Module dadi.Misc
Miscellaneous utility functions. Including ms simulation.
Expand source code
"""
Miscellaneous utility functions. Including ms simulation.
"""
import collections,os,sys,time, warnings
def simple_warning(message, category, filename, lineno, file=None, line=None):
return '%s' % message
warnings.formatwarning = simple_warning
import numpy
import scipy.linalg
from .Spectrum_mod import Spectrum
# Nucleotide order assumed in Q matrices.
code = 'CGTA'
#: Storage for times at which each stream was flushed.
__times_last_flushed = {}
def delayed_flush(stream=sys.stdout, delay=1):
"""
Flush a stream, ensuring that it is only flushed every 'delay' *minutes*.
Note that upon the first call to this method, the stream is not flushed.
stream: The stream to flush. For this to work with simple 'print'
statements, the stream should be sys.stdout.
delay: Minimum time *in minutes* between flushes.
This function is useful to prevent I/O overload on the cluster.
"""
global __times_last_flushed
curr_time = time.time()
# If this is the first time this method has been called with this stream,
# we need to fill in the times_last_flushed dict. setdefault will do this
# without overwriting any entry that may be there already.
if stream not in __times_last_flushed:
__times_last_flushed[stream] = curr_time
last_flushed = __times_last_flushed[stream]
# Note that time.time() returns values in seconds, hence the factor of 60.
if (curr_time - last_flushed) >= delay*60:
stream.flush()
__times_last_flushed[stream] = curr_time
def ensure_1arg_func(var):
"""
Ensure that var is actually a one-argument function.
This is primarily used to convert arguments that are constants into
trivial functions of time for use in integrations where parameters are
allowed to change over time.
"""
if numpy.isscalar(var):
# If a constant was passed in, use lambda to make it a nice
# simple function.
var_f_tmp = lambda t: var
else:
var_f_tmp = var
# Wrapping arguments in float64 eases working with CUDA
var_f = lambda t: numpy.float64(var_f_tmp(t))
if not callable(var_f):
raise ValueError('Argument is not a constant or a function.')
try:
var_f(0.0)
except TypeError:
raise ValueError('Argument is not a constant or a one-argument '
'function.')
return var_f
def ms_command(theta, ns, core, iter, recomb=0, rsites=None, seeds=None):
"""
Generate ms command for simulation from core.
theta: Assumed theta
ns: Sample sizes
core: Core of ms command that specifies demography.
iter: Iterations to run ms
recomb: Assumed recombination rate
rsites: Sites for recombination. If None, default is 10*theta.
seeds: Seeds for random number generator. If None, ms default is used.
Otherwise, three integers should be passed. Example: (132, 435, 123)
"""
if len(ns) > 1:
ms_command = "ms %(total_chrom)i %(iter)i -t %(theta)f -I %(numpops)i "\
"%(sample_sizes)s %(core)s"
else:
ms_command = "ms %(total_chrom)i %(iter)i -t %(theta)f %(core)s"
if recomb:
ms_command = ms_command + " -r %(recomb)f %(rsites)i"
if not rsites:
rsites = theta*10
sub_dict = {'total_chrom': numpy.sum(ns), 'iter': iter, 'theta': theta,
'numpops': len(ns), 'sample_sizes': ' '.join(map(str, ns)),
'core': core, 'recomb': recomb, 'rsites': rsites}
ms_command = ms_command % sub_dict
if seeds is not None:
seed_command = " -seeds %i %i %i" % (seeds[0], seeds[1], seeds[2])
ms_command = ms_command + seed_command
return ms_command
def perturb_params(params, fold=1, lower_bound=None, upper_bound=None):
"""
Generate a perturbed set of parameters.
Each element of params is radomly perturbed <fold> factors of 2 up or down.
fold: Number of factors of 2 to perturb by
lower_bound: If not None, the resulting parameter set is adjusted to have
all value greater than lower_bound.
upper_bound: If not None, the resulting parameter set is adjusted to have
all value less than upper_bound.
"""
pnew = params * 2**(fold * (2*numpy.random.uniform(size=len(params))-1))
if lower_bound is not None:
for ii,bound in enumerate(lower_bound):
if bound is None:
lower_bound[ii] = -numpy.inf
pnew = numpy.maximum(pnew, 1.01*numpy.asarray(lower_bound))
if upper_bound is not None:
for ii,bound in enumerate(upper_bound):
if bound is None:
upper_bound[ii] = numpy.inf
pnew = numpy.minimum(pnew, 0.99*numpy.asarray(upper_bound))
return pnew
def make_fux_table(fid, ts, Q, tri_freq):
"""
Make file of 1-fux for use in ancestral misidentification correction.
fid: Filename to output to.
ts: Expected number of substitutions per site between ingroup and outgroup.
Q: Trinucleotide transition rate matrix. This should be a 64x64 matrix, in
which entries are ordered using the code CGTA -> 0,1,2,3. For example,
ACT -> 3*16+0*4+2*1=50. The transition rate from ACT to AGT is then
entry 50,54.
tri_freq: Dictionary in which each entry maps a trinucleotide to its
ancestral frequency. e.g. {'AAA': 0.01, 'AAC':0.012...}
Note that should be the frequency in the entire region scanned
for variation, not just sites where there are SNPs.
"""
# Ensure that the *columns* of Q sum to zero.
# That is the correct condition when Q_{i,j} is the rate from i to j.
# This indicates a typo in Hernandez, Williamson, and Bustamante.
for ii in range(Q.shape[1]):
s = Q[:,ii].sum() - Q[ii,ii]
Q[ii,ii] = -s
eQhalf = scipy.linalg.matfuncs.expm(Q * ts/2.)
if not hasattr(fid, 'write'):
newfile = True
fid = open(fid, 'w')
outlines = []
for first_ii,first in enumerate(code):
for x_ii,x in enumerate(code):
for third_ii,third in enumerate(code):
# This the index into Q and eQ
xind = 16*first_ii+4*x_ii+1*third_ii
for u_ii,u in enumerate(code):
# This the index into Q and eQ
uind = 16*first_ii+4*u_ii+1*third_ii
## Note that the Q terms factor out in our final
## calculation, because for both PMuUu and PMuUx the final
## factor in Eqn 2 is P(S={u,x}|M=u).
#Qux = Q[uind,xind]
#denomu = Q[uind].sum() - Q[uind,uind]
PMuUu, PMuUx = 0,0
# Equation 2 in HWB. We have to generalize slightly to
# calculate PMuUx. In calculate PMuUx, we're summing over
# alpha the probability that the MRCA was alpha, and it
# substituted to x on the outgroup branch, and it
# substituted to u on the ingroup branch, and it mutated to
# x in the ingroup (conditional on it having mutated in the
# ingroup). Note that the mutation to x condition cancels
# in fux, so we don't bother to calculate it.
for aa,alpha in enumerate(code):
aind = 16*first_ii+4*aa+1*third_ii
pia = tri_freq[first+alpha+third]
Pau = eQhalf[aind,uind]
Pax = eQhalf[aind,xind]
PMuUu += pia * Pau*Pau
PMuUx += pia * Pau*Pax
# This is 1-fux. For a given SNP with actual ancestral state
# u and derived allele x, this is 1 minus the probability
# that the outgroup will have u.
# Eqn 3 in HWB.
res = 1 - PMuUu/(PMuUu + PMuUx)
# These aren't SNPs, so we can arbitrarily set them to 0
if u == x:
res = 0
outlines.append('%c%c%c %c %.6f' % (first,x,third,u,res))
fid.write(os.linesep.join(outlines))
if newfile:
fid.close()
def zero_diag(Q):
"""
Copy of Q altered such that diagonal entries are all 0.
"""
Q_nodiag = Q.copy()
for ii in range(Q.shape[0]):
Q_nodiag[ii,ii] = 0
return Q_nodiag
def tri_freq_dict_to_array(tri_freq_dict):
"""
Convert dictionary of trinucleotide frequencies to array in correct order.
"""
tripi = numpy.zeros(64)
for ii,left in enumerate(code):
for jj,center in enumerate(code):
for kk,right in enumerate(code):
row = ii*16 + jj*4 + kk
tripi[row] = tri_freq_dict[left+center+right]
return tripi
def total_instantaneous_rate(Q, pi):
"""
Total instantaneous substitution rate.
