def test_population_configurations(self):
def f(configs):
return msprime.simulator_factory(population_configurations=configs)
for bad_type in [10, ["sdf"], "sdfsd"]:
self.assertRaises(TypeError, f, bad_type)
# Just test the basic equalities here. The actual
# configuration options are tested elewhere.
for N in range(1, 10):
pop_configs = [
msprime.PopulationConfiguration(5, initial_size=5)
for _ in range(N)
]
sample_size = 5 * N
sim = msprime.simulator_factory(
population_configurations=pop_configs)
self.assertEqual(len(sim.demography.populations), len(pop_configs))
for pop, pop_config in zip(sim.demography.populations,
pop_configs):
self.assertEqual(pop.initial_size, pop_config.initial_size)
self.assertEqual(pop.growth_rate, pop_config.growth_rate)
self.assertEqual(len(sim.samples), sample_size)
self.assertEqual(len(sim.population_configuration), N)
# The default is a single population
sim = msprime.simulator_factory(10)
self.assertEqual(len(sim.population_configuration), 1)
def test_new_old_style_model_changes_equal(self):
models = [
msprime.SweepGenicSelection(
position=j,
start_frequency=j,
end_frequency=j,
alpha=j,
dt=j,
) for j in range(1, 10)
]
# Old style
sim = msprime.simulator_factory(
sample_size=2,
Ne=10,
demographic_events=[
msprime.SimulationModelChange(None, model) for model in models
],
)
self.assertEqual(len(sim.model_change_events), len(models))
for event, model in zip(sim.model_change_events, models):
self.assertEqual(event.model, model)
sim2 = msprime.simulator_factory(
sample_size=2,
Ne=10,
model=[None] +
[msprime.SimulationModelChange(None, model) for model in models],
)
self.assertEqual(sim.model_change_events, sim2.model_change_events)
def test_model_change_old_style(self):
main_model = msprime.SmcApproxCoalescent()
sim = msprime.simulator_factory(
Ne=100,
sample_size=2,
model=main_model,
demographic_events=[
msprime.SimulationModelChange(
1, msprime.DiscreteTimeWrightFisher()),
msprime.SimulationModelChange(2, None),
],
)
self.assertEqual(len(sim.model_change_events), 2)
self.assertEqual(sim.model_change_events[0].time, 1)
# When model=None we change to the standard coalescent
self.assertEqual(sim.model_change_events[1].time, 2)
self.assertEqual(sim.model_change_events[1].model.name, "hudson")
# This should be the same in new notation
sim = msprime.simulator_factory(
Ne=100,
sample_size=2,
model=[main_model, (1, "dtwf"), (2, None)],
)
self.assertEqual(len(sim.model_change_events), 2)
self.assertEqual(sim.model_change_events[0].time, 1)
# When model=None we change to the standard coalescent
self.assertEqual(sim.model_change_events[1].time, 2)
self.assertEqual(sim.model_change_events[1].model.name, "hudson")
def test_named_model_variants(self):
simulation_models = [("hudson", msprime.StandardCoalescent),
("smc", msprime.SmcApproxCoalescent),
("smc_prime", msprime.SmcPrimeApproxCoalescent)]
for name, model in simulation_models:
sim = msprime.simulator_factory(sample_size=10, model=name.upper())
self.assertIsInstance(sim.get_model(), model)
sim = msprime.simulator_factory(sample_size=10, model=name.title())
self.assertIsInstance(sim.get_model(), model)
def test_random_seed(self):
seed = 12345
sim = msprime.simulator_factory(10, random_seed=seed)
self.assertEqual(sim.random_generator.get_seed(), seed)
# It's an error to specify both seed and generator.
with self.assertRaises(ValueError):
msprime.simulator_factory(
10,
random_seed=1234,
random_generator=_msprime.RandomGenerator(1234))
def test_named_model_variants(self):
simulation_models = [
("hudson", msprime.StandardCoalescent),
("smc", msprime.SmcApproxCoalescent),
("smc_prime", msprime.SmcPrimeApproxCoalescent),
("dtwf", msprime.DiscreteTimeWrightFisher)
]
for name, model in simulation_models:
sim = msprime.simulator_factory(sample_size=10, model=name.upper())
self.assertIsInstance(sim.model, model)
sim = msprime.simulator_factory(sample_size=10, model=name.title())
self.assertIsInstance(sim.model, model)
def test_hudson(self):
threshold = 20
sim = msprime.simulator_factory(sample_size=10, recombination_rate=5)
sim.run()
self.assertGreater(sim.get_num_common_ancestor_events(), threshold)
self.assertGreater(sim.get_num_recombination_events(), threshold)
self.assertEqual(sim.get_num_rejected_common_ancestor_events(), 0)
sim = msprime.simulator_factory(
sample_size=10, recombination_rate=5, model="hudson")
sim.run()
self.assertGreater(sim.get_num_common_ancestor_events(), threshold)
self.assertGreater(sim.get_num_recombination_events(), threshold)
self.assertEqual(sim.get_num_rejected_common_ancestor_events(), 0)
def verify_simulation(self, n, m, r):
"""
Verifies a simulation for the specified parameters.
