def get_mutation_fixation_trajectories(population):
mutations, depth_tuple = parse_annotated_timecourse(population)
population_avg_depth_times, population_avg_depths, clone_avg_depth_times, clone_avg_depths = depth_tuple
state_times, state_trajectories = parse_well_mixed_state_timecourse(population)
times = mutations[0][9]
Ms = numpy.zeros_like(times)*1.0
fixed_Ms = numpy.zeros_like(times)*1.0
#transit_times[population] = []
for mutation_idx in range(0,len(mutations)):
#location, gene_name, allele, var_type, test_statistic, pvalue, cutoff_idx, depth_fold_change, depth_change_pvalue, times, alts, depths, clone_times, clone_alts, clone_depths = mutations[mutation_idx]
location, gene_name, allele, var_type, codon, position_in_codon, AAs_count, test_statistic, pvalue, cutoff_idx, depth_fold_change, depth_change_pvalue, times, alts, depths, clone_times, clone_alts, clone_depths = mutations[mutation_idx]
state_Ls = state_trajectories[mutation_idx]
good_idxs, filtered_alts, filtered_depths = timecourse_utils.mask_timepoints(times, alts, depths, var_type, cutoff_idx, depth_fold_change, depth_change_pvalue)
freqs = timecourse_utils.estimate_frequencies(filtered_alts, filtered_depths)
masked_times = times[good_idxs]
masked_freqs = freqs[good_idxs]
masked_state_Ls = state_Ls[good_idxs]
t0,tf,transit_time = timecourse_utils.calculate_appearance_fixation_time_from_hmm(masked_times, masked_freqs, masked_state_Ls)
#print(t0,tf,transit_time)
if t0==tf==transit_time==None:
continue
#print(masked_times, masked_freqs)
interpolating_function = timecourse_utils.create_interpolation_function(masked_times, masked_freqs, tmax=100000)
fs = interpolating_function(times)
fs[fs<0]=0
# Record
Ms += fs
if masked_state_Ls[-1] in well_mixed_fixed_states:
fixed_Ms += (times>=tf)
return times, Ms, fixed_Ms
state_times, state_trajectories = parse_file.parse_well_mixed_state_timecourse(population)
times = mutations[0][10]
Ms = numpy.zeros_like(times)*1.0
fixed_Ms = numpy.zeros_like(times)*1.0
transit_times[population] = []
for mutation_idx in xrange(0,len(mutations)):
location, gene_name, allele, var_type, test_statistic, pvalue, cutoff_idx, depth_fold_change, depth_change_pvalue, times, alts, depths, clone_times, clone_alts, clone_depths = mutations[mutation_idx]
Ls = haplotype_trajectories[mutation_idx]
state_Ls = state_trajectories[mutation_idx]
good_idxs, filtered_alts, filtered_depths = timecourse_utils.mask_timepoints(times, alts, depths, var_type, cutoff_idx, depth_fold_change, depth_change_pvalue)
freqs = timecourse_utils.estimate_frequencies(filtered_alts, filtered_depths)
masked_times = times[good_idxs]
masked_freqs = freqs[good_idxs]
masked_state_Ls = state_Ls[good_idxs]
t0,tf,transit_time = timecourse_utils.calculate_appearance_fixation_time_from_hmm(masked_times, masked_freqs, masked_state_Ls)
transit_times[population].append(transit_time)
interpolating_function = timecourse_utils.create_interpolation_function(masked_times, masked_freqs, tmax=100000)
fs = interpolating_function(times)
fs[fs<0]=0
# if pvalue is lower than threshold and not a weird insertion gene
passed_str = 'FAIL'
if (pvalue <= threshold_pvalue) and gene_name != 'repeat' and (
location != parse_file.ancestral_araA_mutation_location
) and (location != parse_file.ancestral_recD_mutation_location) and (