"""
Qzero = zero_diag(Q)
return numpy.dot(pi, Qzero).sum()
def make_data_dict(filename):
"""
Parse SNP file and store info in a properly formatted dictionary.
filename: Name of file to work with.
This is specific to the particular data format described on the wiki.
Modification for other formats should be straightforward.
The file can be zipped (extension .zip) or gzipped (extension .gz). If
zipped, there must be only a single file in the zip archive.
"""
if os.path.splitext(filename)[1] == '.gz':
import gzip
f = gzip.open(filename)
elif os.path.splitext(filename)[1] == '.zip':
import zipfile
archive = zipfile.ZipFile(filename)
namelist = archive.namelist()
if len(namelist) != 1:
raise ValueError('Must be only a single data file in zip '
'archive: %s' % filename)
f = archive.open(namelist[0])
else:
f = open(filename)
# Skip to the header
while True:
header = f.readline()
if not header.startswith('#'):
break
allele2_index = header.split().index('Allele2')
# Pull out our pop ids
pops = header.split()[3:allele2_index]
# The empty data dictionary
data_dict = {}
# Now walk down the file
for SNP_ii, line in enumerate(f):
if line.startswith('#'):
continue
# Split the into fields by whitespace
spl = line.split()
data_this_snp = {}
# We convert to upper case to avoid any issues with mixed case between
# SNPs.
data_this_snp['context'] = spl[0].upper()
data_this_snp['outgroup_context'] = spl[1].upper()
data_this_snp['outgroup_allele'] = spl[1][1].upper()
data_this_snp['segregating'] = spl[2].upper(),spl[allele2_index].upper()
calls_dict = {}
for ii,pop in enumerate(pops):
calls_dict[pop] = int(spl[3+ii]), int(spl[allele2_index+1+ii])
data_this_snp['calls'] = calls_dict
# We name our SNPs using the final columns
snp_id = '_'.join(spl[allele2_index+1+len(pops):])
if snp_id == '':
snp_id = 'SNP_{0}'.format(SNP_ii)
data_dict[snp_id] = data_this_snp
return data_dict
def count_data_dict(data_dict, pop_ids):
"""
Summarize data in data_dict by mapping SNP configurations to counts.
data_dict: data_dict formatted as in Misc.make_data_dict
pop_ids: IDs of populations to collect data for.
Returns a dictionary with keys (successful_calls, derived_calls,
polarized) mapping to counts of SNPs. Here successful_calls is a tuple
with the number of good calls per population, derived_calls is a tuple
of derived calls per pop, and polarized indicates whether that SNP was
polarized using an ancestral state.
"""
count_dict = collections.defaultdict(int)
for snp_info in data_dict.values():
# Skip SNPs that aren't biallelic.
if len(snp_info['segregating']) != 2:
continue
allele1,allele2 = snp_info['segregating']
if 'outgroup_allele' in snp_info and snp_info['outgroup_allele'] != '-'\
and snp_info['outgroup_allele'] in snp_info['segregating']:
outgroup_allele = snp_info['outgroup_allele']
this_snp_polarized = True
else:
outgroup_allele = allele1
this_snp_polarized = False
# Extract the allele calls for each population.
allele1_calls = [snp_info['calls'][pop][0] for pop in pop_ids]
allele2_calls = [snp_info['calls'][pop][1] for pop in pop_ids]
# How many chromosomes did we call successfully in each population?
successful_calls = [a1+a2 for (a1,a2)
in zip(allele1_calls, allele2_calls)]
# Which allele is derived (different from outgroup)?
if allele1 == outgroup_allele:
derived_calls = allele2_calls
elif allele2 == outgroup_allele:
derived_calls = allele1_calls
# Update count_dict
count_dict[tuple(successful_calls),tuple(derived_calls),
this_snp_polarized] += 1
return count_dict
def dd_from_SLiM_files(fnames, mut_types=None, chr='SLIM_'):
"""
Create a data dictionary from a sequence of SLiM output files.
Returns the data dictionary and a sequence of sample sizes.
It is assumed that each file corresponds to samples from a different
population. For example, in SLiM:
p1.outputSample(10, filePath='p1.slimout');
p2.outputSample(10, filePath='p2.slimout');
The populations will be named 0,1,2,... corresponding
to their order in fnames.
fnames: Filenames to parse
mut_types: Sequence of mutation types to include. If None, all mutations
will be included.
chr: Prefix to be used for indicating mutation locations.
The keys in the resulting dictionary are of the form
<chr>_<position>.<globalid>
"""
# Open all the files
try:
fids = [open(_) for _ in fnames]
except TypeError:
fids = fnames
# For each population, we'll first map the mutation ids used in the file to
# the simulation-level global mutation ids
mut_dicts = [{} for _ in fids]
loc_dict = {}
for fid, mut_dict in zip(fids, mut_dicts):
fid.readline(); fid.readline()
line = fid.readline()
while not line.startswith('Genomes:'):
local_id, global_id, mut_type, location, _ = line.split(None, 4)
if mut_types is None or mut_type in mut_types:
mut_dict[local_id] = global_id
loc_dict[global_id] = location
line = fid.readline()
# Now for each population we count each mutation
mut_counts = [collections.defaultdict(int) for _ in fids]
# We also use this pass to measure the sample size for each pop.
sample_sizes = [0 for _ in fids]
for pop_ii, (fid, mut_count) in enumerate(zip(fids, mut_counts)):
# For each genome
for line in fid:
sample_sizes[pop_ii] += 1
mutations = line.split()[2:]
for m in mutations:
mut_count[m] += 1
# Now collect all mutations from all pops.
all_muts = set()
for mut_dict in mut_dicts:
all_muts.update(mut_dict.values())
# Create the empty data dictionary
dd = {}
for global_id in all_muts:
key = 'SLiM_'+loc_dict[global_id] + '.' + global_id
dd[key] = {'segregating': [0,1],
'outgroup_allele': 0,
'calls': {}}
for pop_id, n in enumerate(sample_sizes):
# We initialize each entry to indicate that the allele is not
# segregating in the population, since in this case it's
# not reported in that population's file.
dd[key]['calls'][pop_id] = [n, 0]
# Now update the alleles that are segregating in each population.
for pop_ii, (mut_count, mut_dict)\
in enumerate(zip(mut_counts, mut_dicts)):
for local_id, count in mut_count.items():
try:
# Lookup in mut_dict will fail if mutation isn't of appropriate
# type to include in this run.
global_id = mut_dict[local_id]
key = 'SLiM_'+loc_dict[global_id] + '.' + global_id
dd[key]['calls'][pop_ii] = (sample_sizes[pop_ii]-count, count)
except KeyError:
pass
return dd, sample_sizes
def make_data_dict_vcf(vcf_filename, popinfo_filename, subsample=None, filter=True,
flanking_info=[None, None]):
"""
Parse a VCF file containing genomic sequence information, along with a file
identifying the population of each sample, and store the information in
a properly formatted dictionary.
Each file may be zipped (.zip) or gzipped (.gz). If a file is zipped,
it must be the only file in the archive, and the two files cannot be zipped
together. Both files must be present for the function to work.
vcf_filename : Name of VCF file to work with. The function currently works
for biallelic SNPs only, so if REF or ALT is anything other
than a single base pair (A, C, T, or G), the allele will be
skipped. Additionally, genotype information must be present
in the FORMAT field GT, and genotype info must be known for
every sample, else the SNP will be skipped. If the ancestral
allele is known it should be specified in INFO field 'AA'.
Otherwise, it will be set to '-'.
popinfo_filename : Name of file containing the population assignments for
each sample in the VCF. If a sample in the VCF file does
not have a corresponding entry in this file, it will be
skipped. See _get_popinfo for information on how this
file must be formatted.
subsample : Dictionary with population names used in the popinfo_filename
as keys and the desired sample size (in number of individuals)
for subsampling as values. E.g., {"pop1": n1, "pop2": n2} for
two populations.
filter : If set to True, alleles will be skipped if they have not passed
all filters (i.e. either 'PASS' or '.' must be present in FILTER
column.
flanking_info : Flanking information for the reference and/or ancestral
allele can be provided as field(s) in the INFO column. To
add this information to the dict, flanking_info should
specify the names of the fields that contain this info as a
list (e.g. ['RFL', 'AFL'].) If context info is given for
only one allele, set the other item in the list to None,
(e.g. ['RFL', None]). Information can be provided as a 3
base-pair sequence or 2 base-pair sequence, where the first
base-pair is the one immediately preceding the SNP, and the
last base-pair is the one immediately following the SNP.
