def edges_as_df(
*circuits: Circuit,
sanitise=False) -> Tuple[pd.DataFrame, Dict[Circuit, pd.DataFrame]]:
"""
circuit, pre_id, pre_name, pre_side, pre_segment, post_id, post_name, post_side, post_segment, contact_number, synaptic_area
:param circuits:
:return:
"""
if not circuits:
circuits = sorted(Circuit, key=lambda x: str(x).lower())
headers = ("circuit", "pre_id", "pre_name", "pre_side", "pre_segment",
"post_id", "post_name", "post_side", "post_segment",
"contact_number", "synaptic_area")
all_data = []
per_circuit = dict()
for circuit in circuits:
g = multidigraph_to_digraph(get_data(circuit))
rows = []
total_count = 0
for pre, post, edata in iter_data(g):
row = [circuit]
for node in (pre, post):
row.extend([node.id, node.name, node.side, node.segment])
row.append(edata["count"])
row.append(edata["area"])
assert len(row) == len(headers)
if sanitise:
row = sanitise_row(row)
rows.append(row)
total_count += edata["count"]
logger.info(f"Total synapse count for %s: %s", circuit, total_count)
rows = sorted(rows)
per_circuit[circuit] = pd.DataFrame(data=rows, columns=headers)
all_data.extend(rows)
return pd.DataFrame(data=all_data, columns=headers), per_circuit
def __init__(self, graph: nx.MultiDiGraph, name=None):
"""
Note: this is made more complicated by:
- some ambiguous chordotonals (e.g. lch5-2/4)
- resolved by summing their counts/areas
- some LNs which do not have a side (e.g. Ladder)
- Resolved by defining edge side by the postsynaptic partner
Assumes that only presynaptic neurons can be ambiguous,
and that that all postsynaptic partners have a side,
and that there are no contralateral edges.
"""
super().__init__(graph, name)
self.graph = multidigraph_to_digraph(graph)
# self.graph, self.filtered_nodes = self._filter_unilateral_nodes()
# self.graph, self.filtered_edges = self._filter_unilateral_edges()
self.name = name
def df_to_graph(detected_df: pd.DataFrame) -> nx.DiGraph:
g = nx.MultiDiGraph()
syn_to_skels = defaultdict(list)
for row in detected_df.itertuples():
syn_to_skels[row.synapse_object_id].append(row)
if row.skeleton_id not in g.nodes:
g.add_node(row.skeleton_id, synapses=dict())
g.node[row.skeleton_id]["synapses"][row.synapse_object_id] = (
row.contact_px * PX_AREA)
for this_skid, synapses in g.nodes(data="synapses"):
for synapse_id, contact_px in synapses.items():
for row in syn_to_skels[synapse_id]:
if row.skeleton_id != this_skid:
g.add_edge(
this_skid,
row.skeleton_id,
area=row.contact_px * PX_AREA,
synapse_object_id=synapse_id,
pre_area=contact_px * PX_AREA,
)
return multidigraph_to_digraph(g)
def __init__(self, graph: nx.MultiDiGraph, name=""):
super().__init__(graph, name)
self.graph = multidigraph_to_digraph(self.graph)
done_edges.add((pre, post))
done_edges.add((post, pre))
erroneous_edges = set()
for pre_post in g_auto.edges:
if pre_post not in done_edges:
erroneous_edges.add(pre_post)
print(f"{len(erroneous_edges)} erroneous edges found: ")
for pre, post in sorted(erroneous_edges):
print(f"\t{pre} -> {post}")
fig, ax_arr = plt.subplots(1, 2)
count_ax, area_ax = ax_arr.flatten()
manual_auto_scatter(manual_counts, auto_counts, "Contact number", count_ax)
manual_auto_scatter(manual_areas, auto_areas, "Contact area", area_ax)
if __name__ == "__main__":
catmaid = get_catmaid()
multi_g = get_data("ORN-PN")
g_manual = multidigraph_to_digraph(multi_g)
skids = list(g_manual.nodes)
detected_df = catmaid.get_detected_synapses_between(*skids)
g_auto = df_to_graph(detected_df)
compare_edges(g_manual, g_auto)
plt.show()
data_dir = TABLES_DIR / "out" / "count_frac_area"
data_dir.mkdir(parents=True, exist_ok=True)
HEADERS = ("circuit", "pre_id", "pre_name", "pre_side", "pre_segment",
"post_id", "post_name", "post_side", "post_segment",
"contact_number", "contact_fraction", "synaptic_area")
all_data = []
tgt_total_counts_df = pd.read_csv(CATMAID_CSV_DIR /
"dendritic_synapse_counts.csv",
index_col=0)
for circuit in Circuit:
g = multidigraph_to_digraph(get_data(circuit))
rows = []
total_count = 0
for pre, post, edata in iter_data(g):
row = [str(circuit)]
for node in (pre, post):
row.extend([node.id, node.name, str(node.side), str(node.segment)])
row.append(edata["count"])
frac = edata["count"] / tgt_total_counts_df["post_count"].loc[post.id]
row.append(frac)
row.append(edata["area"])
rows.append(row)