def test_design_with_overhang_primers(repeat, span_cost):
goal = random_record(3000)
make_circular_and_id([goal])
r1 = random_record(100) + goal[1000:2000] + random_record(100)
p1 = goal[970:1030]
p2 = goal[1970:2030].reverse_complement()
r2 = goal[2000:] + goal[:1000]
p3 = goal[1970:2030]
make_linear_and_id([r1, p1, p2, r2, p3])
design = Design(span_cost)
design.add_materials(primers=[p1, p2, p3],
templates=[r1, r2],
queries=[goal],
fragments=[])
expected_path = [
(970, False, "A", True),
(2030, False, "B", True),
(1970, False, "A", True),
(4000, True, "B", True),
]
check_design_result(design, expected_path)
def test_design_with_overlaps_with_templates(span_cost):
"""Fragments with overlaps."""
goal = random_record(3000)
make_circular_and_id([goal])
r1 = goal[-40:] + goal[:1000]
r2 = goal[970:2000]
r3 = goal[1950:]
p1 = goal[-40:]
make_linear_and_id([r1, r2, r3, p1])
design = Design(span_cost)
design.add_materials(primers=[p1],
fragments=[],
queries=[goal],
templates=[r1, r2, r3])
expected_path = [
(970, True, "A", True),
(1950, True, "B", False),
(1950, True, "A", False),
(3000, True, "B", True),
(3000 - 40, False, "A", True),
(4000, True, "B", True),
]
check_design_result(design, expected_path, check_path=True)
def test_design_task_with_gaps(span_cost):
"""Fragments with overlaps."""
goal = random_record(3000)
make_circular_and_id([goal])
r1 = goal[:950]
r2 = goal[1000:2000]
r3 = goal[2050:]
make_linear_and_id([r1, r2, r3])
design = Design(span_cost)
design.add_materials(primers=[],
templates=[r1, r2, r3],
queries=[goal],
fragments=[])
expected_path = [
(0, True, "A", False),
(950, True, "B", False),
(1000, True, "A", False),
(2000, True, "B", False),
(2050, True, "A", False),
(3000, True, "B", False),
]
check_design_result(design, expected_path)
def test_single_fragment(span_cost):
goal = random_record(3000)
make_circular_and_id([goal])
r1 = goal[177:2255]
make_linear_and_id([r1])
design = Design(span_cost)
design.add_materials(primers=[],
templates=[r1],
queries=[goal],
fragments=[])
expected_path = [(177, True, "A", False), (2255, True, "B", False)]
check_design_result(design, expected_path)
def test_a_reverse_pcr_fragment(span_cost):
goal = random_record(3000)
make_circular_and_id([goal])
t1 = goal[1000:2500].reverse_complement()
p1 = goal[2500 - 20:2510].reverse_complement()
make_linear_and_id([p1, t1])
design = Design(span_cost)
design.add_materials(primers=[p1],
templates=[t1],
queries=[goal],
fragments=[])
expected_path = [(1000, True, "A", False), (2510, False, "B", False)]
check_design_result(design, expected_path)
def test_fully_overlapped(span_cost):
goal = random_record(2000)
make_circular_and_id([goal])
r1 = goal[1100:1300]
p1 = goal[1177:1177 + 30]
p2 = goal[1188:1188 + 30]
p3 = goal[1225 - 30:1225].reverse_complement()
make_linear_and_id([r1, p1, p2, p3])
design = Design(span_cost)
design.add_materials(primers=[p1, p2, p3],
templates=[r1],
queries=[goal],
fragments=[])
expected_path = [(1177, False, "A", False), (1300, True, "B", False)]
check_design_result(design, expected_path)
def test_blast_has_same_results(span_cost):
goal = random_record(3000)
make_circular_and_id([goal])
r1 = random_record(100) + goal[1000:2000] + random_record(100)
p1 = goal[970:1030]
p2 = goal[1970:2030].reverse_complement()
r2 = goal[2000:] + goal[:1000]
p3 = goal[1970:2030]
make_linear_and_id([r1, p1, p2, r2, p3])
size_of_groups = []
for i in range(20):
design = Design(span_cost)
design.logger.set_level("INFO")
design.add_materials(primers=[p1, p2, p3],
templates=[r1, r2],
queries=[goal],
fragments=[])
design.compile()
for container in design.container_list:
size_of_groups.append(len(container.groups()))
assert len(size_of_groups) == 20
assert len(set(size_of_groups)) == 1
def test_num_groups_vs_endpoints(here, paths, query, span_cost):
primers = make_linear(load_fasta_glob(paths["primers"]))
templates = load_genbank_glob(paths["templates"])
query_path = join(here, "data/test_data/genbank/designs", query)
queries = make_circular(load_genbank_glob(query_path))
design = Design(span_cost)
design.add_materials(primers=primers, templates=templates, queries=queries)
design._blast()
containers = design.container_list
assert len(containers) == 1
container = containers[0]
container.expand()
groups = container.groups()
print(len(groups)**2)
a_arr = set()
b_arr = set()
for g in groups:
a_arr.add(g.query_region.a)
b_arr.add(g.query_region.b)
print(len(a_arr) * len(b_arr))
def test_add_special_partition_node(self, span_cost):
"""This test adds a new unique node 'n3' with a unique type to simulate
adding a partitioning to the graph.