"""
do_subsampling = False
if subsample is not None:
do_subsampling = True
warnings.warn('Note on subsampling: If you will be including inbreeding in your model, '
'do not project your data to smaller sample sizes in later steps of your analysis.')
if os.path.splitext(popinfo_filename)[1] == '.gz':
import gzip
popinfo_file = gzip.open(popinfo_filename)
elif os.path.splitext(popinfo_filename)[1] == '.zip':
import zipfile
archive = zipfile.ZipFile(popinfo_filename)
namelist = archive.namelist()
if len(namelist) != 1:
raise ValueError("Must be only a single popinfo file in zip "
"archive: {}".format(popinfo_filename))
popinfo_file = archive.open(namelist[0])
else:
popinfo_file = open(popinfo_filename)
# pop_dict has key, value pairs of "SAMPLE_NAME" : "POP_NAME"
try:
popinfo_dict = _get_popinfo(popinfo_file)
except:
raise ValueError('Failed in parsing popinfo file.')
popinfo_file.close()
# Open VCF file
if os.path.splitext(vcf_filename)[1] == '.gz':
import gzip
vcf_file = gzip.open(vcf_filename)
elif os.path.splitext(vcf_filename)[1] == '.zip':
import zipfile
archive = zipfile.ZipFile(vcf_filename)
namelist = archive.namelist()
if len(namelist) != 1:
raise ValueError("Must be only a single vcf file in zip "
"archive: {}".format(vcf_filename))
vcf_file = archive.open(namelist[0])
else:
vcf_file = open(vcf_filename)
data_dict = {}
for line in vcf_file:
# decoding lines for Python 3 - probably a better way to handle this
try:
line = line.decode()
except AttributeError:
pass
# Skip metainformation
if line.startswith('##'):
continue
# Read header
if line.startswith('#'):
header_cols = line.split()
# Ensure there is at least one sample
if len(header_cols) <= 9:
raise ValueError("No samples in VCF file")
# Use popinfo_dict to get the order of populations present in VCF
poplist = [popinfo_dict[sample] if sample in popinfo_dict else None
for sample in header_cols[9:]]
continue
# Read SNP data
# Data lines in VCF file are tab-delimited
# See https://samtools.github.io/hts-specs/VCFv4.2.pdf
cols = line.split("\t")
snp_id = '_'.join(cols[:2]) # CHROM_POS
snp_dict = {}
# Skip SNP if filter is set to True and it fails a filter test
if filter and cols[6] != 'PASS' and cols[6] != '.':
continue
# Add reference and alternate allele info to dict
ref, alt = (allele.upper() for allele in cols[3:5])
if ref not in ['A', 'C', 'G', 'T'] or alt not in ['A', 'C', 'G', 'T']:
# Skip line if site is not an SNP
continue
snp_dict['segregating'] = (ref, alt)
snp_dict['context'] = '-' + ref + '-'
# Add ancestral allele information if available
info = cols[7].split(';')
for field in info:
if field.startswith('AA=') or field.startswith('AA_ensembl=') or field.startswith('AA_chimp='):
outgroup_allele = field.split('=')[1].upper()
if outgroup_allele not in ['A','C','G','T']:
# Skip if ancestral not single base A, C, G, or T
outgroup_allele = '-'
break
else:
outgroup_allele = '-'
snp_dict['outgroup_allele'] = outgroup_allele
snp_dict['outgroup_context'] = '-' + outgroup_allele + '-'
# Add flanking info if it is present
rflank, aflank = flanking_info
for field in info:
if rflank and field.startswith(rflank):
flank = field[len(rflank)+1:].upper()
if not (len(flank) == 2 or len(flank) == 3):
continue
prevb, nextb = flank[0], flank[-1]
if prevb not in ['A','C','T','G']:
prevb = '-'
if nextb not in ['A','C','T','G']:
nextb = '-'
snp_dict['context'] = prevb + ref + nextb
continue
if aflank and field.startswith(aflank):
flank = field[len(aflank)+1:].upper()
if not (len(flank) == 2 or len(flank) == 3):
continue
prevb, nextb = flank[0], flank[-1]
if prevb not in ['A','C','T','G']:
prevb = '-'
if nextb not in ['A','C','T','G']:
nextb = '-'
snp_dict['outgroup_context'] = prevb + outgroup_allele + nextb
calls_dict = {}
subsample_dict = {}
gtindex = cols[8].split(':').index('GT')
if do_subsampling:
# Collect data for all genotyped samples
for pop, sample in zip(poplist, cols[9:]):
if pop is None:
continue
gt = sample.split(':')[gtindex]
if pop not in subsample_dict:
subsample_dict[pop] = []
if '.' not in gt:
subsample_dict[pop].append(gt)
# key-value pairs here are population names
# and a list of genotypes to subsample from
for pop, genotypes in subsample_dict.items():
if pop not in calls_dict:
calls_dict[pop] = (0, 0)
if len(genotypes) < subsample[pop]:
# Not enough calls for this SNP
break
# Choose which individuals to use
idx = numpy.random.choice([i for i in range(0,len(genotypes))], subsample[pop], replace=False)
for ii in idx:
gt = subsample_dict[pop][ii]
refcalls, altcalls = calls_dict[pop]
refcalls += gt[::2].count('0')
altcalls += gt[::2].count('1')
calls_dict[pop] = (refcalls, altcalls)
else:
# Only runs if we didn't break out of this loop
snp_dict['calls'] = calls_dict
data_dict[snp_id] = snp_dict
else:
for pop, sample in zip(poplist, cols[9:]):
if pop is None:
continue
if pop not in calls_dict:
calls_dict[pop] = (0,0)
# Genotype in VCF format 0|1|1|0:...
gt = sample.split(':')[gtindex]
#g1, g2 = gt[0], gt[2]
#if g1 == '.' or g2 == '.':
# continue
#full_info = False
#break
refcalls, altcalls = calls_dict[pop]
#refcalls += int(g1 == '0') + int(g2 == '0')
#altcalls += int(g1 == '1') + int(g2 == '1')
# Assume biallelic variants
refcalls += gt[::2].count('0')
altcalls += gt[::2].count('1')
calls_dict[pop] = (refcalls, altcalls)
snp_dict['calls'] = calls_dict
data_dict[snp_id] = snp_dict
vcf_file.close()
return data_dict
def _get_popinfo(popinfo_file):
"""
Helper function for make_data_dict_vcf. Takes an open file that contains
information on the population designations of each sample within a VCF file,
and returns a dictionary containing {"SAMPLE_NAME" : "POP_NAME"} pairs.
The file should be formatted as a table, with columns delimited by
whitespace, and rows delimited by new lines. Lines beginning with '#' are
considered comments and will be ignored. Each sample must appear on its own
line. If no header information is provided, the first column will be assumed
to be the SAMPLE_NAME column, while the second column will be assumed to be
the POP_NAME column. If a header is present, it must be the first
non-comment line of the file. The column positions of the words "SAMPLE" and
"POP" (ignoring case) in this header will be used to determine proper
positions of the SAMPLE_NAME and POP_NAME columns in the table.
popinfo_file : An open text file of the format described above.
"""
popinfo_dict = {}
sample_col = 0
pop_col = 1
header = False
# check for header info
for line in popinfo_file:
if line.startswith('#'):
continue
cols = [col.lower() for col in line.split()]
if 'sample' in cols:
header = True
sample_col = cols.index('sample')
if 'pop' in cols:
header = True
pop_col = cols.index('pop')
break
# read in population information for each sample
popinfo_file.seek(0)
for line in popinfo_file:
if line.startswith('#') or not line.strip():
continue
cols = line.split()
sample = cols[sample_col]
pop = cols[pop_col]
# avoid adding header to dict
if (sample.lower() == 'sample' or pop.lower() == 'pop') and header:
header = False
continue
popinfo_dict[sample] = pop
return popinfo_dict
def annotate_from_annovar(dd, annovar_file, variant_type):
"""
Return a data dictionary with only the sites of a requested type of variation based on an ANNOVAR '.exonic_variant_function' output file.
dd: Data dictionary of sites of a requested type of annotation
annovar_file: Output file from ANNOVAR with the '.exonic_variant_function' extension
variant_type: The type of variant you want to make a data dictionary sites to contain
"""
anno_list = []
var_fid = open(annovar_file)
for line in var_fid:
variant = line.split('\t')[1]
position = '_'.join(line.split()[4:6])
if variant_type in variant:
anno_list.append(position)
var_fid.close()
dd_anno = {}
for key in dd:
if key in anno_list:
dd_anno[key] = dd[key]
return dd_anno
def combine_pops(fs, idx=[0,1]):
"""
Combine the frequency spectra of two populations.
fs: Spectrum object (2D or 3D).
idx: Indices for populations being collapsed. (defaul=[0,1])
The function will always return the combined populations along the
first axis of the sfs. The resulting spectrum is also returned
as a numpy array, but can be converted to a Spectrum object
using the dadi.Spectrum() function.
