def get_one_IGC_event(self, seq, sub_IGC_init_Q, sub_IGC_tract_Q,
sub_total_IGC_init_Q, ordered_orlg, display):
np.fill_diagonal(sub_IGC_tract_Q,
0.0) # fill diagonal entries with 0, just in case..
# sample an IGC event orlg pair
IGC_pos = draw_from_distribution(sub_total_IGC_init_Q.ravel(), 1,
range(sub_total_IGC_init_Q.size))
orlg_from_num = int(floor(IGC_pos / sub_total_IGC_init_Q.ndim))
orlg_to_num = IGC_pos - orlg_from_num * sub_total_IGC_init_Q.ndim
orlg_from = ordered_orlg[orlg_from_num]
orlg_to = ordered_orlg[orlg_to_num]
# now sample a starting pos and tract length
tract_p = sub_IGC_tract_Q[orlg_from_num, orlg_to_num]
init_prob_array = np.array([1.0 / tract_p] + [1.0] *
(self.nsites / 3 - 1))
start_pos = draw_from_distribution(
init_prob_array / sum(init_prob_array), 1,
range(len(init_prob_array))) * 3
tract_length = np.random.geometric(tract_p, 1)[0] * 3
stop_pos = start_pos + tract_length - 1 # tract_length is a positive integer
if stop_pos > self.nsites - 1:
stop_pos = self.nsites - 1
#print start_pos, stop_pos, self.nsites
seq, IGC_info = self.IGC_copy(start_pos, stop_pos, orlg_from, orlg_to,
seq, display, tract_length)
return seq, IGC_info
def get_one_point_mutation(
self, seq, seq_rate_dict,
display): # modified from IGCSimulation.IGCSimulator
# only allow all sequences with same length
assert (len(set([len(seq_rate_dict[orlg])
for orlg in seq_rate_dict])) == 1)
orlg_group = sorted(seq_rate_dict.keys())
# Now concatenate all rates to one giant list to draw from distribution
# Use a dumb but safe way rather than comprehension
concatenated_rate = list()
for orlg in orlg_group:
concatenated_rate.extend(seq_rate_dict[orlg])
concatenated_rate = np.array(concatenated_rate) / sum(
concatenated_rate)
# Now sample a point mutation position
mut_pos = draw_from_distribution(concatenated_rate, 1,
range(len(concatenated_rate)))
mut_paralog_num = int(
floor(mut_pos / len(seq_rate_dict[orlg_group[0]])))
mut_paralog = orlg_group[mut_paralog_num]
seq_pos = mut_pos - mut_paralog_num * len(seq_rate_dict[orlg_group[0]])
# Now add in data_type difference: codon/nucleotide
if self.PMModel.data_type == 'cd':
states = self.PMModel.codon_nonstop
translated_seq = [
''.join(seq[mut_paralog][i:i + 3])
for i in range(0, len(seq[mut_paralog]), 3)
]
else:
states = 'ACGT'
translated_seq = seq[mut_paralog]
# Now perform point mutation at the position
old_state = translated_seq[seq_pos]
prob = np.array(self.PMModel.Q_mut[states.index(old_state), :])
new_state = states[draw_from_distribution(prob / sum(prob), 1,
range(len(prob)))]
translated_seq[seq_pos] = new_state
seq[mut_paralog] = [i for i in ''.join(translated_seq)]
# TODO: implement log
# mutation_orlg, mut_pos, old_state, new_state
mutation_info = [str(mut_paralog), str(seq_pos), old_state, new_state]
if display:
print ' '.join(mutation_info)
return seq, mutation_info
def sim_root(self):
root_name = self.tree.phylo_tree.root.name
root_conf = self.tree.node_to_conf[root_name]
root_orlg = divide_configuration(root_conf)
self.node_to_seq[root_name] = dict()
if self.PMModel.name == 'HKY':
distn = [self.PMModel.parameters['Pi_' + nt] for nt in 'ACGT']
for orlg in root_orlg['loc']:
seq = draw_from_distribution(distn, self.nsites, 'ACGT')
self.node_to_seq[root_name][orlg] = seq
elif self.PMModel.name == 'MG94':
distn = [
