def cryptic_parameters(id_map, labeled_nodes, related_pairs):
# TODO: Add argument for this
recomb_dir = abspath(
join(dirname(__file__), "../data/recombination_rates/"))
cm_data = centimorgan_data_from_directory(recomb_dir)
ibd_detector = SharedSegmentDetector(0, 5, cm_data)
# We try to only include the labeled nodes the analyst would have
# access to This only works if we use the deterministic random
# argument when we evaluate the classifier.
labeled_copy = sorted(labeled_nodes)
rand_state = getstate()
seed(42)
shuffle(labeled_copy)
setstate(rand_state)
labeled_node_pairs = set(combinations(labeled_copy[:1000], 2))
related_pairs = set(related_pairs)
cryptic_pairs = set(x for x in labeled_node_pairs
if x not in related_pairs)
lengths = []
for node_a_id, node_b_id in cryptic_pairs:
node_a = id_map[node_a_id]
node_b = id_map[node_b_id]
genome_a = node_a.genome
genome_b = node_b.genome
length = ibd_detector.shared_segment_length(genome_a, genome_b)
lengths.append(length)
np_lengths = np.array(lengths, dtype=np.uint64)
params = fit_hurdle_gamma(np_lengths)
assert all(x is not None for x in params)
return params
def distributions_from_directory(directory, id_mapping):
"""
Calculate distributions from a directory created by
calculate_shared_to_directory.
"""
distributions = dict()
for labeled_filename in tqdm(listdir(directory)):
lengths = defaultdict(list)
labeled = int(labeled_filename)
with open(join(directory, labeled_filename), "r") as labeled_file:
for line in labeled_file:
# If the program crashed, the output can be left in an
# inconsistent state.
try:
unlabeled_id, shared_str = line.split("\t")
except ValueError:
warn("Malformed line:\n{}".format(line), stacklevel=0)
continue
unlabeled = int(unlabeled_id)
if unlabeled not in id_mapping:
error_string = "No such unlabeled node with id {}."
warn(error_string.format(unlabeled_id), stacklevel=0)
continue
try:
shared_float = float(shared_str)
except ValueError:
error_string = "Error formatting value as float: {}."
warn(error_string.format(shared_str), stacklevel=0)
continue
lengths[unlabeled].append(shared_float)
for unlabeled, lengths in lengths.items():
shape, scale, zero_prob = fit_hurdle_gamma(
np.array(lengths, dtype=np.float64))
if shape is None:
continue
shape, scale = adjust_shape_scale(shape, scale)
params = HurdleGammaParams(shape, scale, zero_prob)
distributions[unlabeled, labeled] = params
return distributions
if args.cm_ibd_threshold > 0:
cur_path = realpath(__file__)
parent = split(split(cur_path)[0])[0]
rates_dir = join(parent, "data", "recombination_rates")
print("Loading recombination data for centimorgan cutoff.", flush = True)
recomb_data = centimorgan_data_from_directory(rates_dir)
ibd_detector = SharedSegmentDetector(0, 5, recomb_data)
else:
ibd_detector = SharedSegmentDetector(5000000)
labeled_nodes = classifier._labeled_nodes
labeled_node_pairs = set(combinations(labeled_nodes, 2))
related_pairs = set(classifier._distributions.keys())
cryptic_pairs = set(x for x in combinations(labeled_nodes, 2)
if x not in related_pairs)
print("Calculating IBD for pairs.")
lengths = []
id_map = population.id_mapping
for node_a_id, node_b_id in progressbar(cryptic_pairs):
node_a = id_map[node_a_id]
node_b = id_map[node_b_id]
genome_a = node_a.genome
genome_b = node_b.genome
length = ibd_detector.shared_segment_length(genome_a, genome_b)
lengths.append(length)
np_lengths = np.array(lengths, dtype = np.uint64)
print(fit_hurdle_gamma(np_lengths))