"""
ns = fs.sample_sizes
if len(ns) == 3:
fs_tmp = numpy.array(fs)
if idx == [0,1]:
fs2 = numpy.zeros((ns[0]+ns[1]+1,ns[2]+1))
for ii in range(ns[0]+1):
for jj in range(ns[1]+1):
for kk in range(ns[2]+1):
fs2[ii+jj,kk] += fs_tmp[ii,jj,kk]
elif idx == [0,2]:
fs2 = numpy.zeros((ns[0]+ns[2]+1,ns[1]+1))
for ii in range(ns[0]+1):
for jj in range(ns[2]+1):
for kk in range(ns[1]+1):
fs2[ii+jj,kk] += fs_tmp[ii,kk,jj]
elif idx == [1,2]:
fs2 = numpy.zeros((ns[1]+ns[2]+1,ns[0]+1))
for ii in range(ns[1]+1):
for jj in range(ns[2]+1):
for kk in range(ns[0]+1):
fs2[ii+jj,kk] += fs_tmp[kk,ii,jj]
else:
print("Error: did not recognize population indices: {}".format(idx))
exit(-1)
elif len(ns) == 2:
fs_tmp = numpy.array(fs)
fs2 = numpy.zeros((ns[0]+ns[1]+1,))
for ii in range(ns[0]+1):
for jj in range(ns[1]+1):
fs2[ii+jj] += fs_tmp[ii,jj]
else:
print("Error: could not combine populations.")
exit(-1)
fs2 = Spectrum(fs2)
fs2.extrap_x = fs.extrap_x
return fs2
def fragment_data_dict(dd, chunk_size):
"""
Split data dictionary for bootstrapping.
For bootstrapping, split the data dictionary in subdictionaries
for each chunk of the genome. The chunk_size is given in
basepairs.
This method assumes that keys in the dictionary are
chromosome_position[.additional_info]
The [.additional_info] is optional, and can be used to distinguish
recurrent mutations at the same site.
dd: Data dictionary to split
chunk_size: Size of genomic chunks in basepairs
Return: List of dictionaries corresponding to each chunk.
"""
# split dictionary by chromosome name
ndd = collections.defaultdict(list)
for k in dd.keys():
spl = k.split('.')[0].split('_')
chrname, position = '_'.join(k.split('_')[:-1]), k.split('_')[-1]
# Track additional_info
if not '.' in position:
add_info = None
else:
position, add_info = position.split('.',1)
ndd[chrname].append((int(position), add_info))
# generate chunks with given chunk size
chunks_dict = collections.defaultdict(list)
for chrname in ndd.keys():
positions = sorted(ndd[chrname])
end = chunk_size
chunk_index = 0
chunks_dict[chrname].append([])
for p, add_info in positions:
while p > end:
# Need a new chunk
end += chunk_size
chunk_index += 1
chunks_dict[chrname].append([])
chunks_dict[chrname][chunk_index].append((p, add_info))
# Break data dictionary into dictionaries for each chunk
new_dds = []
for chrname, chunks in chunks_dict.items():
for pos_list in chunks:
new_dds.append({})
for pos, add_info in pos_list:
if not add_info:
key = '{0}_{1}'.format(chrname,pos)
else:
key = '{0}_{1}.{2}'.format(chrname,pos,add_info)
new_dds[-1][key] = dd[key]
return new_dds
import numpy as np
import random
from .Spectrum_mod import Spectrum
import functools, operator
def bootstraps_from_dd_chunks(fragments, Nboot, pop_ids, projections, mask_corners=True, polarized=True):
"""
Bootstrap frequency spectra from data dictionary fragments
fragments: Fragmented data dictionary
Nboot: Number of bootstrap spectra to generate
Remaining arguments are as in Spectrum.from_data_dict
"""
spectra = [Spectrum.from_data_dict(dd, pop_ids, projections, mask_corners, polarized)
for dd in fragments]
bootstraps = []
for ii in range(Nboot):
chosen = random.choices(spectra, k=len(spectra))
bootstraps.append(functools.reduce(operator.add, chosen))
bootstraps = [Spectrum(_, mask_corners=mask_corners, data_folded=not polarized, pop_ids=pop_ids)
for _ in bootstraps]
return bootstraps
def bootstraps_subsample_vcf(vcf_filename, popinfo_filename, subsample, Nboot, chunk_size, pop_ids, filter=True,
flanking_info=[None, None], mask_corners=True, polarized=True):
"""
Bootstrap frequency spectra from subsampling in VCF file
This method is useful when you have subsampled individuals. You might do this
because you're modeling inbreeding, or because you want to avoid composite likelihood
complications (if your data is unlinked).
Nboot: Number of boostrap spectra to generate
chunk_size: Size of regions to divide genome into (in basepairs)
Other arguments are as in make_data_dict_vcf and Spectrum.from_data_dict.
"""
bootstraps = []
projections = [subsample[pop]*2 for pop in pop_ids]
for ii in range(Nboot):
dd = make_data_dict_vcf(vcf_filename, popinfo_filename, subsample=subsample, filter=filter,
flanking_info=flanking_info)
fragments = fragment_data_dict(dd, chunk_size)
fs = bootstraps_from_dd_chunks(fragments, 1, pop_ids, projections, mask_corners, polarized)[0]
bootstraps.append(fs)
return bootstrapsFunctions
def annotate_from_annovar(dd, annovar_file, variant_type)-
Return a data dictionary with only the sites of a requested type of variation based on an ANNOVAR '.exonic_variant_function' output file.
dd: Data dictionary of sites of a requested type of annotation annovar_file: Output file from ANNOVAR with the '.exonic_variant_function' extension variant_type: The type of variant you want to make a data dictionary sites to contain
Expand source code
def annotate_from_annovar(dd, annovar_file, variant_type): """ Return a data dictionary with only the sites of a requested type of variation based on an ANNOVAR '.exonic_variant_function' output file. dd: Data dictionary of sites of a requested type of annotation annovar_file: Output file from ANNOVAR with the '.exonic_variant_function' extension variant_type: The type of variant you want to make a data dictionary sites to contain """ anno_list = [] var_fid = open(annovar_file) for line in var_fid: variant = line.split('\t')[1] position = '_'.join(line.split()[4:6]) if variant_type in variant: anno_list.append(position) var_fid.close() dd_anno = {} for key in dd: if key in anno_list: dd_anno[key] = dd[key] return dd_anno def bootstraps_from_dd_chunks(fragments, Nboot, pop_ids, projections, mask_corners=True, polarized=True)-
Bootstrap frequency spectra from data dictionary fragments
fragments: Fragmented data dictionary Nboot: Number of bootstrap spectra to generate Remaining arguments are as in Spectrum.from_data_dict
Expand source code
def bootstraps_from_dd_chunks(fragments, Nboot, pop_ids, projections, mask_corners=True, polarized=True): """ Bootstrap frequency spectra from data dictionary fragments fragments: Fragmented data dictionary Nboot: Number of bootstrap spectra to generate Remaining arguments are as in Spectrum.from_data_dict """ spectra = [Spectrum.from_data_dict(dd, pop_ids, projections, mask_corners, polarized) for dd in fragments] bootstraps = [] for ii in range(Nboot): chosen = random.choices(spectra, k=len(spectra)) bootstraps.append(functools.reduce(operator.add, chosen)) bootstraps = [Spectrum(_, mask_corners=mask_corners, data_folded=not polarized, pop_ids=pop_ids) for _ in bootstraps] return bootstraps def bootstraps_subsample_vcf(vcf_filename, popinfo_filename, subsample, Nboot, chunk_size, pop_ids, filter=True, flanking_info=[None, None], mask_corners=True, polarized=True)-
Bootstrap frequency spectra from subsampling in VCF file
This method is useful when you have subsampled individuals. You might do this because you're modeling inbreeding, or because you want to avoid composite likelihood complications (if your data is unlinked).