reduce(mul,
[self.PMModel.parameters['Pi_' + b] for b in codon], 1)
for codon in self.PMModel.codon_nonstop
]
distn = np.array(distn) / sum(distn)
for orlg in root_orlg['loc']:
seq = draw_from_distribution(distn, self.nsites / 3,
self.PMModel.codon_nonstop)
self.node_to_seq[root_name][orlg] = [i for i in ''.join(seq)]
def sim_one_branch(self, edge, display):
# First, make sure this branch is not simulated
assert (edge[0] in self.node_to_seq
and not edge[1] in self.node_to_seq)
blen = self.tree.edge_to_blen[edge]
starting_seq = self.node_to_seq[edge[0]]
conf = self.tree.node_to_conf[edge[0]]
branch_orlg = divide_configuration(conf)
current_seq = deepcopy(
starting_seq
) # it's passed on to the next node, need a new allocation of memory
branch_orlg = divide_configuration(conf)
#assert(all([orlg in branch_orlg['loc'] for orlg in starting_seq.keys()]))
# Get sub IGC init matrix from
ordered_orlg = sorted(branch_orlg['loc'])
if len(ordered_orlg) == 1:
Total_IGC_init_rate = 0.0
else:
branch_IGC_init_Q = np.zeros(
(len(ordered_orlg), len(ordered_orlg)), dtype=np.floating)
branch_IGC_tract_Q = np.zeros(
(len(ordered_orlg), len(ordered_orlg)), dtype=np.floating)
Total_IGC_init_Q = np.zeros((len(ordered_orlg), len(ordered_orlg)),
dtype=np.floating)
for i in range(len(ordered_orlg)):
for j in range(len(ordered_orlg)):
branch_IGC_init_Q[i, j] = self.IGCModel.Q_init[
ordered_orlg[i], ordered_orlg[j]]
branch_IGC_tract_Q[i, j] = self.IGCModel.Q_tract[
ordered_orlg[i], ordered_orlg[j]]
if i != j:
if branch_IGC_tract_Q[i, j] != 0:
Total_IGC_init_Q[i, j] = branch_IGC_init_Q[
i, j] * (self.nsites / 3 - 1 +
1.0 / branch_IGC_tract_Q[i, j])
IGC_init_rate_diag = branch_IGC_init_Q.sum(axis=1) # row sum
Total_IGC_init_rate = Total_IGC_init_Q.sum()
sub_IGC_tract_Q = branch_IGC_tract_Q
sub_total_IGC_init_Q = Total_IGC_init_Q
sub_IGC_init_Q = branch_IGC_init_Q
cummulate_time = 0.0
while (cummulate_time < blen):
# Now sample exponential distributed waiting time for next event
# point mutation or IGC event
# need to update total point mutation rate with every new event
# no need to update total IGC rate on the same branch since it's modeled as context independent
seq_rate_dict, Total_PM_rate = self.get_mutation_rate(current_seq)
Total_rate = Total_PM_rate + Total_IGC_init_rate
cummulate_time += np.random.exponential(1.0 / Total_rate)
if display:
print cummulate_time
if cummulate_time > blen:
break
else:
# Now decide whether it's a point mutation or IGC event
event = draw_from_distribution(
np.array([Total_PM_rate, Total_IGC_init_rate]) /
Total_rate, 1, range(2))
if event == 0:
# It's a point mutation event
current_seq, mutation_info = self.get_one_point_mutation(
current_seq, seq_rate_dict, display)
to_write_info = ['_'.join(edge),
str(cummulate_time)] + mutation_info
self.append_to_log_file(to_write_info, 'PM')
elif event == 1:
# It's an IGC event
current_seq, IGC_info = self.get_one_IGC_event(
current_seq, branch_IGC_init_Q, branch_IGC_tract_Q,
Total_IGC_init_Q, ordered_orlg, display)
to_write_info = ['_'.join(edge),
str(cummulate_time)] + IGC_info
self.append_to_log_file(to_write_info, 'IGC')
else:
# draw from distribution failure
assert (False)
# Now need to pass the seq to new node
# need to consider gene duplication loss events here
self.pass_seq_to_node(current_seq, edge)
