khp-informatics / adr-graph Goto Github PK

filterNetworksByCommonSource prints a count for source nodes using

networkA['sourcenodes'] = networkA['graph'].vs(type_eq=networkA['sourcetype'])['name']

however if sourcetype and targettype are the same, this will be the count of ALL nodes, not just source nodes

I have downloaded the data and docker image. The initiation of the EdgePrediction image throws an error. This is from the second line in the ExampleUseDoc.rst.

ep = EdgePrediction.EdgePrediction()
AttributeError Traceback (most recent call last)
in
----> 1 ep = EdgePrediction.EdgePrediction()

AttributeError: class EdgePrediction has no attribute 'EdgePrediction'

Even running the following
ep = EdgePrediction()
I run into other errors later.

Thank you for your time in developing this, and I do recognize that this hasn't seen a lot of attention but despite that, I would really like to see this package work.

#3

particularly important now this used as a package

Wondering if there is another place to set, require_all_predictors to 'False' other than "init"? I was hoping to increase the number of models the 'predictAll' method would produce. Switching the default value to 'False' did not seem to change the result.

self.require_all_predictors = False in the init method

epall = ep.EdgePrediction()
s = epall.CSV_to_graph('data.csv', header=True, srcNameCol=0, srcTypeCol=1, tgtNameCol=4, tgtTypeCol=3, edgeTypeCol=2)
epall.preprocess()
epall.to_predict = 'HAS_SIDE_EFFECT'
epall.network_order = ['HAS_SIDE_EFFECT', 'DRUG_TARGETS', 'INDICATED_FOR']

AllResult = epall.predictAll(calculate_auc = True)

still produces several results that indicate:

self.require_all_predictors is True and x/3 networks have 0 predictor overlap
not optimizing a model for CXXXXX

Again my main goal is to just generate a higher percentage of models. Thank you for your time and attention.