFECT_ALLELE,
"panel_variant_id", "panel_variant_a0", "panel_variant_a1", "swap"
]].drop_duplicates()
logging.info("Mapping variants")
columns = [x for x in b.columns.values]
b = GWASAndModels.align_data_to_alleles(b, snp_map_, Constants.SNP,
PF.K_RSID)
if GWAS.ZSCORE in b:
b = b.assign(zscore=b.zscore * b.swap)
if GWAS.BETA in b:
b = b.assign(beta=b.beta * b.swap)
b = b.rename(columns={GWAS.SNP:"gwas_snp", GWAS.EFFECT_ALLELE:"gwas_effect_allele", GWAS.NON_EFFECT_ALLELE:"gwas_non_effect_allele"})\
.drop(columns=[GWASAndModels.EA_BASE, GWASAndModels.NEA_BASE])\
.rename(columns={"panel_variant_id":GWAS.SNP, "panel_variant_a0":GWASAndModels.NEA, "panel_variant_a1":GWASAndModels.EA})\
[["gwas_snp", "gwas_effect_allele", "gwas_non_effect_allele"]+columns]
if model is not None:
logging.info("Aligning GWAS to models")
PF = PredictionModel.WDBQF
base = model.weights[[
PF.K_RSID, PF.K_EFFECT_ALLELE, PF.K_NON_EFFECT_ALLELE
]].drop_duplicates()
b = GWASAndModels.align_data_to_alleles(b, base, Constants.SNP,
PF.K_RSID)
b = b.drop(columns=[GWASAndModels.EA_BASE, GWASAndModels.NEA_BASE])
b = b.fillna("NA")
if model is not None:
logging.info("Trimming output")
keep = [GWAS.SNP, GWAS.ZSCORE]
if GWAS.BETA in b: keep.append(GWAS.BETA)
b = b[keep]
return b
def test_load_gwas(self):
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format, strict=False)
assert_basic_gwas(self, gwas)
#full format, OR+SE (which is like beta+se)
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_or":"OR",
"column_se":"SE",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_fbse(self, gwas)
# full format, beta+SE
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_beta":"BETA",
"column_se":"SE",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_fbse(self, gwas)
# full format, pvalue+beta
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_beta":"BETA",
"column_pvalue":"P",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_pb(self, gwas)
# full format, pvalue+beta_sign
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_beta_sign":"BETA_SIGN",
"column_pvalue":"P",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_pb(self, gwas)
# full format, pvalue+or
gwas_format = {
"column_snp":"SNPID",
"column_non_effect_allele":"A2",
"column_effect_allele":"A1",
"column_or":"OR",
"column_pvalue":"P",
"column_chromosome":"HG19CHRC",
"column_position":"BP"
}
gwas = GWAS.load_gwas("tests/_td/GWAS/scz2/scz2.gwas.results.txt.gz", gwas_format)
assert_gwas_zscore_pb(self, gwas)