"""
recomb_map = msprime.RecombinationMap.uniform_map(m, r, num_loci=m)
rng = msprime.RandomGenerator(1)
sim = msprime.simulator_factory(
n, recombination_map=recomb_map, random_generator=rng)
self.assertEqual(sim.random_generator, rng)
sim.run()
self.assertEqual(sim.num_breakpoints, len(sim.breakpoints))
self.assertGreater(sim.time, 0)
self.assertGreater(sim.num_avl_node_blocks, 0)
self.assertGreater(sim.num_segment_blocks, 0)
self.assertGreater(sim.num_node_mapping_blocks, 0)
tree_sequence = sim.get_tree_sequence()
t = 0.0
for record in tree_sequence.nodes():
if record.time > t:
t = record.time
self.assertEqual(sim.time, t)
self.assertGreater(sim.num_common_ancestor_events, 0)
self.assertGreaterEqual(sim.num_recombination_events, 0)
self.assertGreaterEqual(sim.total_num_migration_events, 0)
self.assertGreaterEqual(sim.num_multiple_recombination_events, 0)
self.verify_sparse_trees(tree_sequence)
self.verify_dump_load(tree_sequence)
def test_multimerger(self):
rng = msprime.RandomGenerator(1234)
sim = msprime.simulator_factory(
100, recombination_rate=0.1, record_full_arg=True,
random_generator=rng, demographic_events=[
msprime.InstantaneousBottleneck(time=0.1, population=0, strength=5)])
self.verify(sim, multiple_mergers=True)
def test_multimerger(self):
rng = msprime.RandomGenerator(1234)
sim = msprime.simulator_factory(
100, recombination_rate=0.1, record_full_arg=True,
random_generator=rng, demographic_events=[
msprime.InstantaneousBottleneck(time=0.1, population=0, strength=5)])
self.verify(sim, multiple_mergers=True)
def test_single_growth_rate(self):
# Set out our values in units of generations and absolute sizes.
Ne = 1000
growth_rate = -0.01
end_time = 20
end_size = Ne * math.exp(-growth_rate * end_time)
population_configurations = [
msprime.PopulationConfiguration(
sample_size=2, initial_size=Ne, growth_rate=growth_rate)]
demographic_events = [
msprime.PopulationParametersChange(time=end_time, growth_rate=0)]
simulator = msprime.simulator_factory(
Ne=Ne,
population_configurations=population_configurations,
demographic_events=demographic_events)
ll_sim = simulator.create_ll_instance()
ll_end_time = ll_sim.debug_demography()
self.assertEqual(end_time, ll_end_time)
populations = [
msprime.Population(**d)
for d in ll_sim.get_population_configuration()]
self.assertEqual(len(populations), 1)
pop = populations[0]
self.assertEqual(pop.growth_rate, growth_rate)
self.assertEqual(pop.initial_size, Ne)
self.assertEqual(pop.get_size(end_time), end_size)
# Now fast forward to the next time slice.
ll_end_time = ll_sim.debug_demography()
self.assertTrue(math.isinf(ll_end_time))
populations = [
msprime.Population(**d) for d in ll_sim.get_population_configuration()]
pop = populations[0]
self.assertEqual(pop.growth_rate, 0)
self.assertEqual(pop.initial_size, end_size)
self.assertEqual(pop.get_size(10), end_size)
def run_replicate(self, j, treefile):
recomb_map = msprime.RecombinationMap.uniform_map(
self.num_loci, self.recombination_rate, self.num_loci)
sim = msprime.simulator_factory(self.sample_size,
recombination_map=recomb_map,
Ne=self.effective_population_size)
sim.run()
mutation_rate = 0
if self.generate_haplotypes:
mutation_rate = self.mutation_rate
mutation_generator = msprime.MutationGenerator(
msprime.RandomGenerator(random.randint(1, 2**31)), mutation_rate)
tree_sequence = sim.get_tree_sequence(mutation_generator)
if self.generate_trees:
tree_sequence.dump(treefile)
if self.generate_haplotypes:
with open(treefile, "w") as f:
for h in tree_sequence.haplotypes():
print(h, file=f)
self.tree_file_size[j] = os.path.getsize(treefile)
self.used_memory[j] = sim.used_memory
self.num_trees[j] = sim.num_breakpoints
self.num_multiple_re_events = sim.num_multiple_recombination_events
self.num_re_events[j] = sim.num_recombination_events
self.num_ca_events[j] = sim.num_common_ancestor_events
self.num_records[j] = tree_sequence.num_edges
self.num_nodes[j] = tree_sequence.num_nodes
self.num_records_per_tree = get_mean_records_per_tree(tree_sequence)
def test_recombination_n100(self):
rng = _msprime.RandomGenerator(100)
sim = msprime.simulator_factory(100,
recombination_rate=0.2,
record_full_arg=True,
random_generator=rng)
self.verify(sim)
def verify_simulation(self, n, m, r):
"""
Verifies a simulation for the specified parameters.
"""
recomb_map = msprime.RecombinationMap.uniform_map(m, r)
rng = _msprime.RandomGenerator(1)
sim = msprime.simulator_factory(
n,
recombination_map=recomb_map,
random_generator=rng,
discrete_genome=True,
)
self.assertEqual(sim.random_generator, rng)
sim.run()
self.assertEqual(sim.num_breakpoints, len(sim.breakpoints))
self.assertGreater(sim.time, 0)
self.assertGreater(sim.num_avl_node_blocks, 0)
self.assertGreater(sim.num_segment_blocks, 0)
self.assertGreater(sim.num_node_mapping_blocks, 0)
tree_sequence = next(sim.run_replicates(1))
t = 0.0
for record in tree_sequence.nodes():
if record.time > t:
t = record.time
self.assertEqual(sim.time, t)
self.assertGreater(sim.num_common_ancestor_events, 0)
self.assertGreaterEqual(sim.num_recombination_events, 0)
self.assertGreaterEqual(np.sum(sim.num_migration_events), 0)
self.assertGreaterEqual(sim.num_multiple_recombination_events, 0)
def verify_simulation(self, n, m, r):
"""
Verifies a simulation for the specified parameters.
