def test_read_msmc():
true_sizes = [
0.00086526447673996, 9.79088625144171e-05, 0.000171382542974173,
0.000457984501804459, 0.000272927253969727, 0.000131071906047658,
6.82882858274492e-05, 4.10030998343475e-05, 2.79371748811273e-05,
2.12348489352847e-05, 1.99020025394955e-05, 1.99020025394955e-05,
2.2919799039202e-05, 2.2919799039202e-05, 2.98841094236551e-05,
2.98841094236551e-05, 4.04164511122607e-05, 4.04164511122607e-05,
5.57810348943842e-05, 5.57810348943842e-05, 8.00343187158654e-05,
8.00343187158654e-05, 0.000116691016192045, 0.000116691016192045,
0.000167278907468, 0.000167278907468, 0.000226324223028942,
0.000226324223028942, 0.000287970327537451, 0.000287970327537451,
0.000345006037605658, 0.000345006037605658, 0.000392215310516861,
0.000392215310516861, 0.000424300540558889, 0.000424300540558889,
0.000447447313078885, 0.000447447313078885, 0.000841088166225936,
0.000841088166225936
]
true_times = [
1.58858e-06, 3.21843e-06, 4.89174e-06, 6.61091e-06, 8.3785e-06,
1.01973e-05, 1.20705e-05, 1.40013e-05, 1.59934e-05, 1.80508e-05,
2.01779e-05, 2.23798e-05, 2.46617e-05, 2.70297e-05, 2.94906e-05,
3.2052e-05, 3.47225e-05, 3.75116e-05, 4.04305e-05, 4.34919e-05,
4.67103e-05, 5.01028e-05, 5.36893e-05, 5.74932e-05, 6.15427e-05,
6.58717e-05, 7.05216e-05, 7.5544e-05, 8.10035e-05, 8.69838e-05,
9.35947e-05, 0.000100985, 0.000109364, 0.000119036, 0.000130476,
0.000144477, 0.000162528, 0.000187969, 0.000231461
]
sizes, times = read_msmc(joinpath(THIS_DIR, 'msmc_test.final.txt'), 1.0)
assert np.allclose(sizes, true_sizes)
assert np.allclose(times, true_times)
sizes, times = read_msmc(joinpath(THIS_DIR, 'msmc_test.final.txt'),
1.25e-8)
assert np.allclose(sizes, np.array(true_sizes) / 1.25e-8)
assert np.allclose(times, np.array(true_times) / 1.25e-8)
def test_issue6():
sizes, times = read_msmc(joinpath(THIS_DIR, 'issue6_msmc.txt'), 1.0)
N_ref = .0005 / 4.
times = [t / (2. * N_ref) for t in times]
sizes = [p / N_ref for p in sizes]
new_sizes, new_times = decimate_sizes(sizes, times, 0.0, None)
assert np.abs(new_times[0] - 1.85199e-07 / (2. * N_ref)) < 1e-13
def _main(args):
n_ref = 0.0005 / (4. * args.mu)
if args.msmc_file:
if args.smcpp_file or args.epochtimes or args.popsizes:
raise IOError('Can only specify one of msmc_file, smcpp_file, or '
'popsizes')
pop_sizes, times = read_msmc(args.msmc_file, args.mu)
elif args.smcpp_file:
if args.msmc_file or args.epochtimes or args.popsizes:
raise IOError('Can only specify one of msmc_file, smcpp_file, or '
'popsizes')
pop_sizes, times = read_smcpp(args.smcpp_file)
else:
times = []
if args.epochtimes:
times = list(map(float, args.epochtimes.split(',')))
pop_sizes = list(map(float, args.popsizes.split(',')))
times = [t / (2. * n_ref) for t in times]
pop_sizes = [p / n_ref for p in pop_sizes]
if len(pop_sizes) != len(times)+1:
raise IOError('Number of population sizes must match '
'number of epochs.')
pop_sizes, times = decimate_sizes(pop_sizes,
times,
args.decimate_rel_tol,
args.decimate_anc_size)
logging.info('Size history to be used when computing lookup table is\n'
+ 'Scaled Size\tScaled Left Time\tScaled Right Time\n'
+ '\n'.join([str(p) + '\t' + str(t1) + '\t' + str(t2)
for p, t1, t2 in zip(pop_sizes,
[0] + times,
times + [float('inf')])]))
max_size = args.samplesize
num_particles = max_size
if args.moran_pop_size:
if not args.approx:
raise IOError('Cannot use moran_pop_size when computing an exact '
'lookup table. Turn off --aprox flag.')
