demo

Example codes

Sample code to generate the embedding space and predict drug-disease associations are provided in run_demo.ipynb.

Input File formats

Hierarchy file

The input hierarchy file for semantic similarity calculation should be a .csv file with following information:

child	parent
child1	parent1
parent1	parent2
parent2	disease
...	...
parent n	drug
...	...

Caution: All leafs should be branched out from root named as either 'disease' or 'drug'

Network file

The input network file should be a tab- or space-separated .txt file with following orders:

source target weight edge_type edge_id

Node type label file

The input node type label file for heterogeneous Skip-gram should be a '.tsv' file with following information:

node	type
node 1	gene
node 2	disease
node 3	gene
...	...
node n	drug
...	...

Caution: The drug, disease and gene node types should be given as drug, disease, gene for the model to properly recognize its type.

Positive/negative drug-disease association file

Positive/negative pair information for drug-disease association prediction task should be input as a .tsv file with following structure:

drug	disease	label
drug1	disease1	1
drug2	disease2	0
...	...	...

where label == 1 refers to positive pairs and label == 0 refers to negative pairs.