"""
recomb_map = msprime.RecombinationMap.uniform_map(m, r, num_loci=m)
rng = msprime.RandomGenerator(1)
sim = msprime.simulator_factory(n,
recombination_map=recomb_map,
random_generator=rng)
self.assertEqual(sim.random_generator, rng)
sim.run()
self.assertEqual(sim.num_breakpoints, len(sim.breakpoints))
self.assertGreater(sim.time, 0)
self.assertGreater(sim.num_avl_node_blocks, 0)
self.assertGreater(sim.num_segment_blocks, 0)
self.assertGreater(sim.num_node_mapping_blocks, 0)
tree_sequence = sim.get_tree_sequence()
t = 0.0
for record in tree_sequence.nodes():
if record.time > t:
t = record.time
self.assertEqual(sim.time, t)
self.assertGreater(sim.num_common_ancestor_events, 0)
self.assertGreaterEqual(sim.num_recombination_events, 0)
self.assertGreaterEqual(sim.total_num_migration_events, 0)
self.assertGreaterEqual(sim.num_multiple_recombination_events, 0)
self.verify_sparse_trees(tree_sequence)
self.verify_dump_load(tree_sequence)
def __init__(self,
sample_size=1,
num_loci=1,
scaled_recombination_rate=0,
num_replicates=1,
migration_matrix=None,
population_configurations=None,
demographic_events=None,
scaled_mutation_rate=0,
print_trees=False,
precision=3,
random_seeds=None,
scaled_gene_conversion_rate=0,
gene_conversion_track_length=1,
hotspots=None):
self._sample_size = sample_size
self._num_loci = num_loci
self._num_replicates = num_replicates
self._recombination_rate = scaled_recombination_rate
self._mutation_rate = scaled_mutation_rate
# For strict ms-compability we want to have m non-recombining loci
if hotspots is None:
self._recomb_map = msprime.RecombinationMap.uniform_map(
num_loci, self._recombination_rate, discrete=True)
else:
self._recomb_map = hotspots_to_recomb_map(hotspots,
self._recombination_rate,
num_loci)
# If we have specified any population_configurations we don't want
# to give the overall sample size.
sample_size = self._sample_size
if population_configurations is not None:
sample_size = None
# msprime measure's time in units of generations, given a specific
# Ne value whereas ms uses coalescent time. To be compatible with ms,
# we therefore need to use an Ne value of 1/4.
self._simulator = msprime.simulator_factory(
Ne=0.25,
sample_size=sample_size,
recombination_map=self._recomb_map,
population_configurations=population_configurations,
migration_matrix=migration_matrix,
demographic_events=demographic_events,
gene_conversion_rate=scaled_gene_conversion_rate,
gene_conversion_track_length=gene_conversion_track_length)
self._precision = precision
self._print_trees = print_trees
# sort out the random seeds
ms_seeds = random_seeds
if random_seeds is None:
ms_seeds = generate_seeds()
seed = get_single_seed(ms_seeds)
self._random_generator = msprime.RandomGenerator(seed)
self._ms_random_seeds = ms_seeds
self._simulator.random_generator = self._random_generator
self._mutation_generator = msprime.MutationGenerator(
self._random_generator, self._mutation_rate)
def test_migration_matrix(self):
# Cannot specify a migration matrix without population
# configurations
self.assertRaises(ValueError,
msprime.simulator_factory,
10,
migration_matrix=[])
for N in range(1, 10):
pop_configs = [
msprime.PopulationConfiguration(5) for _ in range(N)
]
sim = msprime.simulator_factory(
population_configurations=pop_configs)
ll_sim = sim.create_ll_instance()
# If we don't specify a matrix, it's 0 everywhere.
matrix = [0 for j in range(N * N)]
np.testing.assert_array_equal(ll_sim.get_migration_matrix(),
matrix)
def f(hl_matrix):
return msprime.simulator_factory(
population_configurations=pop_configs,
migration_matrix=hl_matrix)
hl_matrix = [[(j + k) * int(j != k) for j in range(N)]
for k in range(N)]
sim = f(hl_matrix)
self.assertEqual(sim.migration_matrix, hl_matrix)
# Try with equivalent numpy array.
sim = f(np.array(hl_matrix))
self.assertEqual(sim.migration_matrix, hl_matrix)
ll_sim = sim.create_ll_instance()
ll_matrix = [v for row in hl_matrix for v in row]
np.testing.assert_array_equal(ll_sim.get_migration_matrix(),
ll_matrix)
for bad_type in [234, 1.2]:
self.assertRaises(TypeError, f, bad_type)
# Iterables should raise a value error.
for bad_type in [{}, ""]:
self.assertRaises(ValueError, f, bad_type)
# Now check for the structure of the matrix.
hl_matrix[0][0] = "bad value"
sim = f(hl_matrix)
self.assertRaises(TypeError, sim.create_ll_instance)
hl_matrix[0] = None
self.assertRaises(TypeError, f, hl_matrix)
hl_matrix[0] = []
self.assertRaises(ValueError, f, hl_matrix)
