def _imap_consensus_reduce(self):
"""
Get consensus sequence for all samples in clade.
"""
# skip if no imap
if not self.imap:
return
# empty array of shape imap groups
iarr = np.zeros((len(self.imap), self.seqarr.shape[1]), dtype=np.uint8)
inames = np.array(sorted(self.imap.keys()))
# iterate over imap groups
for ikey, ivals in self.imap.items():
# get subarray for this group
sidxs = [np.where(self.names == i)[0][0] for i in ivals]
subarr = self.seqarr[sidxs, :]
# get consensus sequence
cons = reftrick(subarr, GETCONS)[:, 0]
# insert to iarr
iidx = np.where(inames == ikey)[0][0]
iarr[iidx] = cons
# save as new data
iarr[iarr == 0] = 78
self.seqarr = iarr
self.names = inames
self._longname = 1 + max([len(i) for i in self.names])
self._pnames = np.array([
"{}{}".format(name, " " * (self._longname - len(name)))
for name in self.names
])
def _loci_to_arr(loci, taxdict, mindict):
"""
return a frequency array from a loci file for all loci with taxa from
taxdict and min coverage from mindict.
"""
## get max length of loci
maxlen = np.max(np.array([len(locus.split("\n")[0]) for locus in loci]))
## make the array (4 or 5) and a mask array to remove loci without cov
nloci = len(loci)
maxlen = np.max(np.array([len(locus.split("\n")[0]) for locus in loci]))
keep = np.zeros(nloci, dtype=np.bool_)
arr = np.zeros((nloci, 4, maxlen), dtype=np.float64)
## six rows b/c one for each p3, and for the fused p3 ancestor
if len(taxdict) == 5:
# arr = np.zeros((nloci, 6, 300), dtype=np.float64)
arr = np.zeros((nloci, 6, maxlen), dtype=np.float64)
## if not mindict, make one that requires 1 in each taxon
if isinstance(mindict, int):
mindict = {i: mindict for i in taxdict}
elif isinstance(mindict, dict):
mindict = {i: mindict[i] for i in taxdict}
else:
mindict = {i: 1 for i in taxdict}
## raise error if names are not 'p[int]'
allowed_names = ['p1', 'p2', 'p3', 'p4', 'p5']
if any([i not in allowed_names for i in taxdict]):
raise IPyradError(\
"keys in taxdict must be named 'p1' through 'p4' or 'p5'")
## parse key names
keys = sorted([i for i in taxdict.keys() if i[0] == 'p'])
outg = keys[-1]
## grab seqs just for the good guys
for loc in range(nloci):
## parse the locus
lines = loci[loc].split("\n")[:-1]
names = [i.split()[0] for i in lines]
seqs = np.array([list(i.split()[1]) for i in lines])
## check that names cover the taxdict (still need to check by site)
covs = [sum([j in names for j in taxdict[tax]]) >= mindict[tax] \
for tax in taxdict]
## keep locus
if all(covs):
keep[loc] = True
## get the refseq
refidx = np.where([i in taxdict[outg] for i in names])[0]
refseq = seqs[refidx].view(np.uint8)
ancestral = np.array([reftrick(refseq, GETCONS2)[:, 0]])
## freq of ref in outgroup
iseq = _reffreq2(ancestral, refseq, GETCONS2)
arr[loc, -1, :iseq.shape[1]] = iseq
## enter 4-taxon freqs
if len(taxdict) == 4:
for tidx, key in enumerate(keys[:-1]):
## get idx of names in test tax
nidx = np.where([i in taxdict[key] for i in names])[0]
sidx = seqs[nidx].view(np.uint8)
## get freq of sidx
iseq = _reffreq2(ancestral, sidx, GETCONS2)
## fill it in
arr[loc, tidx, :iseq.shape[1]] = iseq
else:
## entere p5; and fill it in
iseq = _reffreq2(ancestral, refseq, GETCONS2)
arr[loc, -1, :iseq.shape[1]] = iseq
## enter p1
nidx = np.where([i in taxdict['p1'] for i in names])[0]
sidx = seqs[nidx].view(np.uint8)
iseq = _reffreq2(ancestral, sidx, GETCONS2)
arr[loc, 0, :iseq.shape[1]] = iseq
## enter p2
nidx = np.where([i in taxdict['p2'] for i in names])[0]
sidx = seqs[nidx].view(np.uint8)
iseq = _reffreq2(ancestral, sidx, GETCONS2)
arr[loc, 1, :iseq.shape[1]] = iseq
## enter p3 with p4 masked, and p4 with p3 masked
nidx = np.where([i in taxdict['p3'] for i in names])[0]
nidy = np.where([i in taxdict['p4'] for i in names])[0]
sidx = seqs[nidx].view(np.uint8)
sidy = seqs[nidy].view(np.uint8)
xseq = _reffreq2(ancestral, sidx, GETCONS2)
yseq = _reffreq2(ancestral, sidy, GETCONS2)
mask3 = xseq != 0
mask4 = yseq != 0
xseq[mask4] = 0
yseq[mask3] = 0
arr[loc, 2, :xseq.shape[1]] = xseq
arr[loc, 3, :yseq.shape[1]] = yseq
## enter p34
nidx = nidx.tolist() + nidy.tolist()
sidx = seqs[nidx].view(np.uint8)
iseq = _reffreq2(ancestral, sidx, GETCONS2)
arr[loc, 4, :iseq.shape[1]] = iseq
## size-down array to the number of loci that have taxa for the test
arr = arr[keep, :, :]
## size-down sites to
arr = masknulls(arr)
return arr, keep