Nboot: Number of boostrap spectra to generate chunk_size: Size of regions to divide genome into (in basepairs) Other arguments are as in make_data_dict_vcf and Spectrum.from_data_dict.
Expand source code
def bootstraps_subsample_vcf(vcf_filename, popinfo_filename, subsample, Nboot, chunk_size, pop_ids, filter=True, flanking_info=[None, None], mask_corners=True, polarized=True): """ Bootstrap frequency spectra from subsampling in VCF file This method is useful when you have subsampled individuals. You might do this because you're modeling inbreeding, or because you want to avoid composite likelihood complications (if your data is unlinked). Nboot: Number of boostrap spectra to generate chunk_size: Size of regions to divide genome into (in basepairs) Other arguments are as in make_data_dict_vcf and Spectrum.from_data_dict. """ bootstraps = [] projections = [subsample[pop]*2 for pop in pop_ids] for ii in range(Nboot): dd = make_data_dict_vcf(vcf_filename, popinfo_filename, subsample=subsample, filter=filter, flanking_info=flanking_info) fragments = fragment_data_dict(dd, chunk_size) fs = bootstraps_from_dd_chunks(fragments, 1, pop_ids, projections, mask_corners, polarized)[0] bootstraps.append(fs) return bootstraps def combine_pops(fs, idx=[0, 1])-
Combine the frequency spectra of two populations. fs: Spectrum object (2D or 3D). idx: Indices for populations being collapsed. (defaul=[0,1])
The function will always return the combined populations along the first axis of the sfs. The resulting spectrum is also returned as a numpy array, but can be converted to a Spectrum object using the dadi.Spectrum() function.
Expand source code
def combine_pops(fs, idx=[0,1]): """ Combine the frequency spectra of two populations. fs: Spectrum object (2D or 3D). idx: Indices for populations being collapsed. (defaul=[0,1]) The function will always return the combined populations along the first axis of the sfs. The resulting spectrum is also returned as a numpy array, but can be converted to a Spectrum object using the dadi.Spectrum() function. """ ns = fs.sample_sizes if len(ns) == 3: fs_tmp = numpy.array(fs) if idx == [0,1]: fs2 = numpy.zeros((ns[0]+ns[1]+1,ns[2]+1)) for ii in range(ns[0]+1): for jj in range(ns[1]+1): for kk in range(ns[2]+1): fs2[ii+jj,kk] += fs_tmp[ii,jj,kk] elif idx == [0,2]: fs2 = numpy.zeros((ns[0]+ns[2]+1,ns[1]+1)) for ii in range(ns[0]+1): for jj in range(ns[2]+1): for kk in range(ns[1]+1): fs2[ii+jj,kk] += fs_tmp[ii,kk,jj] elif idx == [1,2]: fs2 = numpy.zeros((ns[1]+ns[2]+1,ns[0]+1)) for ii in range(ns[1]+1): for jj in range(ns[2]+1): for kk in range(ns[0]+1): fs2[ii+jj,kk] += fs_tmp[kk,ii,jj] else: print("Error: did not recognize population indices: {}".format(idx)) exit(-1) elif len(ns) == 2: fs_tmp = numpy.array(fs) fs2 = numpy.zeros((ns[0]+ns[1]+1,)) for ii in range(ns[0]+1): for jj in range(ns[1]+1): fs2[ii+jj] += fs_tmp[ii,jj] else: print("Error: could not combine populations.") exit(-1) fs2 = Spectrum(fs2) fs2.extrap_x = fs.extrap_x return fs2 def count_data_dict(data_dict, pop_ids)-
Summarize data in data_dict by mapping SNP configurations to counts.
data_dict: data_dict formatted as in Misc.make_data_dict pop_ids: IDs of populations to collect data for.
Returns a dictionary with keys (successful_calls, derived_calls, polarized) mapping to counts of SNPs. Here successful_calls is a tuple with the number of good calls per population, derived_calls is a tuple of derived calls per pop, and polarized indicates whether that SNP was polarized using an ancestral state.
Expand source code
def count_data_dict(data_dict, pop_ids): """ Summarize data in data_dict by mapping SNP configurations to counts. data_dict: data_dict formatted as in Misc.make_data_dict pop_ids: IDs of populations to collect data for. Returns a dictionary with keys (successful_calls, derived_calls, polarized) mapping to counts of SNPs. Here successful_calls is a tuple with the number of good calls per population, derived_calls is a tuple of derived calls per pop, and polarized indicates whether that SNP was polarized using an ancestral state. """ count_dict = collections.defaultdict(int) for snp_info in data_dict.values(): # Skip SNPs that aren't biallelic. if len(snp_info['segregating']) != 2: continue allele1,allele2 = snp_info['segregating'] if 'outgroup_allele' in snp_info and snp_info['outgroup_allele'] != '-'\ and snp_info['outgroup_allele'] in snp_info['segregating']: outgroup_allele = snp_info['outgroup_allele'] this_snp_polarized = True else: outgroup_allele = allele1 this_snp_polarized = False # Extract the allele calls for each population. allele1_calls = [snp_info['calls'][pop][0] for pop in pop_ids] allele2_calls = [snp_info['calls'][pop][1] for pop in pop_ids] # How many chromosomes did we call successfully in each population? successful_calls = [a1+a2 for (a1,a2) in zip(allele1_calls, allele2_calls)] # Which allele is derived (different from outgroup)? if allele1 == outgroup_allele: derived_calls = allele2_calls elif allele2 == outgroup_allele: derived_calls = allele1_calls # Update count_dict count_dict[tuple(successful_calls),tuple(derived_calls), this_snp_polarized] += 1 return count_dict def dd_from_SLiM_files(fnames, mut_types=None, chr='SLIM_')-
Create a data dictionary from a sequence of SLiM output files.
Returns the data dictionary and a sequence of sample sizes.
It is assumed that each file corresponds to samples from a different population. For example, in SLiM: p1.outputSample(10, filePath='p1.slimout'); p2.outputSample(10, filePath='p2.slimout');
The populations will be named 0,1,2,… corresponding to their order in fnames.
fnames: Filenames to parse mut_types: Sequence of mutation types to include. If None, all mutations will be included. chr: Prefix to be used for indicating mutation locations.
The keys in the resulting dictionary are of the form
_ . Expand source code
def dd_from_SLiM_files(fnames, mut_types=None, chr='SLIM_'): """ Create a data dictionary from a sequence of SLiM output files. Returns the data dictionary and a sequence of sample sizes. It is assumed that each file corresponds to samples from a different population. For example, in SLiM: p1.outputSample(10, filePath='p1.slimout'); p2.outputSample(10, filePath='p2.slimout'); The populations will be named 0,1,2,... corresponding to their order in fnames. fnames: Filenames to parse mut_types: Sequence of mutation types to include. If None, all mutations will be included. chr: Prefix to be used for indicating mutation locations. The keys in the resulting dictionary are of the form <chr>_<position>.<globalid> """ # Open all the files try: fids = [open(_) for _ in fnames] except TypeError: fids = fnames # For each population, we'll first map the mutation ids used in the file to # the simulation-level global mutation ids mut_dicts = [{} for _ in fids] loc_dict = {} for fid, mut_dict in zip(fids, mut_dicts): fid.readline(); fid.readline() line = fid.readline() while not line.startswith('Genomes:'): local_id, global_id, mut_type, location, _ = line.split(None, 4) if mut_types is None or mut_type in mut_types: mut_dict[local_id] = global_id loc_dict[global_id] = location line = fid.readline() # Now for each population we count each mutation mut_counts = [collections.defaultdict(int) for _ in fids] # We also use this pass to measure the sample size for each pop. sample_sizes = [0 for _ in fids] for pop_ii, (fid, mut_count) in enumerate(zip(fids, mut_counts)): # For each genome for line in fid: sample_sizes[pop_ii] += 1 mutations = line.split()[2:] for m in mutations: mut_count[m] += 1 # Now collect all mutations from all pops. all_muts = set() for mut_dict in mut_dicts: all_muts.update(mut_dict.values()) # Create the empty data dictionary dd = {} for global_id in all_muts: key = 'SLiM_'+loc_dict[global_id] + '.' + global_id dd[key] = {'segregating': [0,1], 'outgroup_allele': 0, 'calls': {}} for pop_id, n in enumerate(sample_sizes): # We initialize each entry to indicate that the allele is not # segregating in the population, since in this case it's # not reported in that population's file. dd[key]['calls'][pop_id] = [n, 0] # Now update the alleles that are segregating in each population. for pop_ii, (mut_count, mut_dict)\ in enumerate(zip(mut_counts, mut_dicts)): for local_id, count in mut_count.items(): try: # Lookup in mut_dict will fail if mutation isn't of appropriate # type to include in this run. global_id = mut_dict[local_id] key = 'SLiM_'+loc_dict[global_id] + '.' + global_id dd[key]['calls'][pop_ii] = (sample_sizes[pop_ii]-count, count) except KeyError: pass return dd, sample_sizes def delayed_flush(stream=sys.stdout, delay=1)-
Flush a stream, ensuring that it is only flushed every 'delay' minutes. Note that upon the first call to this method, the stream is not flushed.
