def poa(groups, max_sequences_per_poa=100, gpu_mem_per_batch=0.9):
"""
Generate consensus for POA groups.
Args:
groups : A list of lists of sequences for which consensus is to be generated.
"""
free, total = cuda.cuda_get_mem_info(cuda.cuda_get_device())
gpu_mem_per_batch *= free
batch = CudaPoaBatch(max_sequences_per_poa,
gpu_mem_per_batch,
stream=None,
output_type="consensus")
results = []
for i, group in enumerate(groups, start=1):
group_status, seq_status = batch.add_poa_group(group)
# Once batch is full, run POA processing
if group_status == 1 or i == len(groups):
batch.generate_poa()
consensus, coverage, status = batch.get_consensus()
results.extend(consensus)
batch.reset()
group_status, seq_status = batch.add_poa_group(group)
return results
def test_cudapoa_simple_batch():
free, total = cuda.cuda_get_mem_info(cuda.cuda_get_device())
batch = CudaPoaBatch(10, 0.9 * free)
poa_1 = ["ACTGACTG", "ACTTACTG", "ACGGACTG", "ATCGACTG"]
poa_2 = ["ACTGAC", "ACTTAC", "ACGGAC", "ATCGAC"]
batch.add_poa_group(poa_1)
batch.add_poa_group(poa_2)
batch.generate_poa()
consensus, coverage, status = batch.get_consensus()
assert(len(consensus) == 2)
assert(batch.total_poas == 2)
def test_cudapoa_reset_batch():
device = cuda.cuda_get_device()
free, total = cuda.cuda_get_mem_info(device)
batch = CudaPoaBatch(10, 0.9 * free, device_id=device)
poa_1 = ["ACTGACTG", "ACTTACTG", "ACGGACTG", "ATCGACTG"]
batch.add_poa_group(poa_1)
batch.generate_poa()
consensus, coverage, status = batch.get_consensus()
assert(batch.total_poas == 1)
batch.reset()
assert(batch.total_poas == 0)
def test_cudapoa_complex_batch():
random.seed(2)
read_len = 500
ref = ''.join([random.choice(['A', 'C', 'G', 'T']) for _ in range(read_len)])
num_reads = 100
mutation_rate = 0.02
reads = []
for _ in range(num_reads):
new_read = ''.join([r if random.random() > mutation_rate else random.choice(['A', 'C', 'G', 'T']) for r in ref])
reads.append(new_read)
device = cuda.cuda_get_device()
free, total = cuda.cuda_get_mem_info(device)
stream = cuda.CudaStream()
batch = CudaPoaBatch(1000, 0.9 * free, stream=stream, device_id=device)
(add_status, seq_status) = batch.add_poa_group(reads)
batch.generate_poa()
consensus, coverage, status = batch.get_consensus()
consensus = consensus[0]
assert(len(consensus) == len(ref))
match_ratio = SequenceMatcher(None, ref, consensus).ratio()
assert(match_ratio == 1.0)
def test_cudapoa_simple_batch():
device = cuda.cuda_get_device()
free, total = cuda.cuda_get_mem_info(device)
batch = CudaPoaBatch(10,
1024,
0.9 * free,
deivce_id=device,
output_mask='consensus')
poa_1 = ["ACTGACTG", "ACTTACTG", "ACGGACTG", "ATCGACTG"]
poa_2 = ["ACTGAC", "ACTTAC", "ACGGAC", "ATCGAC"]
batch.add_poa_group(poa_1)
batch.add_poa_group(poa_2)
batch.generate_poa()
consensus, coverage, status = batch.get_consensus()
assert (len(consensus) == 2)
assert (batch.total_poas == 2)
def test_cudapoa_graph():
device = cuda.cuda_get_device()
free, total = cuda.cuda_get_mem_info(device)
batch = CudaPoaBatch(10, 0.9 * free, device_id=device)
poa_1 = ["ACTGACTG", "ACTTACTG", "ACTCACTG"]
batch.add_poa_group(poa_1)
batch.generate_poa()
consensus, coverage, status = batch.get_consensus()
assert (batch.total_poas == 1)
# Expected graph
# - -> G -> -
# | |
# A -> C -> T -> T -> A -> C -> T -> G
# | |
# - -> C -> -
graphs, status = batch.get_graphs()
assert (len(graphs) == 1)
digraph = graphs[0]
assert (digraph.number_of_nodes() == 10)
assert (digraph.number_of_edges() == 11)