def generate_trios(pedfile, f1=True):
"""
Given a PED file, specify whether you want
to output trios w/r/t the F1 or F2 generation
(i.e., whether the kid in each trio is an F1 or F2).
"""
from peddy import Ped
ped = Ped(pedfile)
p0s = [k for k in ped.samples() if k.mom is None]
f1s = [k for k in ped.samples() if k.mom and k.dad and k.mom.mom is None]
f2s = [k for k in ped.samples() if k.mom and k.mom.mom]
if f1: trios = f1s
else: trios = f2s
for i in trios:
yield (i.sample_id, i.mom.sample_id, i.dad.sample_id)
def create_samples(self):
ped = Ped(self.ped_path)
cols = [
'sample_id', 'family_id', 'name', 'paternal_id', 'maternal_id',
'sex', 'phenotype'
]
if ped.header is None:
ped.header = [x for x in cols if x != 'name']
samples = [fix_sample_name(s) for s in self.vcf.samples]
cols = [
'sample_id', 'family_id', 'name', 'paternal_id', 'maternal_id',
'sex', 'phenotype'
]
idxs, rows, not_in_vcf = [], [], []
cols.extend(ped.header[6:])
sample_id = 1
for i, s in enumerate(ped.samples(), start=1):
try:
idxs.append(samples.index(fix_sample_name(s.sample_id)))
except ValueError:
not_in_vcf.append(s.sample_id)
continue
rows.append([
sample_id,
s.family_id,
fix_sample_name(s.sample_id),
fix_sample_name(str(s.paternal_id)),
fix_sample_name(str(s.maternal_id)),
'1' if s.sex == 'male' else '2' if s.sex == 'female' else '-9',
'2' if s.affected is True else
'1' if s.affected is False else '-9',
] + s.attrs)
sample_id += 1
if len(not_in_vcf) > 0:
print("not in VCF: %s" % ",".join(not_in_vcf), file=sys.stderr)
scols = [sql.Column('sample_id', sql.Integer, primary_key=True)]
for i, col in enumerate(cols[1:], start=1):
vals = None
try:
vals = [r[i] for r in rows]
l = max(len(v) for v in vals)
scols.append(sql.Column(col, Unicode(l)))
except:
print(col, vals, file=sys.stderr)
raise
t = sql.Table('samples', self.metadata, *scols)
t.drop(checkfirst=True)
t.create()
self.engine.execute(t.insert(), [dict(zip(cols, r)) for r in rows])
# track the order to pull from the genotype fields.
self.sample_idxs = np.array(idxs)
return [r[2] for r in rows]
def create_samples(self):
ped = Ped(self.ped_path)
cols = ["sample_id", "family_id", "name", "paternal_id", "maternal_id", "sex", "phenotype"]
if ped.header is None:
ped.header = [x for x in cols if x != "name"]
samples = [fix_sample_name(s) for s in self.vcf.samples]
cols = ["sample_id", "family_id", "name", "paternal_id", "maternal_id", "sex", "phenotype"]
idxs, rows, not_in_vcf = [], [], []
cols.extend(ped.header[6:])
sample_id = 1
for i, s in enumerate(ped.samples(), start=1):
try:
idxs.append(samples.index(fix_sample_name(s.sample_id)))
except ValueError:
not_in_vcf.append(s.sample_id)
continue
rows.append(
[
sample_id,
s.family_id,
fix_sample_name(s.sample_id),
fix_sample_name(str(s.paternal_id)),
fix_sample_name(str(s.maternal_id)),
"1" if s.sex == "male" else "2" if s.sex == "female" else "-9",
"2" if s.affected is True else "1" if s.affected is False else "-9",
]
+ s.attrs
)
sample_id += 1
if len(not_in_vcf) > 0:
print("not in VCF: %s" % ",".join(not_in_vcf), file=sys.stderr)
scols = [sql.Column("sample_id", sql.Integer, primary_key=True)]
for i, col in enumerate(cols[1:], start=1):
vals = None
try:
vals = [r[i] for r in rows]
l = max(len(v) for v in vals)
scols.append(sql.Column(col, Unicode(l)))
except:
print(col, vals, file=sys.stderr)
raise
t = sql.Table("samples", self.metadata, *scols)
t.drop(checkfirst=True)
t.create()
self.engine.execute(t.insert(), [dict(zip(cols, r)) for r in rows])