stream: The stream to flush. For this to work with simple 'print' statements, the stream should be sys.stdout. delay: Minimum time in minutes between flushes.
This function is useful to prevent I/O overload on the cluster.
Expand source code
def delayed_flush(stream=sys.stdout, delay=1): """ Flush a stream, ensuring that it is only flushed every 'delay' *minutes*. Note that upon the first call to this method, the stream is not flushed. stream: The stream to flush. For this to work with simple 'print' statements, the stream should be sys.stdout. delay: Minimum time *in minutes* between flushes. This function is useful to prevent I/O overload on the cluster. """ global __times_last_flushed curr_time = time.time() # If this is the first time this method has been called with this stream, # we need to fill in the times_last_flushed dict. setdefault will do this # without overwriting any entry that may be there already. if stream not in __times_last_flushed: __times_last_flushed[stream] = curr_time last_flushed = __times_last_flushed[stream] # Note that time.time() returns values in seconds, hence the factor of 60. if (curr_time - last_flushed) >= delay*60: stream.flush() __times_last_flushed[stream] = curr_time def ensure_1arg_func(var)-
Ensure that var is actually a one-argument function.
This is primarily used to convert arguments that are constants into trivial functions of time for use in integrations where parameters are allowed to change over time.
Expand source code
def ensure_1arg_func(var): """ Ensure that var is actually a one-argument function. This is primarily used to convert arguments that are constants into trivial functions of time for use in integrations where parameters are allowed to change over time. """ if numpy.isscalar(var): # If a constant was passed in, use lambda to make it a nice # simple function. var_f_tmp = lambda t: var else: var_f_tmp = var # Wrapping arguments in float64 eases working with CUDA var_f = lambda t: numpy.float64(var_f_tmp(t)) if not callable(var_f): raise ValueError('Argument is not a constant or a function.') try: var_f(0.0) except TypeError: raise ValueError('Argument is not a constant or a one-argument ' 'function.') return var_f def fragment_data_dict(dd, chunk_size)-
Split data dictionary for bootstrapping.
For bootstrapping, split the data dictionary in subdictionaries for each chunk of the genome. The chunk_size is given in basepairs.
This method assumes that keys in the dictionary are chromosome_position[.additional_info] The [.additional_info] is optional, and can be used to distinguish recurrent mutations at the same site.
dd: Data dictionary to split chunk_size: Size of genomic chunks in basepairs
Return: List of dictionaries corresponding to each chunk.
Expand source code
def fragment_data_dict(dd, chunk_size): """ Split data dictionary for bootstrapping. For bootstrapping, split the data dictionary in subdictionaries for each chunk of the genome. The chunk_size is given in basepairs. This method assumes that keys in the dictionary are chromosome_position[.additional_info] The [.additional_info] is optional, and can be used to distinguish recurrent mutations at the same site. dd: Data dictionary to split chunk_size: Size of genomic chunks in basepairs Return: List of dictionaries corresponding to each chunk. """ # split dictionary by chromosome name ndd = collections.defaultdict(list) for k in dd.keys(): spl = k.split('.')[0].split('_') chrname, position = '_'.join(k.split('_')[:-1]), k.split('_')[-1] # Track additional_info if not '.' in position: add_info = None else: position, add_info = position.split('.',1) ndd[chrname].append((int(position), add_info)) # generate chunks with given chunk size chunks_dict = collections.defaultdict(list) for chrname in ndd.keys(): positions = sorted(ndd[chrname]) end = chunk_size chunk_index = 0 chunks_dict[chrname].append([]) for p, add_info in positions: while p > end: # Need a new chunk end += chunk_size chunk_index += 1 chunks_dict[chrname].append([]) chunks_dict[chrname][chunk_index].append((p, add_info)) # Break data dictionary into dictionaries for each chunk new_dds = [] for chrname, chunks in chunks_dict.items(): for pos_list in chunks: new_dds.append({}) for pos, add_info in pos_list: if not add_info: key = '{0}_{1}'.format(chrname,pos) else: key = '{0}_{1}.{2}'.format(chrname,pos,add_info) new_dds[-1][key] = dd[key] return new_dds def make_data_dict(filename)-
Parse SNP file and store info in a properly formatted dictionary.
filename: Name of file to work with.
This is specific to the particular data format described on the wiki. Modification for other formats should be straightforward.
The file can be zipped (extension .zip) or gzipped (extension .gz). If zipped, there must be only a single file in the zip archive.
Expand source code
def make_data_dict(filename): """ Parse SNP file and store info in a properly formatted dictionary. filename: Name of file to work with. This is specific to the particular data format described on the wiki. Modification for other formats should be straightforward. The file can be zipped (extension .zip) or gzipped (extension .gz). If zipped, there must be only a single file in the zip archive. """ if os.path.splitext(filename)[1] == '.gz': import gzip f = gzip.open(filename) elif os.path.splitext(filename)[1] == '.zip': import zipfile archive = zipfile.ZipFile(filename) namelist = archive.namelist() if len(namelist) != 1: raise ValueError('Must be only a single data file in zip ' 'archive: %s' % filename) f = archive.open(namelist[0]) else: f = open(filename) # Skip to the header while True: header = f.readline() if not header.startswith('#'): break allele2_index = header.split().index('Allele2') # Pull out our pop ids pops = header.split()[3:allele2_index] # The empty data dictionary data_dict = {} # Now walk down the file for SNP_ii, line in enumerate(f): if line.startswith('#'): continue # Split the into fields by whitespace spl = line.split() data_this_snp = {} # We convert to upper case to avoid any issues with mixed case between # SNPs. data_this_snp['context'] = spl[0].upper() data_this_snp['outgroup_context'] = spl[1].upper() data_this_snp['outgroup_allele'] = spl[1][1].upper() data_this_snp['segregating'] = spl[2].upper(),spl[allele2_index].upper() calls_dict = {} for ii,pop in enumerate(pops): calls_dict[pop] = int(spl[3+ii]), int(spl[allele2_index+1+ii]) data_this_snp['calls'] = calls_dict # We name our SNPs using the final columns snp_id = '_'.join(spl[allele2_index+1+len(pops):]) if snp_id == '': snp_id = 'SNP_{0}'.format(SNP_ii) data_dict[snp_id] = data_this_snp return data_dict def make_data_dict_vcf(vcf_filename, popinfo_filename, subsample=None, filter=True, flanking_info=[None, None])-
Parse a VCF file containing genomic sequence information, along with a file identifying the population of each sample, and store the information in a properly formatted dictionary.
