Phenotype tool
Example usage of PhenoTool class
First some imports:
from DAE import *
from pheno_tool.tool import *
Then prepare studies to work with:
studies = vDB.get_studies('ALL SSC')
studies.append(vDB.get_study('w1202s766e611'))
Now we can create an instance of PhenoTool class to work with given set of individuals:
tool = PhenoTool(
phdb, studies, roles=['prb', 'mom', 'dad'],
measure_id='ssc_core_descriptive.ssc_diagnosis_nonverbal_iq',
)
We can check the list of individuals that are subject to computation by this particular instance of PhenoTool by using list_of_subjects() method:
tool.list_of_subjects()
will output following dictionary:
{
u'14566.p1': Person(14566.p1; prb; M),
u'14398.p1': Person(14398.p1; prb; M),
u'14524.p1': Person(14524.p1; prb; M),
u'12607.p1': Person(12607.p1; prb; M),
...
}
We also have access to a Pandas data frame representing these subjects:
persons = tool.list_of_subject_df()
persons.head()
will output initial rows of the subjects dataframe:
person_id family_id role gender ssc_core_descriptive.ssc_diagnosis_nonverbal_iq normalized
2 11000.p1 11000 prb M 78.0 78.0
5 11001.p1 11001 prb M 123.0 123.0
8 11002.p1 11002 prb F 111.0 111.0
11 11003.p1 11003 prb M 108.0 108.0
14 11004.p1 11004 prb M 74.0 74.0
Now we are ready for the actual computations. Let first get some set of gene symbols:
from DAE import get_gene_sets_symNS
gt = get_gene_sets_symNS('main')
gene_syms = gt.t2G['autism candidates from Iossifov PNAS 2015'].keys()
To perform the actual computation we use calc method of PhenoTool. We need to pass to this method the specification of the variants:
res = tool.calc(
VT(
effect_types=['LGDs'],
gene_syms=gene_syms,
present_in_child=['autism only', 'autism and unaffected'],
present_in_parent=['neither'],
))
The returned object is of type
PhenoResult. This class
could be queried about the results of the calculation:
In [23]: res.positive_count
Out[23]: 135
In [24]: res.negative_count
Out[24]: 2626
In [25]: res.pvalue
Out[25]: 1.5345885574583403e-07
We also have access to phenotypes and genotypes as dataframes:
In [26]: res.genotypes_df.head()
Out[26]:
person_id gender role variants
2 11000.p1 M prb 0
5 11001.p1 M prb 0
8 11002.p1 F prb 0
11 11003.p1 M prb 0
14 11004.p1 M prb 0
In [27]: res.phenotypes_df.head()
Out[27]:
person_id role gender ssc_core_descriptive.ssc_diagnosis_nonverbal_iq normalized
2 11000.p1 prb M 78.0 78.0
5 11001.p1 prb M 123.0 123.0
8 11002.p1 prb F 111.0 111.0
11 11003.p1 prb M 108.0 108.0
14 11004.p1 prb M 74.0 74.0
and as dictionaries:
In [28]: res.genotypes
Out[28]:
Counter({
u'11046.p1': 0,
u'14398.p1': 0,
u'12547.p1': 1,
u'14021.p1': 0,
u'11264.p1': 0,
u'11925.p1': 1,
u'11249.p1': 0,
....
})
In [29]: res.phenotypes
Out[29]:
{u'14566.p1': {'gender': u'M',
'normalized': 44.0,
'person_id': u'14566.p1',
'role': u'prb',
'ssc_core_descriptive.ssc_diagnosis_nonverbal_iq': 44.0},
u'14398.p1': {'gender': u'M',
'normalized': 111.0,
'person_id': u'14398.p1',
'role': u'prb',
'ssc_core_descriptive.ssc_diagnosis_nonverbal_iq': 111.0},
u'14524.p1': {'gender': u'M',
'normalized': 76.0,
'person_id': u'14524.p1',
'role': u'prb',
'ssc_core_descriptive.ssc_diagnosis_nonverbal_iq': 76.0},
...
}
Example usage of PhenoTool class with pheno filters
The PhenoTool class supports different ways to specify which individuals are subject to the calculations. We can use roles, measure_id, normalize_by and pheno_filters arguments in the PhenoTool constructor. For example if we specify roles and measure_id the tool will use individuals of the specified roles and has the specified measurement:
tool = PhenoTool(phdb, studies, roles=['prb', 'sib'],
measure_id="ssc_commonly_used.head_circumference")
and if whe check list of subjects:
In [41]: len(tool.list_of_subjects())
Out[41]: 5166
If we specify a normalization, then the list of individuals should have measurements for normalization too:
tool = PhenoTool(phdb, studies, roles=['prb', 'sib'],
measure_id="ssc_commonly_used.head_circumference",
normalize_by=['pheno_common.age'])
In [45]: len(tool.list_of_subjects())
Out[45]: 4869
Additionally we can use pheno_filters argument. Let us specify only individuals that are of white race:
tool = PhenoTool(phdb, studies, roles=['prb', 'sib'],
measure_id="ssc_commonly_used.head_circumference",
pheno_filters={
'pheno_common.race': set(['white'])
})
In [47]: len(tool.list_of_subjects())
Out[47]: 3983
or:
tool = PhenoTool(phdb, studies, roles=['prb', 'sib'],
measure_id="ssc_commonly_used.head_circumference",
pheno_filters={
'pheno_common.race': set(['white', 'asian']),
})
In [51]: len(tool.list_of_subjects())
Out[51]: 4174
Using pheno_filters we can also specify ranges of values for given measurement:
tool = PhenoTool(phdb, studies, roles=['prb', 'sib'],
measure_id="ssc_commonly_used.head_circumference",
pheno_filters={
'pheno_common.race': set(['white', 'asian']),
'pheno_common.age': (200, 220),
})
In [53]: len(tool.list_of_subjects())
Out[53]: 145
Example with vineland_ii
Example for iterating on many measures:
from DAE import *
from pheno_tool.tool import *
from pheno_tool.genotype_helper import GenotypeHelper
studies = vDB.get_studies('IossifovWE2014')
genotype_helper = GenotypeHelper(studies)
effect_types = ['LGDs', 'missense', 'nonsynonymous']
genotypes = {}
for et in effect_types:
variants_type = VT(
effect_types=[et],
present_in_parent=['neither'],
present_in_child=['autism only',
'autism and unaffected',
'unaffected only']
)
persons_variants = genotype_helper.get_persons_variants_df(variants_type)
genotypes[et] = persons_variants
result = {}
for measure_id in phdb.get_instrument_measures('vineland_ii'):
measure = phdb.get_measure(measure_id)
if measure.measure_type != "continuous":
continue
cols = [measure_id]
for role in ['prb', 'sib']:
tool = PhenoTool(phdb, studies, roles=[role], measure_id=measure_id)
for et in effect_types:
print("working on: {}, {}, {}".format(measure_id, role, et))
res = tool.calc(genotypes[et])
cols.append("%.3f" % res.pvalue)
result[measure_id] = cols