if max_size > args.moran_pop_size:
raise IOError('moran_pop_size must be at least as large as the '
'desired sample size.')
num_particles = args.moran_pop_size
rho_grid = [i * .1 for i in range(100)] + list(range(10, 101))
logging.info('Beginning Lookup Table. This may take a while')
table = LookupTable(num_particles, 0.0005, rho_grid, pop_sizes,
times, not args.approx, args.numthreads,
store_stationary=args.store_stationary,
load_stationary=args.load_stationary).table
logging.info('\t...complete')
table.columns /= 4. * n_ref
if num_particles > max_size:
logging.info('Downsampling')
table = downsample(table, max_size)
logging.info('\t...complete')
table.to_hdf(args.outfile, 'ldtable', mode='w')
def _main(args):
table = read_hdf(args.tablefile, 'ldtable')
table_size = sum(map(int, table.index.values[0].split()))
if table_size < args.samplesize:
raise IOError('Lookup table was constructed for {} haploids, '
'but --samplesize is {} haploids. Either build '
'a larger lookup table or simulate fewer '
'individuals.'.format(table_size, args.samplesize))
max_rho = table.columns[-1]
table.columns *= 100. / max_rho
block_penalties = list(map(float, args.blockpenalty.split(',')))
window_sizes = list(map(float, args.windowsize.split(',')))
logging.info('Searching over Windowsizes %s, and Block Penalties %s',
window_sizes, block_penalties)
if args.msmc_file:
if args.smcpp_file or args.epochtimes or args.popsizes:
raise IOError('Can only specify one of msmc_file, smcpp_file, or '
'popsizes')
pop_sizes, times = read_msmc(args.msmc_file, args.mu)
elif args.smcpp_file:
if args.msmc_file or args.epochtimes or args.popsizes:
raise IOError('Can only specify one of msmc_file, smcpp_file, or '
'popsizes')
pop_sizes, times = read_smcpp(args.smcpp_file)
else:
pop_sizes = list(map(float, args.popsizes.split(',')))
times = []
if args.epochtimes:
times = list(map(float, args.epochtimes.split(',')))
if len(pop_sizes) != len(times) + 1:
raise IOError('Number of population sizes must '
'match number of epochs.')
pop_sizes, times = decimate_sizes(pop_sizes,
times,
args.decimate_rel_tol,
args.decimate_anc_size)
pop_config = [
msprime.PopulationConfiguration(sample_size=args.samplesize,
initial_size=pop_sizes[0])]
demography = []
if times:
for pop_size, time in zip(pop_sizes[1:], times):
demography.append(
msprime.PopulationParametersChange(time=time * 2,
initial_size=pop_size,
population_id=0))
reco_maps = _load_hapmap()
pool = Pool(args.numthreads, maxtasksperchild=100)
logging.info('Simulating data...')
simulation_args = [((pop_config, args.mu, demography, args.ploidy),
reco_maps) for k in range(args.num_sims)]
test_set = list(pool.imap(_simulate_data, simulation_args, chunksize=10))
logging.info('\tdone simulating')
scores = {}
for block_penalty in block_penalties:
for window_size in window_sizes:
estimates = list(pool.imap(partial(_call_optimize,
metawindow=args.metawindow,
windowsize=window_size,
table=table,
ploidy=args.ploidy,
bpen=block_penalty,
overlap=args.overlap,
max_rho=max_rho),
test_set,
chunksize=10))
scores[(block_penalty,
window_size)] = _score(estimates,
[ts[1] for ts in test_set],
[ts[2] for ts in test_set],
pool)
ofile = open(args.outfile, 'w') if args.outfile else sys.stdout
ofile.write('\t'.join(['Block_Penalty',
'Window_Size',
'Pearson_Corr_1bp',
'Pearson_Corr_10kb',
'Pearson_Corr_100kb',
'Log_Pearson_Corr_1bp',
'Log_Pearson_Corr_10kb',
'Log_Pearson_Corr_100kb',
'Spearman_Corr_1bp',
'Spearman_Corr_10kb',
'Spearman_Corr_100kb',
'L2',
'Log_L2']) + '\n')
for block_penalty, window_size in sorted(scores):
line = ([block_penalty, window_size]
+ scores[block_penalty, window_size])
ofile.write('\t'.join(map(str, line)) + '\n')
if args.outfile:
ofile.close()