Each file may be zipped (.zip) or gzipped (.gz). If a file is zipped, it must be the only file in the archive, and the two files cannot be zipped together. Both files must be present for the function to work.
vcf_filename : Name of VCF file to work with. The function currently works for biallelic SNPs only, so if REF or ALT is anything other than a single base pair (A, C, T, or G), the allele will be skipped. Additionally, genotype information must be present in the FORMAT field GT, and genotype info must be known for every sample, else the SNP will be skipped. If the ancestral allele is known it should be specified in INFO field 'AA'. Otherwise, it will be set to '-'.
popinfo_filename : Name of file containing the population assignments for each sample in the VCF. If a sample in the VCF file does not have a corresponding entry in this file, it will be skipped. See _get_popinfo for information on how this file must be formatted.
subsample : Dictionary with population names used in the popinfo_filename as keys and the desired sample size (in number of individuals) for subsampling as values. E.g., {"pop1": n1, "pop2": n2} for two populations.
filter : If set to True, alleles will be skipped if they have not passed all filters (i.e. either 'PASS' or '.' must be present in FILTER column.
flanking_info : Flanking information for the reference and/or ancestral allele can be provided as field(s) in the INFO column. To add this information to the dict, flanking_info should specify the names of the fields that contain this info as a list (e.g. ['RFL', 'AFL'].) If context info is given for only one allele, set the other item in the list to None, (e.g. ['RFL', None]). Information can be provided as a 3 base-pair sequence or 2 base-pair sequence, where the first base-pair is the one immediately preceding the SNP, and the last base-pair is the one immediately following the SNP.
Expand source code
def make_data_dict_vcf(vcf_filename, popinfo_filename, subsample=None, filter=True, flanking_info=[None, None]): """ Parse a VCF file containing genomic sequence information, along with a file identifying the population of each sample, and store the information in a properly formatted dictionary. Each file may be zipped (.zip) or gzipped (.gz). If a file is zipped, it must be the only file in the archive, and the two files cannot be zipped together. Both files must be present for the function to work. vcf_filename : Name of VCF file to work with. The function currently works for biallelic SNPs only, so if REF or ALT is anything other than a single base pair (A, C, T, or G), the allele will be skipped. Additionally, genotype information must be present in the FORMAT field GT, and genotype info must be known for every sample, else the SNP will be skipped. If the ancestral allele is known it should be specified in INFO field 'AA'. Otherwise, it will be set to '-'. popinfo_filename : Name of file containing the population assignments for each sample in the VCF. If a sample in the VCF file does not have a corresponding entry in this file, it will be skipped. See _get_popinfo for information on how this file must be formatted. subsample : Dictionary with population names used in the popinfo_filename as keys and the desired sample size (in number of individuals) for subsampling as values. E.g., {"pop1": n1, "pop2": n2} for two populations. filter : If set to True, alleles will be skipped if they have not passed all filters (i.e. either 'PASS' or '.' must be present in FILTER column. flanking_info : Flanking information for the reference and/or ancestral allele can be provided as field(s) in the INFO column. To add this information to the dict, flanking_info should specify the names of the fields that contain this info as a list (e.g. ['RFL', 'AFL'].) If context info is given for only one allele, set the other item in the list to None, (e.g. ['RFL', None]). Information can be provided as a 3 base-pair sequence or 2 base-pair sequence, where the first base-pair is the one immediately preceding the SNP, and the last base-pair is the one immediately following the SNP. """ do_subsampling = False if subsample is not None: do_subsampling = True warnings.warn('Note on subsampling: If you will be including inbreeding in your model, ' 'do not project your data to smaller sample sizes in later steps of your analysis.') if os.path.splitext(popinfo_filename)[1] == '.gz': import gzip popinfo_file = gzip.open(popinfo_filename) elif os.path.splitext(popinfo_filename)[1] == '.zip': import zipfile archive = zipfile.ZipFile(popinfo_filename) namelist = archive.namelist() if len(namelist) != 1: raise ValueError("Must be only a single popinfo file in zip " "archive: {}".format(popinfo_filename)) popinfo_file = archive.open(namelist[0]) else: popinfo_file = open(popinfo_filename) # pop_dict has key, value pairs of "SAMPLE_NAME" : "POP_NAME" try: popinfo_dict = _get_popinfo(popinfo_file) except: raise ValueError('Failed in parsing popinfo file.') popinfo_file.close() # Open VCF file if os.path.splitext(vcf_filename)[1] == '.gz': import gzip vcf_file = gzip.open(vcf_filename) elif os.path.splitext(vcf_filename)[1] == '.zip': import zipfile archive = zipfile.ZipFile(vcf_filename) namelist = archive.namelist() if len(namelist) != 1: raise ValueError("Must be only a single vcf file in zip " "archive: {}".format(vcf_filename)) vcf_file = archive.open(namelist[0]) else: vcf_file = open(vcf_filename) data_dict = {} for line in vcf_file: # decoding lines for Python 3 - probably a better way to handle this try: line = line.decode() except AttributeError: pass # Skip metainformation if line.startswith('##'): continue # Read header if line.startswith('#'): header_cols = line.split() # Ensure there is at least one sample if len(header_cols) <= 9: raise ValueError("No samples in VCF file") # Use popinfo_dict to get the order of populations present in VCF poplist = [popinfo_dict[sample] if sample in popinfo_dict else None for sample in header_cols[9:]] continue # Read SNP data # Data lines in VCF file are tab-delimited # See https://samtools.github.io/hts-specs/VCFv4.2.pdf cols = line.split("\t") snp_id = '_'.join(cols[:2]) # CHROM_POS snp_dict = {} # Skip SNP if filter is set to True and it fails a filter test if filter and cols[6] != 'PASS' and cols[6] != '.': continue # Add reference and alternate allele info to dict ref, alt = (allele.upper() for allele in cols[3:5]) if ref not in ['A', 'C', 'G', 'T'] or alt not in ['A', 'C', 'G', 'T']: # Skip line if site is not an SNP continue snp_dict['segregating'] = (ref, alt) snp_dict['context'] = '-' + ref + '-' # Add ancestral allele information if available info = cols[7].split(';') for field in info: if field.startswith('AA=') or field.startswith('AA_ensembl=') or field.startswith('AA_chimp='): outgroup_allele = field.split('=')[1].upper() if outgroup_allele not in ['A','C','G','T']: # Skip if ancestral not single base A, C, G, or T outgroup_allele = '-' break else: outgroup_allele = '-' snp_dict['outgroup_allele'] = outgroup_allele snp_dict['outgroup_context'] = '-' + outgroup_allele + '-' # Add flanking info if it is present rflank, aflank = flanking_info for field in info: if rflank and field.startswith(rflank): flank = field[len(rflank)+1:].upper() if not (len(flank) == 2 or len(flank) == 3): continue prevb, nextb = flank[0], flank[-1] if prevb not in ['A','C','T','G']: prevb = '-' if nextb not in ['A','C','T','G']: nextb = '-' snp_dict['context'] = prevb + ref + nextb continue if aflank and field.startswith(aflank): flank = field[len(aflank)+1:].upper() if not (len(flank) == 2 or len(flank) == 3): continue prevb, nextb = flank[0], flank[-1] if prevb not in ['A','C','T','G']: prevb = '-' if nextb not in ['A','C','T','G']: nextb = '-' snp_dict['outgroup_context'] = prevb + outgroup_allele + nextb calls_dict = {} subsample_dict = {} gtindex = cols[8].split(':').index('GT') if do_subsampling: # Collect data for all genotyped samples for pop, sample in zip(poplist, cols[9:]): if pop is None: continue gt = sample.split(':')[gtindex] if pop not in subsample_dict: subsample_dict[pop] = [] if '.' not in gt: subsample_dict[pop].append(gt) # key-value pairs here are population names # and a list of genotypes to subsample from for pop, genotypes in subsample_dict.items(): if pop not in calls_dict: calls_dict[pop] = (0, 0) if len(genotypes) < subsample[pop]: # Not enough calls for this SNP break # Choose which individuals to use idx = numpy.random.choice([i for i in range(0,len(genotypes))], subsample[pop], replace=False) for ii in idx: gt = subsample_dict[pop][ii] refcalls, altcalls = calls_dict[pop] refcalls += gt[::2].count('0') altcalls += gt[::2].count('1') calls_dict[pop] = (refcalls, altcalls) else: # Only runs if we didn't break out of this loop snp_dict['calls'] = calls_dict data_dict[snp_id] = snp_dict else: for pop, sample in zip(poplist, cols[9:]): if pop is None: continue if pop not in calls_dict: calls_dict[pop] = (0,0) # Genotype in VCF format 0|1|1|0:... gt = sample.split(':')[gtindex] #g1, g2 = gt[0], gt[2] #if g1 == '.' or g2 == '.': # continue #full_info = False #break refcalls, altcalls = calls_dict[pop] #refcalls += int(g1 == '0') + int(g2 == '0') #altcalls += int(g1 == '1') + int(g2 == '1') # Assume biallelic variants refcalls += gt[::2].count('0') altcalls += gt[::2].count('1') calls_dict[pop] = (refcalls, altcalls) snp_dict['calls'] = calls_dict data_dict[snp_id] = snp_dict vcf_file.close() return data_dict def make_fux_table(fid, ts, Q, tri_freq)-
Make file of 1-fux for use in ancestral misidentification correction.
fid: Filename to output to. ts: Expected number of substitutions per site between ingroup and outgroup. Q: Trinucleotide transition rate matrix. This should be a 64x64 matrix, in which entries are ordered using the code CGTA -> 0,1,2,3. For example, ACT -> 316+04+2*1=50. The transition rate from ACT to AGT is then entry 50,54. tri_freq: Dictionary in which each entry maps a trinucleotide to its ancestral frequency. e.g. {'AAA': 0.01, 'AAC':0.012…} Note that should be the frequency in the entire region scanned for variation, not just sites where there are SNPs.