Such a procedure might be used for highly complex sequences.
"""
goal = random_record(4000)
make_circular_and_id([goal])
r1 = goal[1000:2000]
r2 = goal[200:500]
make_linear_and_id([r1, r2])
design = Design(span_cost)
design.add_materials(
primers=[], templates=[r1, r2], queries=[goal], fragments=[]
)
design.compile()
import networkx as nx
for qk, g in design.graphs.items():
query = design.seqdb[qk]
gcopy = nx.DiGraph(g)
for n1, n2, edata in g.edges(data=True):
r = Region(n1.index, n2.index, len(query.seq), cyclic=True)
if n1.type == "B" and n2.type == "A":
# index = int((n1.index + n2.index) / 2)
delta = int(len(r) / 2)
index = r.t(delta + n1.index)
n3 = AssemblyNode(index, False, str(uuid4()), overhang=True)
edata1 = dict(edata)
edata2 = dict(edata)
edata1["material"] = edata["material"] / 10.0
edata2["material"] = edata["material"] / 10.0
edata1["span"] = 0
gcopy.add_edge(n1, n3, **edata1)
gcopy.add_edge(n3, n2, **edata2)
design.graphs[qk] = gcopy
result = list(design.optimize().values())[0]
assembly = result.assemblies[0]
df = assembly.to_df()
assert list(df["query_start"]) == [200, 500, 750, 1000, 2000, 3100]
assert list(df["query_end"]) == [500, 750, 1000, 2000, 3100, 200]
def test_requires_synthesis_with_template_over_origin(span_cost):
goal = random_record(5000)
make_circular_and_id([goal])
r1 = goal[1000:2000]
r2 = goal[3500:] + goal[:500]
make_linear_and_id([r1, r2])
design = Design(span_cost)
design.add_materials(primers=[],
templates=[r1, r2],
queries=[goal],
fragments=[])
expected_path = [
(500, True, "B", False),
(1000, True, "A", False),
(2000, True, "B", False),
(3500, True, "A", False),
]
check_design_result(design, expected_path)
def test_case(span_cost):
"""This is a test case which has previously failed to find a solution.
The case is that there are just two small fragments with a small
<10bp gap. The solution should be to PCR amplify both fragment and
synthesize the rest of the plasmid.
"""
goal = random_record(2000)
make_circular_and_id([goal])
r1 = goal[1188:1230]
r2 = goal[1238:1282]
make_linear_and_id([r1, r2])
design = Design(span_cost)
design.add_materials(primers=[],
templates=[r1, r2],
queries=[goal],
fragments=[])
expected_path = [(1238, True, "A", False), (1282, True, "B", False)]
check_design_result(design, expected_path)
def test_very_long_synthesizable_region(span_cost):
goal = random_record(10000)
make_circular_and_id([goal])
r1 = goal[4177:4255]
r2 = goal[4188:4225]
make_linear_and_id([r1, r2])
design = Design(span_cost)
design.add_materials(primers=[],
templates=[r1],
queries=[goal],
fragments=[])
expected_path = [
(500, True, "B", False),
(1000, True, "A", False),
(2000, True, "B", False),
(2500, True, "A", False),
]
with pytest.raises(NoSolution):
check_design_result(design, expected_path)
def test_large_pkl(span_cost):
"""Expect more than one graph to be output if multiple queries are
provided."""
design = Design.fake(n_designs=3,
n_cyclic_seqs=100,
n_linear_seqs=100,
n_primers=100)
design.compile()
with logger.timeit("DEBUG", "pickling graphs"):
pickle.loads(pickle.dumps(design.graphs))
with logger.timeit("DEBUG", "pickling containers"):
pickle.loads(pickle.dumps(design.container_factory))
with logger.timeit("DEBUG", "pickling span_cost"):
pickle.loads(pickle.dumps(span_cost))
def test_design_near_origin(span_cost):
"""Fragments with overlaps."""
