afs = afs[:, :2718]
# Find consensus sequence
consind = afs.argmax(axis=0)
consensus = alpha[consind]
# Some seqs are not viable and include frameshifts that mess up the
# translation, hence we restrict to positions for which gaps are a minority
is_gap = consensus == '-'
# Exclude full codons
tmp = np.unique(is_gap.nonzero()[0] / 3)
is_gap[np.concatenate([tmp * 3, tmp * 3 + 1, tmp * 3 + 2])] = True
# Exclude stop codons
is_stop = np.zeros_like(is_gap)
tmp = (translate(consensus) == '*').nonzero()[0]
is_stop[np.concatenate([tmp * 3, tmp * 3 + 1, tmp * 3 + 2])] = True
## Plot base prevalence
#for i in xrange(4):
# plt.plot(np.arange(len(consensus)), afs[i],
# lw=1.5, alpha=0.5)
#plt.xlim(0, 2600)
#plt.ylim(-0.05, 1.25)
#plt.xlabel('position in '+gene)
#plt.ylabel('allele frequency')
#plt.legend(alpha, loc=9)
# Good are codons with no gaps and no stops
is_good = (-is_gap) & (-is_stop)
afs = afs[:, :2718]
# Find consensus sequence
consind = afs.argmax(axis=0)
consensus = alpha[consind]
# Some seqs are not viable and include frameshifts that mess up the
# translation, hence we restrict to positions for which gaps are a minority
is_gap = consensus == '-'
# Exclude full codons
tmp = np.unique(is_gap.nonzero()[0] / 3)
is_gap[np.concatenate([tmp * 3, tmp * 3 + 1, tmp * 3 + 2])] = True
# Exclude stop codons
is_stop = np.zeros_like(is_gap)
tmp = (translate(consensus) == '*').nonzero()[0]
is_stop[np.concatenate([tmp * 3, tmp * 3 + 1, tmp * 3 + 2])] = True
# Good are codons with no gaps and no stops
is_good = (-is_gap) & (-is_stop)
# For each codon, calculate the entropy
msa = msa[:, is_good]
consaa = translate(consensus[is_good])
entropy = np.zeros(len(consaa))
from collections import Counter
for i, aa in enumerate(consaa):
tmp = msa[:, i * 3:(i + 1) * 3]
count = Counter(map(''.join, tmp))
abundances = []
for (cod, abundance) in count.iteritems():
is_mut = np.array(
[alpha[al[0]] != consensus[al[1]] for al in alleles],
bool)
alleles = alleles[is_mut]
# Keep only synonymous
all_cla = {x: [] for x in classes}
if len(alleles):
is_syn = np.zeros(len(alleles), bool)
for j, al in enumerate(alleles):
pos = al[1]
mut = alpha[al[0]]
codcons = consensus[pos - pos % 3:pos - pos % 3 + 3]
cod = codcons.copy()
cod[pos % 3] = mut
aacons = translate(codcons)
aa = translate(cod)
is_syn[j] = (aacons == aa)
alleles = alleles[is_syn]
# Test more stringently for synonymity
# (we want to avoid double-hits in one single codon)
if len(alleles):
seqs = np.array(p.seqs_from_visit(p.visit[i]))
is_single = np.zeros(len(alleles), bool)
for j, al in enumerate(alleles):
pos = al[1]
mut = alpha[al[0]]
# Check whether sequences have double-hits
codcons = consensus[pos - pos % 3:pos - pos % 3 + 3]
if not len(alleles):
continue
###############################################################
# Filter only synonymous/nonsynonymous changes
###############################################################
# First test
is_syn = np.zeros(len(alleles), bool)
for j, al in enumerate(alleles):
pos = al[1]
mut = alpha[al[0]]
codcons = consensus[pos - pos % 3:pos - pos % 3 + 3]
cod = codcons.copy()
cod[pos % 3] = mut
is_syn[j] = (translate(codcons) == translate(cod))
alleles_syn = alleles[is_syn]
alleles_nonsyn = alleles[-is_syn]
# Test more stringently (avoid double-hits in one single codon)
if len(alleles_syn):
seqs = np.array(p.seqs_from_visit(p.visit[i]))
is_single = np.zeros(len(alleles_syn), bool)
for j, al in enumerate(alleles_syn):
pos = al[1]
mut = alpha[al[0]]
# Check whether sequences have double-hits
# If a double mutant is *ever* observed, discard the allele.
# Note: This is a very conservative measure and must
# be avoided when estimating densities.