# track the order to pull from the genotype fields.
self.sample_idxs = np.array(idxs)
return [r[2] for r in rows]
def test_6():
p = Ped(op.join(HERE, 'peddy/tests/a6.ped'))
assert len(list(p.samples())) == 14
for sam in p.samples():
assert sam.family_id[:3] == "fam"
def test_getattr():
p = Ped(op.join(HERE, 'peddy/tests/a.ped'))
li = list(p.samples(ethnicity='caucasianNEuropean'))
assert len(li) == 5
for item in li:
assert item.ethnicity == 'caucasianNEuropean'
def test_ped():
p = Ped(op.join(HERE, 'peddy/tests/a.ped'))
assert len(p.families) == 4
assert len(list(p.samples())) == 14
def run(args):
ped = Ped(args.ped)
vcf = VCF(args.vcf, gts012=True)
ped_samples = [s.sample_id for s in ped.samples()]
vcf_samples = set(vcf.samples)
samples = [s for s in ped_samples if s in vcf_samples]
exclude = read_exclude(args.exclude)
vcf = VCF(args.vcf, samples=samples, gts012=True)
if args.region:
vcf_iter = vcf(args.region)
else:
vcf_iter = vcf
pctile1 = 10
# build a dict of sample_id to sample index
smp2idx = dict(zip(vcf.samples, range(len(vcf.samples))))
# get the Ped objects for the family of interest
if args.families is None:
fams = ped.families.values()
else:
fams = [ped.families[f] for f in args.families.split(",")]
if len(fams) == 0:
sys.exit('Families %s not found in ped file' % args.families)
# create a simple dictionary of info for each family member
fs = [get_family_dict(fam, smp2idx, args) for fam in fams]
del fam
fsites = open("%s.sites" % args.prefix, "w")
# fcalls contains the crossovers for all samples.
try:
vcf["AB"]
has_abs = True
except KeyError:
has_abs = False
smp2gt = defaultdict(int)
nused, i, report_at, t0 = 0, 0, 20000, time.time()
for i, v in enumerate(vcf_iter, start=1):
if i % report_at == 0:
persec = i / float(time.time() - t0)
print("%s:%d (%.1f/sec) %.2f%% informative (%d/%d variants)" %
(v.CHROM, v.POS, persec, 100.0 * nused / i, nused, i),
file=sys.stderr)
if i == 20000:
report_at = 100000
if i == 100000:
report_at = 200000
for f in fs:
for k in f:
if k.startswith('fh'): f[k].flush()
sys.stderr.flush()
if v.var_type != 'snp': continue ### no indels
if len(v.ALT) > 1: continue
#if len(v.REF) > 3 or len(v.ALT) > 1 or len(v.ALT[0]) > 3:
# continue
#if v.call_rate < 0.95: continue
if v.call_rate < 0.90: continue
if v.FILTER is not None: continue
if int(v.INFO.get('AC')) == 1: continue
if exclude is not None and 0 != len(exclude[v.CHROM].search(
v.start, v.end)):
continue
# expensive to get gt_bases and we only need it at the crossover.
gt_bases = None
gt_types, gt_quals, gt_depths = v.gt_types, v.gt_quals, v.gt_depths
### added by tom 2019-11-13, as 1st percentile of depths on chr17
### were often negative...
#gt_depths[gt_depths < 0] = 0
gt_phases = v.gt_phases
ipctiles, pctiles = None, None
sample_abs = None
nsites = 0 # track the number of families that had this as an informative site.
