import csv
import logging
from owmeta_core.datasource import GenericTranslation
from owmeta_core.dataobject import ObjectProperty
from owmeta_core.data_trans.csv_ds import CSVDataTranslator, CSVDataSource
from .. import CONTEXT
from ..utils import normalize_cell_name
from ..connection import Connection
from ..cell import Cell
from ..document import Document
from ..evidence import Evidence
from ..neuron import Neuron
from ..muscle import Muscle, BodyWallMuscle
from ..worm import Worm
from ..network import Network
from .common_data import DSMixin, DTMixin
from .data_with_evidence_ds import DataWithEvidenceDataSource
L = logging.getLogger(__name__)
[docs]class ConnectomeCSVDataSource(DSMixin, CSVDataSource):
'''
A CSV data source whose CSV file describes a neural connectome
Basically, this is just a marker type to indicate what's described in the CSV --
there's no consistent schema
'''
class_context = CONTEXT
[docs]class NeuronConnectomeCSVTranslation(GenericTranslation):
class_context = CONTEXT
neurons_source = ObjectProperty(value_type=DataWithEvidenceDataSource)
muscles_source = ObjectProperty(value_type=DataWithEvidenceDataSource)
key_properties = (GenericTranslation.source, muscles_source, neurons_source)
[docs]class NeuronConnectomeCSVTranslator(DTMixin, CSVDataTranslator):
class_context = CONTEXT
input_type = (ConnectomeCSVDataSource, DataWithEvidenceDataSource)
output_type = DataWithEvidenceDataSource
translation_type = NeuronConnectomeCSVTranslation
[docs] def make_translation(self, sources):
tr = super(NeuronConnectomeCSVTranslator, self).make_translation()
tr.source(sources[0])
tr.neurons_source(sources[1])
tr.muscles_source(sources[2])
return tr
[docs] def translate(self, data_source, neurons_source, muscles_source):
# muscle cells that are generically defined in source and need to be broken
# into pair of L and R before being added to owmeta
# muscle cells that have different names in connectome source and cell list.
# Their wormbase cell list names will be used in owmeta
changed_muscles = ['ANAL', 'INTR', 'INTL', 'SPH']
# counters for terminal printing
neuron_connections = 0
muscle_connections = 0
other_connections = 0
res = self.make_new_output(sources=(data_source, neurons_source, muscles_source))
# XXX: should there be a different src for muscles?
w_q = muscles_source.data_context.stored(Worm)()
n_q = neurons_source.data_context.stored(Network)()
i_n = next(n_q.load(), n_q)
i_w = next(w_q.load(), w_q)
# XXX: If the context we query for is the same ID as the default
# context, then we should query the merge of all graphs (using the
# contract of the ConjunctiveGraph with 'default_is_union==True').
# Otherwise, we should query only for what's in that graph.
#
neuron_objs = list(set(i_n.neurons()))
muscle_objs = list(i_w.muscles())
# get lists of neuron and muscles names
neurons = set(neuron.name() for neuron in neuron_objs)
muscles = set(muscle.name() for muscle in muscle_objs)
o_w = res.data_context(Worm)()
o_n = res.data_context(Network)(worm=o_w)
# Evidence object to assert each connection
ctxd_Document = res.evidence_context(Document)
doc = ctxd_Document(key="emmons2015",
title='Whole-animal C. elegans connectomes',
year=2015,
uri='http://abstracts.genetics-gsa.org/cgi-bin/'
'celegans15s/wsrch15.pl?author=emmons&sort=ptimes&'
'sbutton=Detail&absno=155110844&sid=668862',
rdfs_comment="Data related by personal communication")
doc.author('Emmons, S.')
doc.author('Cook, S.')
doc.author('Jarrell, T.')
doc.author('Wang, Y.')
doc.author('Yakolev, M.')
doc.author('Nguyen, K.')
doc.author('Hall, D.')