# Simple numpy array.
hl_matrix = np.ones((N, N))
np.fill_diagonal(hl_matrix, 0)
sim = f(hl_matrix)
np.testing.assert_array_equal(np.array(sim.migration_matrix),
hl_matrix)
sim.run()
events = np.array(sim.num_migration_events)
self.assertEqual(events.shape, (N, N))
self.assertTrue(np.all(events >= 0))
def test_hudson(self):
threshold = 20
sim = msprime.simulator_factory(sample_size=10, recombination_rate=10)
sim.random_generator = msprime.RandomGenerator(2)
sim.run()
self.assertGreater(sim.num_common_ancestor_events, threshold)
self.assertGreater(sim.num_recombination_events, threshold)
self.assertEqual(sim.num_rejected_common_ancestor_events, 0)
sim2 = msprime.simulator_factory(
sample_size=10, recombination_rate=10, model="hudson")
sim2.random_generator = msprime.RandomGenerator(2)
sim2.run()
self.assertEqual(
sim2.num_common_ancestor_events, sim.num_common_ancestor_events)
self.assertEqual(
sim2.num_recombination_events, sim.num_recombination_events)
self.assertEqual(sim2.num_rejected_common_ancestor_events, 0)
def test_smc_variants(self):
for model in ["smc", "smc_prime"]:
threshold = 20
sim = msprime.simulator_factory(
sample_size=10, recombination_rate=5, model=model)
sim.run()
self.assertGreater(sim.get_num_common_ancestor_events(), threshold)
self.assertGreater(sim.get_num_recombination_events(), threshold)
self.assertGreater(sim.get_num_rejected_common_ancestor_events(), 0)
def test_no_recombination(self):
rng = msprime.RandomGenerator(1)
sim = msprime.simulator_factory(10, random_generator=rng, record_full_arg=True)
ts = self.verify(sim)
ts_simplified = ts.simplify()
t1 = ts.tables
t2 = ts_simplified.tables
self.assertEqual(t1.nodes, t2.nodes)
self.assertEqual(t1.edges, t2.edges)
def test_no_recombination(self):
rng = msprime.RandomGenerator(1)
sim = msprime.simulator_factory(10, random_generator=rng, record_full_arg=True)
ts = self.verify(sim)
ts_simplified = ts.simplify()
t1 = ts.tables
t2 = ts_simplified.tables
self.assertEqual(t1.nodes, t2.nodes)
self.assertEqual(t1.edges, t2.edges)
def test_specify_model_and_Ne(self):
# When them model reference size and Ne are both specified,
# Ne is ignored.
for Ne in [0, 1234, None, "sdf"]:
sim = msprime.simulator_factory(
sample_size=2,
Ne=Ne,
model=msprime.SmcPrimeApproxCoalescent(20))
self.assertEqual(sim.model.reference_size, 20)
def test_event_chunk(self):
sim = msprime.simulator_factory(10)
for bad_chunk in [-(2**32), -1, 0]:
with self.assertRaises(ValueError):
sim.run(event_chunk=bad_chunk)
sim.reset()
sim.run(event_chunk=2**32 + 1)
sim.reset()
sim.run(event_chunk=2**64 + 1)
def test_debug_logging_dtwf(self):
sim = msprime.simulator_factory(3, Ne=10, model="dtwf")
with self.assertLogs("msprime.ancestry", logging.DEBUG) as log:
sim.run(event_chunk=1)
self.assertGreaterEqual(len(log.output), 3)
self.assertTrue(log.output[0].startswith("INFO"))
self.assertTrue(log.output[-1].startswith("INFO"))
self.assertTrue(
log.output[1].startswith("DEBUG:msprime.ancestry:time="))
def test_hudson(self):
threshold = 20
sim = msprime.simulator_factory(sample_size=10, recombination_rate=10)
sim.random_generator = msprime.RandomGenerator(2)
sim.run()
self.assertGreater(sim.num_common_ancestor_events, threshold)
self.assertGreater(sim.num_recombination_events, threshold)
self.assertEqual(sim.num_rejected_common_ancestor_events, 0)
sim2 = msprime.simulator_factory(
sample_size=10, recombination_rate=10, model="hudson")
sim2.random_generator = msprime.RandomGenerator(2)
sim2.run()
self.assertEqual(
sim2.num_common_ancestor_events, sim.num_common_ancestor_events)
self.assertEqual(
sim2.num_recombination_events, sim.num_recombination_events)
self.assertEqual(sim2.num_rejected_common_ancestor_events, 0)
def test_effective_population_size(self):
def f(Ne):
return msprime.simulator_factory(10, Ne=Ne)
for bad_value in [-1, -1e16, 0]:
self.assertRaises(ValueError, f, bad_value)
for Ne in [1, 10, 1e5]:
sim = f(Ne)
self.assertEqual(sim.model.population_size, Ne)
# Test the default.
sim = msprime.simulator_factory(10)
def test_population_configurations(self):
def f(configs):
return msprime.simulator_factory(population_configurations=configs)
for bad_type in [10, ["sdf"], "sdfsd"]:
self.assertRaises(TypeError, f, bad_type)
# Just test the basic equalities here. The actual
# configuration options are tested elewhere.
for N in range(1, 10):
pop_configs = [msprime.PopulationConfiguration(5) for _ in range(N)]
sample_size = 5 * N
sim = msprime.simulator_factory(population_configurations=pop_configs)
self.assertEqual(sim.population_configurations, pop_configs)
self.assertEqual(len(sim.samples), sample_size)
ll_sim = sim.create_ll_instance()
self.assertEqual(len(ll_sim.get_population_configuration()), N)
# The default is a single population
sim = msprime.simulator_factory(10)
ll_sim = sim.create_ll_instance()
self.assertEqual(len(ll_sim.get_population_configuration()), 1)
def test_effective_population_size(self):
def f(Ne):
return msprime.simulator_factory(10, Ne=Ne)
for bad_value in [-1, -1e16, 0]:
self.assertRaises(ValueError, f, bad_value)
for Ne in [1, 10, 1e5]:
sim = f(Ne)
self.assertEqual(sim.model.population_size, Ne)