for f in fs:
#if ipctiles is not None and ipctiles[0] < pctile1:
# break
# is_informative only needs gt_types, so we check that first...
add_genotype_info(f, gt_types=gt_types, gt_phases=gt_phases)
# ############## PHASED ####################
if all(f['gt_phase']):
if gt_bases is None:
gt_bases = v.gt_bases
nsites += phased_check(f, v, gt_bases)
continue
# ############ END PHASED ####################
# need exactly 1 het parent for unphased checks.
if 1 != ((f['gt_type'][0] == HET) + (f['gt_type'][1] == HET)):
continue
if not is_informative(f):
continue
# now wee need to add quality and depth.
add_genotype_info(f, gt_quals=gt_quals, gt_depths=gt_depths)
if not passes_quality_control(f, args):
continue
# detect crossovers.
for parent, (p1, p2) in [("dad", (0, 1)), ("mom", (1, 0))]:
if not (f['gt_type'][p1] == HET
and f['gt_type'][p2] == HOM_REF):
continue
if gt_bases is None:
gt_bases = v.gt_bases
if sample_abs is None:
sample_abs = get_allele_balance(v, has_abs)
if sample_abs is None: break
fam_abs = sample_abs[f['idxs']]
if all(np.isnan(fam_abs)): continue
off = 0.31 # require that off <= alt/(ref+alt) <= 1-off
if ((fam_abs[p1] >= 1 - off) | (fam_abs[p1] <= off)): continue
if np.any((1 - off < fam_abs[2:]) | (fam_abs[2:] <= off)):
continue
fam_bases = "\t".join(gt_bases[f['idxs']])
fam_abs = "|".join("%.2f" % val for val in fam_abs)
# calculate on first use. we found that having a low 1st pctile
# was a good indicator of increased chance of spurious XO even
# in families with decent depth.
#print (np.mean(gt_depths), np.median(gt_depths))
if pctiles is None:
ipctiles = np.percentile(gt_depths, (1, 5, 10, 50, 90))
pctiles = "|".join("%.0f" % de for de in ipctiles)
#if ipctiles[0] < pctile1:
# break
fam_depths = "|".join(map(str, gt_depths[f['idxs']]))
nsites += 1
val = 1 if f['gt_type'][2] == f['gt_type'][3] else 0
f['fh-%s' % parent].write('\t'.join(
str(s) for s in [
v.CHROM, v.POS - 1, v.POS, f['ids'][p1],
f['family_id'], val, fam_bases, fam_depths,
"%.2f" % v.call_rate, pctiles, fam_abs
]) + '\n')
fsites.write("%s:%d\t%d\n" % (v.CHROM, v.POS, nsites))
if nsites > 0:
nused += 1
fsites.close()
persec = i / float(time.time() - t0)
print("finished at %s:%d (%.1f/sec) %.2f%% informative (%d/%d variants)" %
(v.CHROM, v.POS, persec, 100.0 * nused / i, nused, i),
file=sys.stderr)
kept = _remove_empty(fs)
call_all(kept, args.prefix, min_sites=20)
def test_ped():
p = Ped('peddy/tests/a.ped')
assert len(p.families) == 4
assert len(list(p.samples())) == 14
def test_6():
p = Ped(op.join(HERE, 'peddy/tests/a6.ped'))
assert len(list(p.samples())) == 14
for sam in p.samples():
assert sam.family_id[:3] == "fam"
def test_getattr():
p = Ped(op.join(HERE, 'peddy/tests/a.ped'))
li = list(p.samples(ethnicity='caucasianNEuropean'))
assert len(li) == 5
for item in li:
assert item.ethnicity == 'caucasianNEuropean'
def test_ped():
p = Ped(op.join(HERE, 'peddy/tests/a.ped'))
assert len(p.families) == 4
assert len(list(p.samples())) == 14
def run(args):
print ('\t'.join(['chrom', 'start', 'end', 'sample_id', 'parent_id', 'n_vars', 'hap_start']))
vcf = VCF(args.vcf)
ped = Ped(args.ped)
samples = [s for s in ped.samples()]
kids = [s for s in samples if s.mom is not None and s.dad is not None]
fams = set([s.family_id for s in samples])
smp2ped = dict(zip([s.sample_id for s in samples], samples))