goal = random_record(3000)
make_circular_and_id([goal])
r1 = goal[-40:] + goal[:1000]
r3 = goal[950:] + goal[:1]
make_linear_and_id([r1, r3])
design = Design(span_cost)
design.add_materials(primers=[],
fragments=[r1, r3],
queries=[goal],
templates=[])
design.run()
df = design.to_df()[1]
print(df)
def test_cost_comparison_library():
design1, library = Design.fake(
n_designs=3,
n_linear_seqs=50,
n_cyclic_seqs=50,
n_primers_from_templates=500,
shared_length=500,
return_with_library=True,
)
design2 = LibraryDesign(seqdb=design1.seqdb)
designs = library["design"]
plasmids = library["cyclic"]
fragments = library["linear"]
primers = library["short"]
design2.add_materials(primers=primers,
fragments=fragments,
templates=plasmids,
queries=designs)
design1.run()
design2.run()
print("#" * 10 + "\nDesign\n" + "#" * 10)
print(json.dumps(design1.status, indent=2))
print("#" * 10 + "\nLibraryDesign\n" + "#" * 10)
print(json.dumps(design2.status, indent=2))
print("%" * 10 + "\nDesign Cost\n" + "%" * 10)
for qk, s in design1.status.items():
print(s["assemblies"])
print("%" * 10 + "\nLibraryDesign Cost\n" + "%" * 10)
for qk, s in design2.status.items():
print(s["assemblies"])
design2.report().plot_coverage(show=True)
print(design2.to_df()[1])
def test_reindex_invariant(reindex):
design1, library = Design.fake(
n_designs=1,
n_linear_seqs=50,
n_cyclic_seqs=50,
n_primers_from_templates=500,
shared_length=500,
return_with_library=True,
)
designs = library["design"]
plasmids = library["cyclic"]
fragments = library["linear"]
primers = library["short"]
new_designs = [design[reindex:] + design[:reindex] for design in designs]
make_cyclic(new_designs)
design2 = Design()
design2.add_materials(templates=plasmids,
fragments=fragments,
primers=primers,
queries=new_designs)
design1.run()
design2.run()
results1 = design1.out()
results2 = design2.out()
assemblies1 = list(results1["designs"].values())[0]["assemblies"][0]
assemblies2 = list(results2["designs"].values())[0]["assemblies"][0]
print(assemblies1["cost"])
print(assemblies2["cost"])
assert assemblies1["cost"] == assemblies2["cost"]
def example_design(self, span_cost):
goal = random_record(3000)
make_circular_and_id([goal])
r1 = random_record(100) + goal[1000:2000] + random_record(100)
p1 = goal[970:1030]
p2 = goal[1970:2030].reverse_complement()
r2 = goal[2000:] + goal[:1000]
p3 = goal[1970:2030]
make_linear_and_id([r1, p1, p2, r2, p3])
design = Design(span_cost)
design.add_materials(primers=[p1, p2, p3],
templates=[r1, r2],
queries=[goal],
fragments=[])
design.compile()
results = design.optimize()
return design, results
def test_design_optimize_cannot_run_before_compile(cached_span_cost):
design = Design.fake(n_designs=1)
with pytest.raises(DasiDesignException):
design.optimize()
design.compile()
design.optimize()
def test_example_0(args):
i0, i1, DesignCls = args
random.seed(0)
np.random.seed(0)
def open_gb(path):
with open(join(fixtures, path)) as f:
return list(SeqIO.parse(f, format="genbank"))
fragments = open_gb("fragments_0.gb")
primers = open_gb("primers_0.gb")
plasmids = open_gb("plasmids_0.gb")
goals = open_gb("goals_0.gb")
design = Design()
design.add_fragments(fragments)
design.add_primers(primers)
design.add_templates(plasmids)
s = slice(i0, i1)
design.add_queries(goals[s])
design.run(n_jobs=4)
# assert successful runs
print(design.status)
for v in design.status.values():
assert v["success"] is True
# output JSON
out = design.out()
print(out)
print(design.to_df()[1])
def run(self, n_jobs: int = 10):
"""Run a design job.
:param n_jobs: number of parrallel jobs to run. (default: 10)
:return:
"""
import warnings
warnings.simplefilter(action="ignore", category=RuntimeWarning)
warnings.simplefilter(action="ignore", category=BiopythonParserWarning)
self._logger.info("Loading sequence files")
primers = make_linear(load_fasta_glob(self._primers))
templates = make_circular(load_genbank_glob(self._templates))
fragments = make_linear(load_genbank_glob(self._fragments))
goals = make_circular(load_genbank_glob(self._goals))
design = Design()
design.n_jobs = n_jobs
design.add_materials(primers=primers,
templates=templates,
fragments=fragments,
queries=goals)
self._logger.info("Getting span cost model")
span_cost = self._get_span_cost()
design.span_cost = span_cost
self._logger.info("Compiling possible molecular assemblies")
design.compile()
self._logger.info("Optimizing molecular assemblies")
design.optimize()
self._logger.info("Designing assembly primers and fragments")
df, adf, design_json = design.to_df()
adf.to_csv("summary.csv")
df.to_csv("sequence_design.csv")
records = []
for result in design.results.values():
if result.assemblies:
a = result.assemblies[0]
for i, role, m in a.molecules:
records.append(m.sequence)
SeqIO.write(records, os.path.join(self._directory, "sequences.gb"),
"genbank")
def design(self):
d = Design.fake(n_designs=3)
d.run()
return d