alts[0] * 1.0 / (depths[0] + (depths[0] == 0)) < 0.2):
# would otherwise pass
# check if clone fail
# determine whether clone data suggests that the mutation
# is a duplication. do not pass these
# first estimate frequencies at good timepoints
good_idxs, filtered_alts, filtered_depths = timecourse_utils.mask_timepoints(
times, alts, depths, var_type, deletion_idx, fold_reduction,
deletion_pvalue)
freqs = timecourse_utils.estimate_frequencies(
filtered_alts, filtered_depths)
masked_times = times[good_idxs]
masked_freqs = freqs[good_idxs]
masked_depth_ratios = depths[good_idxs] / pop_avg_depths[good_idxs]
interpolation_function = timecourse_utils.create_interpolation_function(
masked_times, masked_freqs, tmax=100000)
masked_clone_times, masked_clone_freqs = timecourse_utils.estimate_clone_frequencies(
clone_times, clone_alts, clone_depths)
if len(masked_clone_times) == 0:
good_mutations = []
sys.stderr.write("Processing %s\n" % population)
mutation_data, depth_tuple = parse_file.parse_annotated_timecourse(
population)
times = mutation_data[0][10]
for mutation_idx in xrange(0, len(mutation_data)):
location, gene_name, allele, var_type, test_statistic, pvalue, cutoff_idx, depth_fold_change, depth_change_pvalue, times, alts, depths, clone_times, clone_alts, clone_depths = mutation_data[
mutation_idx]
good_idxs, masked_alts, masked_depths = timecourse_utils.mask_timepoints(
times, alts, depths, var_type, cutoff_idx, depth_fold_change,
depth_change_pvalue, min_coverage)
max_freq = (masked_alts * 1.0 / (masked_depths +
(masked_depths < 1))).max()
num_good_timepoints = (masked_depths > 0.5).sum()
if num_good_timepoints > 5:
mutations.append((masked_alts * 1.0, masked_depths * 1.0))
if var_type != 'sv' and var_type != 'indel' and masked_depths[
-1] > 1:
#if masked_depths[-1]>1:
good_mutations.append((masked_alts * 1.0, masked_depths * 1.0))
else:
mutations.append(
(numpy.zeros_like(times) * 1.0, numpy.zeros_like(times)))
def run_analyses():
r2s_obs_dict = {}
#r2s_null_dict = {}
for treatment in ['0', '1', '2']:
r2s_obs_dict[treatment] = {}
for taxon in taxa:
r2s_all = []
ratio_f_all = []
abs_delta_f_all = []
for replicate in replicates:
population = treatment + taxon + replicate
sys.stderr.write("Processing %s...\n" % population)
mutations, depth_tuple = parse_file.parse_annotated_timecourse(
population)
population_avg_depth_times, population_avg_depths, clone_avg_depth_times, clone_avg_depths = depth_tuple
state_times, state_trajectories = parse_file.parse_well_mixed_state_timecourse(
population)
times = mutations[0][12]
Ms = np.zeros_like(times) * 1.0
fixed_Ms = np.zeros_like(times) * 1.0
for mutation_idx_i in range(0, len(mutations)):
location_i, gene_name_i, allele_i, var_type_i, codon_i, position_in_codon_i, AAs_count_i, test_statistic_i, pvalue_i, cutoff_idx_i, depth_fold_change_i, depth_change_pvalue_i, times_i, alts_i, depths_i, clone_times_i, clone_alts_i, clone_depths_i = mutations[
mutation_idx_i]
state_Ls_i = state_trajectories[mutation_idx_i]
good_idx_i, filtered_alts_i, filtered_depths_i = timecourse_utils.mask_timepoints(
times_i, alts_i, depths_i, var_type_i, cutoff_idx_i,
depth_fold_change_i, depth_change_pvalue_i)
freqs_i = timecourse_utils.estimate_frequencies(
filtered_alts_i, filtered_depths_i)
masked_times_i = times[good_idx_i]
masked_freqs_i = freqs_i[good_idx_i]