exclude = read_exclude(args.exclude)
#kid = smp2ped[args.kid].sample_id
#dad, mom = smp2ped[kid].paternal_id, smp2ped[kid].maternal_id
#samples_in_ped = [s.sample_id for s in samples]
#samples_in_fam = [s.sample_id for s in samples if s.family_id == smp2ped[kid].family_id]
# restrict VCF to the samples in the current family
#vcf.set_samples(samples_in_ped)
smp2idx = dict(zip(vcf.samples, range(len(vcf.samples))))
bad_positions = []
v_feats = []
haps = defaultdict()
inf_positions = defaultdict(list)
fam_dict = defaultdict(lambda: defaultdict())
for fam in fams:
mom = [s for s in samples if s.family_id == fam and s.mom is None and s.dad is None and s.sex == 'female'][0]
dad = [s for s in samples if s.family_id == fam and s.mom is None and s.dad is None and s.sex == 'male'][0]
sibs = [s.sample_id for s in kids if s.dad == dad and s.mom == mom]
fam_dict[fam]['mom'] = mom
fam_dict[fam]['dad'] = dad
fam_dict[fam]['sibs'] = ','.join(sibs)
nused, i, report_at, t0 = 0, 0, 1000, time.time()
for i,v in enumerate(vcf(args.chrom)):
if i != 0 and i % report_at == 0:
persec = i / float(time.time() - t0)
print("%s:%d (%.1f/sec) %.2f%% informative (%d/%d variants)" % (v.CHROM, v.POS,
persec, 100.0 * nused/i, nused, i), file=sys.stderr)
if args.exclude and len(exclude[v.CHROM].search(v.start, v.end)) > 0: continue
if v.var_type != 'snp': continue
if v.FILTER not in ('PASS', None): continue
if v.call_rate < 0.90: continue
if len(v.ALT) > 1: continue
if len(v.ALT[0]) > 1: continue
gts = v.gt_types
quals = v.gt_quals
rd, ad = v.gt_ref_depths, v.gt_alt_depths
for fam in fam_dict:
mom = fam_dict[fam]['mom']
dad = fam_dict[fam]['dad']
sibs = fam_dict[fam]['sibs'].split(',')
if args.inf_parent and args.inf_parent not in (mom, dad): continue
try:
mi, di = smp2idx[mom.sample_id], smp2idx[dad.sample_id]
except KeyError: sys.exit()
# ensure we're at an informative site
if gts[mi] != HOM_REF and gts[di] != HOM_REF: continue
if gts[mi] == HOM_REF and gts[di] == HOM_REF: continue
sib_gts = [gts[smp2idx[k]] for k in sibs]
inf_parent = None
if (gts[mi] == HET and gts[di] == HOM_REF): inf_parent = mom.sample_id
elif (gts[di] == HET and gts[mi] == HOM_REF): inf_parent = dad.sample_id
else: continue
# check that both parents are "high-quality"
if not is_good_site(mi, quals, gts, ad, rd, het_ab=args.ab): continue
if not is_good_site(di, quals, gts, ad, rd, het_ab=args.ab): continue
if args.inf_parent and inf_parent != args.inf_parent: continue
# catalog the "states" of each child w/r/t the informative parent
# if kids are HETs, their state w/r/t to the INF parent is 0
states = []
sib_pass = []
for i,k in enumerate(sibs):
k_idx = smp2idx[k]
k_pass = is_good_site(k_idx, quals, gts, ad, rd, het_ab=args.ab)
sib_pass.append(k_pass)
k_state = -1
if v.CHROM in ('chrX', 'X') and k.sex == "male":
k_state = 0 if gts[k_idx] == HOM_ALT else 1
else:
k_state = 0 if gts[k_idx] == HET else 1
states.append(k_state)
if not all([x is True for x in sib_pass]): continue
if sum([s < 0 for s in states]) > 0: continue
if inf_parent not in haps:
haps[inf_parent] = np.array(states)
else:
haps[inf_parent] = np.vstack((haps[inf_parent], np.array(states)))
inf_positions[inf_parent].append(v.start)
nused += 1
persec = i / float(time.time() - t0)
xos = get_xo(args, haps, inf_positions)