Expand source code
def make_fux_table(fid, ts, Q, tri_freq): """ Make file of 1-fux for use in ancestral misidentification correction. fid: Filename to output to. ts: Expected number of substitutions per site between ingroup and outgroup. Q: Trinucleotide transition rate matrix. This should be a 64x64 matrix, in which entries are ordered using the code CGTA -> 0,1,2,3. For example, ACT -> 3*16+0*4+2*1=50. The transition rate from ACT to AGT is then entry 50,54. tri_freq: Dictionary in which each entry maps a trinucleotide to its ancestral frequency. e.g. {'AAA': 0.01, 'AAC':0.012...} Note that should be the frequency in the entire region scanned for variation, not just sites where there are SNPs. """ # Ensure that the *columns* of Q sum to zero. # That is the correct condition when Q_{i,j} is the rate from i to j. # This indicates a typo in Hernandez, Williamson, and Bustamante. for ii in range(Q.shape[1]): s = Q[:,ii].sum() - Q[ii,ii] Q[ii,ii] = -s eQhalf = scipy.linalg.matfuncs.expm(Q * ts/2.) if not hasattr(fid, 'write'): newfile = True fid = open(fid, 'w') outlines = [] for first_ii,first in enumerate(code): for x_ii,x in enumerate(code): for third_ii,third in enumerate(code): # This the index into Q and eQ xind = 16*first_ii+4*x_ii+1*third_ii for u_ii,u in enumerate(code): # This the index into Q and eQ uind = 16*first_ii+4*u_ii+1*third_ii ## Note that the Q terms factor out in our final ## calculation, because for both PMuUu and PMuUx the final ## factor in Eqn 2 is P(S={u,x}|M=u). #Qux = Q[uind,xind] #denomu = Q[uind].sum() - Q[uind,uind] PMuUu, PMuUx = 0,0 # Equation 2 in HWB. We have to generalize slightly to # calculate PMuUx. In calculate PMuUx, we're summing over # alpha the probability that the MRCA was alpha, and it # substituted to x on the outgroup branch, and it # substituted to u on the ingroup branch, and it mutated to # x in the ingroup (conditional on it having mutated in the # ingroup). Note that the mutation to x condition cancels # in fux, so we don't bother to calculate it. for aa,alpha in enumerate(code): aind = 16*first_ii+4*aa+1*third_ii pia = tri_freq[first+alpha+third] Pau = eQhalf[aind,uind] Pax = eQhalf[aind,xind] PMuUu += pia * Pau*Pau PMuUx += pia * Pau*Pax # This is 1-fux. For a given SNP with actual ancestral state # u and derived allele x, this is 1 minus the probability # that the outgroup will have u. # Eqn 3 in HWB. res = 1 - PMuUu/(PMuUu + PMuUx) # These aren't SNPs, so we can arbitrarily set them to 0 if u == x: res = 0 outlines.append('%c%c%c %c %.6f' % (first,x,third,u,res)) fid.write(os.linesep.join(outlines)) if newfile: fid.close() def ms_command(theta, ns, core, iter, recomb=0, rsites=None, seeds=None)-
Generate ms command for simulation from core.
theta: Assumed theta ns: Sample sizes core: Core of ms command that specifies demography. iter: Iterations to run ms recomb: Assumed recombination rate rsites: Sites for recombination. If None, default is 10*theta. seeds: Seeds for random number generator. If None, ms default is used. Otherwise, three integers should be passed. Example: (132, 435, 123)
Expand source code
def ms_command(theta, ns, core, iter, recomb=0, rsites=None, seeds=None): """ Generate ms command for simulation from core. theta: Assumed theta ns: Sample sizes core: Core of ms command that specifies demography. iter: Iterations to run ms recomb: Assumed recombination rate rsites: Sites for recombination. If None, default is 10*theta. seeds: Seeds for random number generator. If None, ms default is used. Otherwise, three integers should be passed. Example: (132, 435, 123) """ if len(ns) > 1: ms_command = "ms %(total_chrom)i %(iter)i -t %(theta)f -I %(numpops)i "\ "%(sample_sizes)s %(core)s" else: ms_command = "ms %(total_chrom)i %(iter)i -t %(theta)f %(core)s" if recomb: ms_command = ms_command + " -r %(recomb)f %(rsites)i" if not rsites: rsites = theta*10 sub_dict = {'total_chrom': numpy.sum(ns), 'iter': iter, 'theta': theta, 'numpops': len(ns), 'sample_sizes': ' '.join(map(str, ns)), 'core': core, 'recomb': recomb, 'rsites': rsites} ms_command = ms_command % sub_dict if seeds is not None: seed_command = " -seeds %i %i %i" % (seeds[0], seeds[1], seeds[2]) ms_command = ms_command + seed_command return ms_command def perturb_params(params, fold=1, lower_bound=None, upper_bound=None)-
Generate a perturbed set of parameters.
Each element of params is radomly perturbed
factors of 2 up or down. fold: Number of factors of 2 to perturb by lower_bound: If not None, the resulting parameter set is adjusted to have all value greater than lower_bound. upper_bound: If not None, the resulting parameter set is adjusted to have all value less than upper_bound. Expand source code
def perturb_params(params, fold=1, lower_bound=None, upper_bound=None): """ Generate a perturbed set of parameters. Each element of params is radomly perturbed <fold> factors of 2 up or down. fold: Number of factors of 2 to perturb by lower_bound: If not None, the resulting parameter set is adjusted to have all value greater than lower_bound. upper_bound: If not None, the resulting parameter set is adjusted to have all value less than upper_bound. """ pnew = params * 2**(fold * (2*numpy.random.uniform(size=len(params))-1)) if lower_bound is not None: for ii,bound in enumerate(lower_bound): if bound is None: lower_bound[ii] = -numpy.inf pnew = numpy.maximum(pnew, 1.01*numpy.asarray(lower_bound)) if upper_bound is not None: for ii,bound in enumerate(upper_bound): if bound is None: upper_bound[ii] = numpy.inf pnew = numpy.minimum(pnew, 0.99*numpy.asarray(upper_bound)) return pnew def simple_warning(message, category, filename, lineno, file=None, line=None)-
Expand source code
def simple_warning(message, category, filename, lineno, file=None, line=None): return '%s' % message def total_instantaneous_rate(Q, pi)-
Total instantaneous substitution rate.
Expand source code
def total_instantaneous_rate(Q, pi): """ Total instantaneous substitution rate. """ Qzero = zero_diag(Q) return numpy.dot(pi, Qzero).sum() def tri_freq_dict_to_array(tri_freq_dict)-
Convert dictionary of trinucleotide frequencies to array in correct order.
Expand source code
def tri_freq_dict_to_array(tri_freq_dict): """ Convert dictionary of trinucleotide frequencies to array in correct order. """ tripi = numpy.zeros(64) for ii,left in enumerate(code): for jj,center in enumerate(code): for kk,right in enumerate(code): row = ii*16 + jj*4 + kk tripi[row] = tri_freq_dict[left+center+right] return tripi def zero_diag(Q)-
Copy of Q altered such that diagonal entries are all 0.
Expand source code
def zero_diag(Q): """ Copy of Q altered such that diagonal entries are all 0. """ Q_nodiag = Q.copy() for ii in range(Q.shape[0]): Q_nodiag[ii,ii] = 0 return Q_nodiag