# Test the default.
sim = msprime.simulator_factory(10)
def test_debug_func(self):
sim = msprime.simulator_factory(10)
count = 0
def f(sim):
nonlocal count
count += 1
sim.run(event_chunk=1, debug_func=f)
self.assertGreater(count, 0)
def test_population_configurations(self):
def f(configs):
return msprime.simulator_factory(population_configurations=configs)
for bad_type in [10, ["sdf"], "sdfsd"]:
self.assertRaises(TypeError, f, bad_type)
# Just test the basic equalities here. The actual
# configuration options are tested elewhere.
for N in range(1, 10):
pop_configs = [msprime.PopulationConfiguration(5) for _ in range(N)]
sample_size = 5 * N
sim = msprime.simulator_factory(population_configurations=pop_configs)
self.assertEqual(sim.population_configurations, pop_configs)
self.assertEqual(len(sim.samples), sample_size)
ll_sim = sim.create_ll_instance()
self.assertEqual(len(ll_sim.get_population_configuration()), N)
# The default is a single population
sim = msprime.simulator_factory(10)
ll_sim = sim.create_ll_instance()
self.assertEqual(len(ll_sim.get_population_configuration()), 1)
def test_symmetric_growth_rates(self):
# Test a symmetric model where we start with a negative growth
# rate and then increase back to the same value.
Ne = 10001
growth_rate = 0.0125
delta_t = 50
end_size = Ne * math.exp(-growth_rate * delta_t)
population_configurations = [
msprime.PopulationConfiguration(sample_size=2,
initial_size=Ne,
growth_rate=growth_rate)
]
demographic_events = [
msprime.PopulationParametersChange(time=delta_t,
growth_rate=-growth_rate),
msprime.PopulationParametersChange(time=2 * delta_t, growth_rate=0)
]
simulator = msprime.simulator_factory(
Ne=Ne,
population_configurations=population_configurations,
demographic_events=demographic_events)
ll_sim = simulator.create_ll_instance()
ll_end_time = ll_sim.debug_demography()
t = delta_t
self.assertEqual(t, ll_end_time * 4 * Ne)
populations = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()
]
pop = populations[0]
self.assertEqual(pop.growth_rate, growth_rate)
self.assertEqual(pop.initial_size, Ne)
self.assertEqual(pop.get_size(delta_t), end_size)
# Now fast forward to the next time slice.
t += delta_t
ll_end_time = ll_sim.debug_demography()
self.assertEqual(t, ll_end_time * 4 * Ne)
pop = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()
][0]
self.assertEqual(pop.growth_rate, -growth_rate)
self.assertEqual(pop.initial_size, end_size)
self.assertEqual(pop.get_size(delta_t), Ne)
# Now fast forward to the next time slice.
ll_end_time = ll_sim.debug_demography()
self.assertTrue(math.isinf(ll_end_time))
populations = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()
]
pop = populations[0]
self.assertEqual(pop.growth_rate, 0)
self.assertEqual(pop.initial_size, Ne)
def test_sample_size(self):
self.assertRaises(ValueError, msprime.simulator_factory)
self.assertRaises(ValueError, msprime.simulator_factory, 1)
self.assertRaises(ValueError, msprime.simulator_factory, sample_size=1)
for n in [2, 100, 1000]:
sim = msprime.simulator_factory(n)
self.assertEqual(sim.num_samples, n)
self.assertEqual(len(sim.samples), n)
for sample in sim.samples:
self.assertEqual(sample[0], 0)
self.assertEqual(sample[1], 0)
def test_migration_matrix(self):
m = [
[0, 1, 2],
[3, 0, 4],
[5, 6, 0]]
sim = msprime.simulator_factory(
population_configurations=[
msprime.PopulationConfiguration(1),
msprime.PopulationConfiguration(1),
msprime.PopulationConfiguration(1)],
migration_matrix=m)
self.assertEqual(sim.migration_matrix, m)
def test_info_logging(self):
sim = msprime.simulator_factory(10)
with self.assertLogs("msprime.ancestry", logging.INFO) as log:
sim.run()
self.assertEqual(len(log.output), 2)
self.assertEqual(
log.output[0],
("INFO:msprime.ancestry:Running model {'name': 'hudson'} "
"until max time: inf"),
)
self.assertTrue(log.output[1].startswith(
"INFO:msprime.ancestry:Completed at time"))
def test_migration_matrix(self):
# Cannot specify a migration matrix without population
# configurations
self.assertRaises(ValueError,
msprime.simulator_factory,
10,
migration_matrix=[])
for N in range(1, 10):
pop_configs = [
msprime.PopulationConfiguration(5) for _ in range(N)
]
sim = msprime.simulator_factory(
population_configurations=pop_configs)
# If we don't specify a matrix, it's 0 everywhere.
matrix = np.zeros((N, N))
np.testing.assert_array_equal(sim.migration_matrix, matrix)
def f(matrix):
return msprime.simulator_factory(
population_configurations=pop_configs,
migration_matrix=matrix)
matrix = [[(j + k) * int(j != k) for j in range(N)]
for k in range(N)]
sim = f(matrix)
np.testing.assert_array_equal(sim.demography.migration_matrix,
matrix)
# Try with equivalent numpy array.
sim = f(np.array(matrix))
np.testing.assert_array_equal(sim.demography.migration_matrix,
matrix)
np.testing.assert_array_equal(sim.migration_matrix, matrix)
for bad_type in [{}, "", 234, 1.2]:
self.assertRaises(ValueError, f, bad_type)
# Now check for the structure of the matrix.
matrix[0][0] = "bad value"
self.assertRaises(ValueError, f, matrix)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
matrix[0] = None
self.assertRaises(ValueError, f, matrix)
matrix[0] = []
self.assertRaises(ValueError, f, matrix)