masked_state_Ls_i = state_Ls_i[good_idx_i]
P_idx_i = np.where(masked_state_Ls_i == 3)[0]
if len(P_idx_i) < min_trajectory_length:
continue
first_P_i = P_idx_i[0]
last_P_i = P_idx_i[-1]
masked_freqs_P_i = masked_freqs_i[first_P_i:last_P_i + 1]
masked_times_P_i = masked_times_i[first_P_i:last_P_i + 1]
delta_masked_freqs_P_i = masked_freqs_P_i[
1:] - masked_freqs_P_i[:-1]
delta_masked_times_P_i = masked_times_P_i[:-1]
#abs_delta_f = np.absolute(freqs_i[1:] - freqs_i[:-1])
#freqs_i_no_zero = freqs_i[freqs_i>0]
# we want to get the ratio of freqs
for freqs_i_k, freqs_i_l in zip(freqs_i[1:], freqs_i[:-1]):
if (freqs_i_k == 0) or (freqs_i_l == 0):
continue
abs_delta_f_all.append(
np.absolute(freqs_i_k - freqs_i_l))
ratio_f_all.append(freqs_i_k / freqs_i_l)
#ratio_f = freqs_i_no_zero[]
for mutation_idx_j in range(mutation_idx_i + 1,
len(mutations)):
location_j, gene_name_j, allele_j, var_type_j, codon_j, position_in_codon_j, AAs_count_j, test_statistic_j, pvalue_j, cutoff_jdx_j, depth_fold_change_j, depth_change_pvalue_j, times_j, alts_j, depths_j, clone_times_j, clone_alts_j, clone_depths_j = mutations[
mutation_idx_j]
state_Ls_j = state_trajectories[mutation_idx_j]
good_idx_j, filtered_alts_j, filtered_depths_j = timecourse_utils.mask_timepoints(
times_j, alts_j, depths_j, var_type_j,
cutoff_jdx_j, depth_fold_change_j,
depth_change_pvalue_j)
freqs_j = timecourse_utils.estimate_frequencies(
filtered_alts_j, filtered_depths_j)
masked_times_j = times[good_idx_j]
masked_freqs_j = freqs_j[good_idx_j]
masked_state_Ls_j = state_Ls_j[good_idx_j]
P_jdx_j = np.where(masked_state_Ls_j == 3)[0]
if len(P_jdx_j) < min_trajectory_length:
continue
first_P_j = P_jdx_j[0]
last_P_j = P_jdx_j[-1]
masked_freqs_P_j = masked_freqs_j[first_P_j:last_P_j +
1]
masked_times_P_j = masked_times_j[first_P_j:last_P_j +
1]
delta_masked_freqs_P_j = masked_freqs_P_j[
1:] - masked_freqs_P_j[:-1]
# delta_f = f_t_plus_1 - f_t
delta_masked_times_P_j = masked_times_P_j[:-1]
intersect_times = np.intersect1d(
delta_masked_times_P_i, delta_masked_times_P_j)
if len(intersect_times) >= 3:
intersect_idx_i = [
np.where(delta_masked_times_P_i ==
intersect_time)[0][0]
for intersect_time in intersect_times
]
intersect_delta_i = delta_masked_freqs_P_i[
intersect_idx_i]
intersect_idx_j = [
np.where(delta_masked_times_P_j ==
intersect_time)[0][0]
for intersect_time in intersect_times
]
intersect_delta_j = delta_masked_freqs_P_j[
intersect_idx_j]
if len(intersect_delta_i) != len(
intersect_delta_j):
print(len(intersect_delta_j),
len(intersect_delta_j))
r2 = stats.pearsonr(intersect_delta_i,
intersect_delta_j)[0]**2
r2s_all.append(r2)
r2s_all = np.asarray(r2s_all)
ratio_f_all = np.asarray(ratio_f_all)
abs_delta_f_all = np.asarray(abs_delta_f_all)
#r2s_obs_dict[treatment + taxon] = {}
#r2s_obs_dict[treatment + taxon]['r2'] = r2s_all
#r2s_obs_dict[treatment + taxon]['ratio_f'] = ratio_f_all
#r2s_obs_dict[treatment + taxon]['abs_delta_f'] = abs_delta_f_all
r2s_obs_dict[treatment][taxon] = {}
r2s_obs_dict[treatment][taxon]['r2'] = r2s_all
r2s_obs_dict[treatment][taxon]['ratio_f'] = ratio_f_all
r2s_obs_dict[treatment][taxon]['abs_delta_f'] = abs_delta_f_all
with open(pt.get_path() + '/data/mutation_dynamics.pickle',
'wb') as handle:
pickle.dump(r2s_obs_dict, handle, protocol=pickle.HIGHEST_PROTOCOL)