docctx = res.data_context_for(document=doc)
res.evidence_context(Evidence)(key="emmons2015",
reference=doc,
supports=docctx.rdf_object)
with docctx(Neuron, Muscle, BodyWallMuscle, Cell, Connection) as ctx:
res.data_context.add_import(ctx.context)
with open(data_source.full_path()) as csvfile:
edge_reader = csv.reader(csvfile)
next(edge_reader) # skip header row
for row in edge_reader:
source, target, weight, syn_type = map(str.strip, row)
# set synapse type to something the Connection object
# expects, and normalize the source and target names
if syn_type == 'electrical':
syn_type = 'gapJunction'
elif syn_type == 'chemical':
syn_type = 'send'
source_is_bwm = 'BWM' in source.upper()
target_is_bwm = 'BWM' in target.upper()
source = normalize_cell_name(source).upper()
target = normalize_cell_name(target).upper()
weight = int(weight)
# change certain muscle names to names in wormbase
if source in changed_muscles:
source = changed_muscle(source)
if target in changed_muscles:
target = changed_muscle(target)
sources = convert_to_cell(ctx, source, muscles, neurons, source_is_bwm)
targets = convert_to_cell(ctx, target, muscles, neurons, target_is_bwm)
for s in sources:
for t in targets:
conn = add_synapse(ctx, s, t, weight, syn_type)
o_n.synapse(conn)
kind = conn.termination.onedef()
if kind == 'muscle':
muscle_connections += 1
elif kind == 'neuron':
neuron_connections += 1
else:
other_connections += 1
print('Total neuron to neuron connections added = %i' % neuron_connections)
print('Total neuron to muscle connections added = %i' % muscle_connections)
print('Total other connections added = %i' % other_connections)
print('uploaded connections')
return res
[docs]class NeuronConnectomeSynapseClassTranslation(GenericTranslation):
class_context = CONTEXT
neurotransmitter_source = ObjectProperty()
key_properties = (GenericTranslation.source, neurotransmitter_source)
[docs]class NeuronConnectomeSynapseClassTranslator(DTMixin, CSVDataTranslator):
'''
Adds synapse classes to existing connections
'''
class_context = CONTEXT
translation_type = NeuronConnectomeSynapseClassTranslation
input_type = (DataWithEvidenceDataSource, ConnectomeCSVDataSource)
output_type = DataWithEvidenceDataSource
[docs] def make_translation(self, sources):
tr = super(NeuronConnectomeSynapseClassTranslator, self).make_translation()
tr.source(sources[0])
tr.neurotransmitter_source(sources[1])
return tr
[docs] def translate(self, data_source, neurotransmitter_source):
res = self.make_new_output(sources=(data_source, neurotransmitter_source))
doc = res.evidence_context(Document)(author=['Dimitar Sht. Shterionov', 'Gerda Janssens'],
title='Data acquisition and modeling for learning and'
' reasoning in probabilistic logic environment',
year=2011,
uri='https://groups.google.com/forum/m/#!topic/openworm-discuss/G9wKoR8N-l0/discussion')
e = res.evidence_context(Evidence)(key="shterionov2011", reference=doc)
# We don't use doc.as_context for the statements' context because the document
# itself doesn't actually make the statements below, although it does *support*
# them by describing the process by which they are derived
docctx = res.data_context_for(document=doc)
# There are many entries in the Shterionov data where the number of synapses
# doesn't match that in, say, the Emmons (2015) data. We distinguish these so it's
# easy to look back at the differences later, but the different number of synapses
# isn't recorded here.
docctx_anynum = res.data_context_for(document=doc, number_matches=False)
e.supports(docctx.rdf_object)
res.data_context.add_import(docctx)
res.data_context.add_import(docctx_anynum)
with data_source.data_context.stored(Connection, Neuron) as srcctx:
with self.make_reader(neurotransmitter_source,
skipheader=False,
delimiter=';') as reader:
for row in reader:
pre, post, typ, number, nt = row
conn = srcctx.Connection.query(pre_cell=srcctx.Neuron.query(pre),
post_cell=srcctx.Neuron.query(post),
number=int(number),
syntype=typ)
hit = False
for c in conn.load():
docctx(Connection)(ident=c.identifier).synclass(nt)
hit = True
if not hit:
conn = srcctx.Connection.query(pre_cell=srcctx.Neuron(pre),
post_cell=srcctx.Neuron(post),
syntype=typ)
hit = False
for c in conn.load():
docctx_anynum(Connection)(ident=c.identifier).synclass(nt)
hit = True
if not hit:
L.warning("Didn't find any connections matching: {}".format(conn))
return res
def convert_to_cell(ctx, name, muscles, neurons, is_bwm):
ret = []
res = None
res2 = None
if name in neurons:
res = ctx.Neuron(name)
elif name in TO_EXPAND_MUSCLES:
# All of the muscle cell types we need to expand are pharyngeal, so no need to
# pass through is_bwm
res, res2 = expand_muscle(ctx, name)
elif name in muscles:
if is_bwm:
res = ctx.BodyWallMuscle(name)
else:
res = ctx.Muscle(name)
elif name in OTHER_CELLS:
res = ctx.Cell(name)
if res is not None:
ret.append(res)
if res2 is not None:
ret.append(res2)
return ret
def add_synapse(ctx, source, target, weight, syn_type):
c = ctx.Connection(pre_cell=source, post_cell=target,
number=weight, syntype=syn_type)
if isinstance(source, Neuron) and isinstance(target, Neuron):
c.termination('neuron')
elif isinstance(source, Neuron) and isinstance(target, Muscle) or \
isinstance(source, Muscle) and isinstance(target, Neuron):
c.termination('muscle')
return c
MUSCLES = {
'ANAL': 'MU_ANAL',
'INTR': 'MU_INT_R',
'INTL': 'MU_INT_L',
'SPH': 'MU_SPH'
}
TO_EXPAND_MUSCLES = ['PM1D', 'PM2D', 'PM3D', 'PM4D', 'PM5D']
#
# cells that are neither neurons or muscles. These are marked as
# 'Other Cells' in the wormbase cell list but are still part of the new
# connectome.
#
# TODO: In future work these should be uploaded seperately to
# owmeta in a new upload function and should be referred from there
# instead of this list.
OTHER_CELLS = ['MC1DL', 'MC1DR', 'MC1V', 'MC2DL', 'MC2DR', 'MC2V', 'MC3DL',
'MC3DR', 'MC3V']
def changed_muscle(x):
return MUSCLES[x]
def expand_muscle(ctx, name):
return ctx(Muscle)(name + 'L'), ctx(Muscle)(name + 'R')