# Simple numpy array.
matrix = np.ones((N, N))
np.fill_diagonal(matrix, 0)
sim = f(matrix)
np.testing.assert_array_equal(
np.array(sim.demography.migration_matrix), matrix)
sim.run()
events = np.array(sim.num_migration_events)
self.assertEqual(events.shape, (N, N))
self.assertTrue(np.all(events >= 0))
def test_migration_matrix(self):
# Cannot specify a migration matrix without population
# configurations
self.assertRaises(
ValueError, msprime.simulator_factory, 10,
migration_matrix=[])
for N in range(1, 10):
pop_configs = [
msprime.PopulationConfiguration(5) for _ in range(N)]
sim = msprime.simulator_factory(
population_configurations=pop_configs)
ll_sim = sim.create_ll_instance()
# If we don't specify a matrix, it's 0 everywhere.
matrix = [0 for j in range(N * N)]
self.assertEqual(ll_sim.get_migration_matrix(), matrix)
def f(hl_matrix):
return msprime.simulator_factory(
population_configurations=pop_configs,
migration_matrix=hl_matrix)
hl_matrix = [
[(j + k) * int(j != k) for j in range(N)] for k in range(N)]
sim = f(hl_matrix)
self.assertEqual(sim.migration_matrix, hl_matrix)
# Try with equivalent numpy array.
sim = f(np.array(hl_matrix))
self.assertEqual(sim.migration_matrix, hl_matrix)
ll_sim = sim.create_ll_instance()
ll_matrix = [v for row in hl_matrix for v in row]
self.assertEqual(ll_sim.get_migration_matrix(), ll_matrix)
for bad_type in [234, 1.2]:
self.assertRaises(TypeError, f, bad_type)
# Iterables should raise a value error.
for bad_type in [{}, ""]:
self.assertRaises(ValueError, f, bad_type)
# Now check for the structure of the matrix.
hl_matrix[0][0] = "bad value"
sim = f(hl_matrix)
self.assertRaises(TypeError, sim.create_ll_instance)
hl_matrix[0] = None
self.assertRaises(TypeError, f, hl_matrix)
hl_matrix[0] = []
self.assertRaises(ValueError, f, hl_matrix)
# Simple numpy array.
hl_matrix = np.ones((N, N))
np.fill_diagonal(hl_matrix, 0)
sim = f(hl_matrix)
self.assertTrue(np.array_equal(np.array(sim.migration_matrix), hl_matrix))
sim.run()
events = np.array(sim.num_migration_events)
self.assertEqual(events.shape, (N, N))
self.assertTrue(np.all(events >= 0))
def test_perf_parameters(self):
sim = msprime.simulator_factory(10)
sim.run()
self.assertGreater(sim.avl_node_block_size, 0)
self.assertGreater(sim.segment_block_size, 0)
self.assertGreater(sim.node_mapping_block_size, 0)
sim.reset()
sim.avl_node_block_size = 1
sim.segment_block_size = 1
sim.node_mapping_block_size = 1
self.assertEqual(sim.avl_node_block_size, 1)
self.assertEqual(sim.segment_block_size, 1)
self.assertEqual(sim.node_mapping_block_size, 1)
def test_recombination_map(self):
def f(recomb_map):
return msprime.simulator_factory(10, recombination_map=recomb_map)
self.assertRaises(TypeError, f, "wrong type")
for n in range(2, 10):
positions = list(range(n))
rates = [0.1 * j for j in range(n - 1)]
# Use the old-form RecombinationMap
recomb_map = msprime.RecombinationMap(positions, rates + [0.0])
sim = msprime.simulator_factory(10, recombination_map=recomb_map)
other_map = msprime.RateMap(**sim.recombination_map)
self.assertEqual(list(other_map.position), positions)
self.assertEqual(list(other_map.rate), rates)
self.assertEqual(sim.sequence_length, other_map.sequence_length)
# Use the new-form RateMap
rate_map = msprime.RateMap(positions, rates)
sim = msprime.simulator_factory(10, recombination_map=rate_map)
other_map = msprime.RateMap(**sim.recombination_map)
self.assertEqual(list(other_map.position), positions)
self.assertEqual(list(other_map.rate), rates)
self.assertEqual(sim.sequence_length, other_map.sequence_length)
def test_smc_variants(self):
for model in ["smc", "smc_prime"]:
threshold = 20
sim = msprime.simulator_factory(
sample_size=10,
recombination_rate=5,
model=model,
random_generator=_msprime.RandomGenerator(3),
)
sim.run()
self.assertGreater(sim.num_rejected_common_ancestor_events, 0)
self.assertGreater(sim.num_common_ancestor_events, threshold)
self.assertGreater(sim.num_recombination_events, threshold)
def test_symmetric_growth_rates(self):
# Test a symmetric model where we start with a negative growth
# rate and then increase back to the same value.
Ne = 10001
growth_rate = 0.0125
delta_t = 50
end_size = Ne * math.exp(-growth_rate * delta_t)
population_configurations = [
msprime.PopulationConfiguration(
sample_size=2, initial_size=Ne, growth_rate=growth_rate)]
demographic_events = [
msprime.PopulationParametersChange(
time=delta_t, growth_rate=-growth_rate),
msprime.PopulationParametersChange(
time=2 * delta_t, growth_rate=0)]
simulator = msprime.simulator_factory(
Ne=Ne,
population_configurations=population_configurations,
demographic_events=demographic_events)
ll_sim = simulator.create_ll_instance()
ll_end_time = ll_sim.debug_demography()
t = delta_t
self.assertEqual(t, ll_end_time * 4 * Ne)
populations = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()]
pop = populations[0]
self.assertEqual(pop.growth_rate, growth_rate)
self.assertEqual(pop.initial_size, Ne)
self.assertEqual(pop.get_size(delta_t), end_size)
