def _get_coef(gene, models, matrix_manager, variances, t1, t2):
model_1 = models.loc[t1]
snps_1 = set(model_1.index.get_level_values(0))
if len(snps_1) == 0:
return None
model_2 = models.loc[t2]
snps_2 = set(model_2.index.get_level_values(0))
if len(snps_2) == 0:
return None
s1, s2, matrix = matrix_manager.get_2(gene, snps_1, snps_2)
if len(s1) == 0 or len(s2) == 0:
return None
w1 = model_1.loc[s1].weight.values
w2 = model_2.loc[s2].weight.values
denom = numpy.float64(numpy.sqrt(variances[t1] * variances[t2]))
if denom == 0:
return numpy.nan
num = numpy.float64(_d(_d(w1, matrix), w2))
coef = num / denom
return coef
def _get_coef_2(gene, models, matrix_manager, variances, t1, t2):
model_1 = models[t1]
snps_1 = set(model_1.keys())
if len(snps_1) == 0:
return None
model_2 = models[t2]
snps_2 = set(model_2.keys())
if len(snps_2) == 0:
return None
s1, s2, matrix = matrix_manager.get_2(gene, snps_1, snps_2)
if len(s1) == 0 or len(s2) == 0:
return None
w1 = numpy.array([model_1[x] for x in s1], dtype=numpy.float64)
w2 = numpy.array([model_2[x] for x in s2], dtype=numpy.float64)
denom = numpy.float64(numpy.sqrt(variances[t1] * variances[t2]))
if denom == 0:
return numpy.nan
num = numpy.float64(_d(_d(w1, matrix), w2))
coef = num / denom
return coef
def _get_coef(gene, models, matrix_manager, variances, t1, t2):
model_1 = models.loc[t1]
snps_1 = set(model_1.index.get_level_values(0))
if len(snps_1) == 0:
return None
model_2 = models.loc[t2]
snps_2 = set(model_2.index.get_level_values(0))
if len(snps_2) == 0:
return None
s1, s2, matrix = matrix_manager.get_2(gene, snps_1, snps_2)
if len(s1) == 0 or len(s2) == 0:
return None
w1 = model_1.loc[s1].weight.values
w2 = model_2.loc[s2].weight.values
denom = numpy.float64(numpy.sqrt(variances[t1] * variances[t2]))
if denom == 0:
return numpy.nan
num = numpy.float64(_d(_d(w1, matrix), w2))
coef = num/denom
return coef
def _get_coef_2(gene, models, matrix_manager, variances, t1, t2):
model_1 = models[t1]
snps_1 = set(model_1.keys())
if len(snps_1) == 0:
return None
model_2 = models[t2]
snps_2 = set(model_2.keys())
if len(snps_2) == 0:
return None
s1, s2, matrix = matrix_manager.get_2(gene, snps_1, snps_2)
if len(s1) == 0 or len(s2) == 0:
return None
w1 = numpy.array([model_1[x] for x in s1], dtype=numpy.float64)
w2 = numpy.array([model_2[x] for x in s2], dtype=numpy.float64)
denom = numpy.float64(numpy.sqrt(variances[t1] * variances[t2]))
if denom == 0:
return numpy.nan
num = numpy.float64(_d(_d(w1, matrix), w2))
coef = num/denom
return coef
def _get_variance(models, matrix_manager, gene, tissue):
model = models.loc[tissue]
snps = set(model.index.get_level_values(0).values)
#Remember that only those snps with data in the GWAS will get loaded.
snps, matrix = matrix_manager.get(gene, snps, strict_whitelist=False)
if len(snps) == 0:
return None
weights = model.loc[snps].weight.values
variance = _d(_d(weights, matrix), weights)
variance = numpy.float64(variance)
return variance
def _get_variance_2(models, matrix_manager, gene, tissue):
model = models[tissue]
snps = set(model.keys())
#Remember that only those snps with data in the GWAS will get loaded.
snps, matrix = matrix_manager.get(gene, snps, strict_whitelist=False)
if len(snps) == 0:
return None
weights = numpy.array([model[x] for x in snps], dtype=numpy.float64)
variance = _d(_d(weights, matrix), weights)
variance = numpy.float64(variance)
return variance
def _get_variance(models, matrix_manager, gene, tissue):
model = models.loc[tissue]
snps = set(model.index.get_level_values(0).values)
#Remember that only those snps with data in the GWAS will get loaded.
snps, matrix = matrix_manager.get(gene, snps, strict_whitelist=False)
if len(snps) == 0:
return None
weights = model.loc[snps].weight.values
variance = _d(_d(weights, matrix), weights)
variance = numpy.float64(variance)
return variance
def _get_variance_2(models, matrix_manager, gene, tissue):
model = models[tissue]
snps = set(model.keys())
#Remember that only those snps with data in the GWAS will get loaded.
snps, matrix = matrix_manager.get(gene, snps, strict_whitelist=False)
if len(snps) == 0:
return None
weights = numpy.array([model[x] for x in snps], dtype=numpy.float64)
variance = _d(_d(weights, matrix), weights)
variance = numpy.float64(variance)
return variance