# Now fast forward to the next time slice.
t += delta_t
ll_end_time = ll_sim.debug_demography()
self.assertEqual(t, ll_end_time * 4 * Ne)
pop = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()][0]
self.assertEqual(pop.growth_rate, -growth_rate)
self.assertEqual(pop.initial_size, end_size)
self.assertEqual(pop.get_size(delta_t), Ne)
# Now fast forward to the next time slice.
ll_end_time = ll_sim.debug_demography()
self.assertTrue(math.isinf(ll_end_time))
populations = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()]
pop = populations[0]
self.assertEqual(pop.growth_rate, 0)
self.assertEqual(pop.initial_size, Ne)
def test_sample_size_population_configuration(self):
for d in range(1, 5):
# Zero sample size is always an error
configs = [msprime.PopulationConfiguration(0) for _ in range(d)]
self.assertRaises(
ValueError, msprime.simulator_factory, population_configurations=configs)
configs = [msprime.PopulationConfiguration(2) for _ in range(d)]
sim = msprime.simulator_factory(population_configurations=configs)
self.assertEqual(len(sim.samples), 2 * d)
samples = []
for j in range(d):
samples += [msprime.Sample(population=j, time=0) for _ in range(2)]
self.assertEqual(sim.samples, samples)
ll_sim = sim.create_ll_instance()
self.assertEqual(ll_sim.get_samples(), samples)
def test_recombination_map(self):
def f(recomb_map):
return msprime.simulator_factory(10, recombination_map=recomb_map)
self.assertRaises(TypeError, f, "wrong type")
for n in range(2, 10):
positions = list(range(n))
rates = [0.1 * j for j in range(n - 1)] + [0.0]
recomb_map = msprime.RecombinationMap(positions, rates)
sim = msprime.simulator_factory(10, recombination_map=recomb_map)
self.assertEqual(sim.recombination_map, recomb_map)
self.assertEqual(recomb_map.get_positions(), positions)
self.assertEqual(recomb_map.get_rates(), rates)
self.assertEqual(sim.num_loci, recomb_map.get_num_loci())
ll_sim = sim.create_ll_instance()
self.assertEqual(ll_sim.get_num_loci(), recomb_map.get_num_loci())
def test_model_instances(self):
for bad_type in [1234, {}]:
self.assertRaises(
TypeError, msprime.simulator_factory, sample_size=2, model=bad_type)
models = [
msprime.StandardCoalescent(),
msprime.SmcApproxCoalescent(),
msprime.SmcPrimeApproxCoalescent(),
msprime.DiscreteTimeWrightFisher(),
msprime.BetaCoalescent(),
msprime.DiracCoalescent(),
]
for model in models:
sim = msprime.simulator_factory(sample_size=10, model=model)
self.assertEqual(sim.model, model)
def test_samples(self):
pop_configs = [
msprime.PopulationConfiguration(),
msprime.PopulationConfiguration(),
msprime.PopulationConfiguration()]
samples = [
msprime.Sample(population=0, time=0),
msprime.Sample(population=1, time=1),
msprime.Sample(population=2, time=2)]
# Ne = 1/4 to keep in coalescence units.
sim = msprime.simulator_factory(
Ne=1/4, samples=samples, population_configurations=pop_configs)
self.assertEqual(sim.samples, samples)
ll_sim = sim.create_ll_instance()
self.assertEqual(ll_sim.get_samples(), samples)
def test_sample_size(self):
self.assertRaises(ValueError, msprime.simulator_factory)
self.assertRaises(ValueError, msprime.simulator_factory, 1)
self.assertRaises(
ValueError, msprime.simulator_factory, sample_size=1)
for n in [2, 100, 1000]:
sim = msprime.simulator_factory(n)
self.assertEqual(len(sim.samples), n)
ll_sim = sim.create_ll_instance()
self.assertEqual(ll_sim.get_num_samples(), n)
samples = ll_sim.get_samples()
self.assertEqual(len(samples), n)
for sample in samples:
self.assertEqual(sample[0], 0)
self.assertEqual(sample[1], 0)
def test_migration_matrix(self):
m = [
[0, 1, 2],
[3, 0, 4],
[5, 6, 0]]
for Ne in [1, 10, 1e6]:
sim = msprime.simulator_factory(
Ne=Ne,
population_configurations=[
msprime.PopulationConfiguration(1),
msprime.PopulationConfiguration(1),
msprime.PopulationConfiguration(1)],
migration_matrix=m)
scaled_m = sim.get_scaled_migration_matrix()
scaled_mp = [
[v * 4 * Ne for v in row] for row in m]
self.assertEqual(scaled_m, scaled_mp)
def test_recombination_rate_scaling(self):
values = [
(10, 0.1, 0.1),
(0.1, 1, 10),
(1e-8, 10**4, 10**8),
(1e-8, 10**5, 10**9),
]
for rate, Ne, length in values:
sim = msprime.simulator_factory(
10, Ne=Ne, recombination_rate=rate, length=length)
num_loci = msprime.RecombinationMap.DEFAULT_NUM_LOCI
total_rate = length * rate
per_locus_rate = total_rate / (num_loci - 1)
# We expect all these rates to be positive.
self.assertGreater(per_locus_rate, 0)
ll_sim = sim.create_ll_instance()
self.assertAlmostEqual(per_locus_rate, ll_sim.get_recombination_rate())
self.assertAlmostEqual(
sim.recombination_map.get_per_locus_recombination_rate(),
per_locus_rate)
def __init__(
self, sample_size=1, num_loci=1, scaled_recombination_rate=0,
num_replicates=1, migration_matrix=None,
population_configurations=None, demographic_events=None,
scaled_mutation_rate=0, print_trees=False,
precision=3, random_seeds=None):
self._sample_size = sample_size
self._num_loci = num_loci
self._num_replicates = num_replicates
self._recombination_rate = scaled_recombination_rate
self._mutation_rate = scaled_mutation_rate
# For strict ms-compability we want to have m non-recombining loci
recomb_map = msprime.RecombinationMap.uniform_map(
num_loci, self._recombination_rate, num_loci)
# If we have specified any population_configurations we don't want
# to give the overall sample size.
sample_size = self._sample_size
if population_configurations is not None:
sample_size = None
# msprime measure's time in units of generations, given a specific
# Ne value whereas ms uses coalescent time. To be compatible with ms,
# we therefore need to use an Ne value of 1/4.
self._simulator = msprime.simulator_factory(
Ne=0.25,
sample_size=sample_size,
recombination_map=recomb_map,
population_configurations=population_configurations,
migration_matrix=migration_matrix,
demographic_events=demographic_events)
self._precision = precision
self._print_trees = print_trees
# sort out the random seeds
ms_seeds = random_seeds
if random_seeds is None:
ms_seeds = generate_seeds()
seed = get_single_seed(ms_seeds)
self._random_generator = msprime.RandomGenerator(seed)
self._ms_random_seeds = ms_seeds
self._simulator.random_generator = self._random_generator
self._mutation_generator = msprime.MutationGenerator(
self._random_generator, self._mutation_rate)
def __init__(
self, sample_size=1, num_loci=1, scaled_recombination_rate=0,
num_replicates=1, migration_matrix=None,
population_configurations=None, demographic_events=None,
scaled_mutation_rate=0, print_trees=False,
precision=3, random_seeds=None):
self._sample_size = sample_size
self._num_loci = num_loci
self._num_replicates = num_replicates
# We use unscaled per-generation rates. By setting Ne = 1 we
# don't need to rescale, but we still need to divide by 4 to
# cancel the factor introduced when calculated the scaled rates.
self._recombination_rate = scaled_recombination_rate / 4
self._mutation_rate = scaled_mutation_rate / 4
# For strict ms-compability we want to have m non-recombining loci
recomb_map = msprime.RecombinationMap.uniform_map(
num_loci, self._recombination_rate, num_loci)
# If we have specified any population_configurations we don't want
# to give the overall sample size.
sample_size = self._sample_size
if population_configurations is not None:
sample_size = None
self._simulator = msprime.simulator_factory(
sample_size=sample_size,
recombination_map=recomb_map,
population_configurations=population_configurations,
migration_matrix=migration_matrix,
demographic_events=demographic_events)
self._precision = precision
self._print_trees = print_trees
# sort out the random seeds
ms_seeds = random_seeds
if random_seeds is None:
ms_seeds = generate_seeds()
seed = get_single_seed(ms_seeds)
self._random_generator = msprime.RandomGenerator(seed)
self._ms_random_seeds = ms_seeds
self._simulator.set_random_generator(self._random_generator)
def test_single_growth_rate_size_change(self):
# Set out our values in units of generations and absolute sizes.
Ne = 1000
growth_rate = -0.01
end_time = 20
end_size = Ne * math.exp(-growth_rate * end_time)
new_size = 4 * Ne
population_configurations = [
msprime.PopulationConfiguration(
sample_size=2, initial_size=Ne, growth_rate=growth_rate)]
demographic_events = [
msprime.PopulationParametersChange(
time=end_time, initial_size=new_size, growth_rate=0)]
simulator = msprime.simulator_factory(
Ne=Ne,
population_configurations=population_configurations,
demographic_events=demographic_events)
ll_sim = simulator.create_ll_instance()
ll_end_time = ll_sim.debug_demography()
self.assertEqual(end_time, ll_end_time * 4 * Ne)
populations = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()]
self.assertEqual(len(populations), 1)
pop = populations[0]
self.assertEqual(pop.growth_rate, growth_rate)
self.assertEqual(pop.initial_size, Ne)
self.assertEqual(pop.get_size(end_time), end_size)
# Now fast forward to the next time slice.
ll_end_time = ll_sim.debug_demography()
self.assertTrue(math.isinf(ll_end_time))
populations = [
msprime.Population(Ne=Ne, **d)
for d in ll_sim.get_population_configuration()]
pop = populations[0]
self.assertEqual(pop.growth_rate, 0)
self.assertEqual(pop.initial_size, new_size)
self.assertEqual(pop.get_size(10), new_size)
def f(Ne):
return msprime.simulator_factory(10, Ne=Ne)
def f(configs):
return msprime.simulator_factory(population_configurations=configs)
def f(hl_matrix):
return msprime.simulator_factory(
population_configurations=pop_configs,
migration_matrix=hl_matrix)
def test_default_migration_matrix(self):
sim = msprime.simulator_factory(10)
ll_sim = sim.create_ll_instance()
self.assertEqual(ll_sim.get_migration_matrix(), [0.0])
def f(recomb_rate):
return msprime.simulator_factory(10, recombination_rate=recomb_rate)
def f(recomb_map):
return msprime.simulator_factory(10, recombination_map=recomb_map)
def test_length(self):
for bad_length in [-1, 0, -1e-6]:
with self.assertRaises(ValueError):
msprime.simulator_factory(10, length=bad_length)
def test_random_seed(self):
seed = 12345
rng = msprime.RandomGenerator(seed)
sim = msprime.simulator_factory(10, random_generator=rng)
self.assertEqual(rng, sim.random_generator)
self.assertEqual(rng.get_seed(), seed)