ALADIN

ALADIN: Advanced Local Drug-Target Interaction Prediction technique

version 1.0, April 20, 2017

---

This package is written by:

Krisztian Buza and Ladislav Peska

Email: [email protected]

---

This package is based on the PyDTI package by Yong Liu, https://github.com/stephenliu0423/PyDTI

and PyHubs package available from http://biointelligence.hu/pyhubs/

---

ALADIN method

ALADIN works on Python 2.7 (tested on Intel Python 2.7.12) and requires NumPy <=1.11, scikit-learn and SciPy to run.

To get the results of different methods, please run eval.py by setting suitable values for the following parameters:

--method 			set DTI prediction method, i.e. blmnii, netlaprls, wnngip, aladin
--dataset: 			choose the benchmark dataset, i.e., nr, gpcr, ic, e, kinase
--method-opt:		set arguments for each method
--predict-num:		a positive integer for predicting top-N novel DTIs for each drug and target (default 0)
    
The following examples evaluates the ALADIN method and its competitors on the GPCR dataset (hyperparameter search is included)
	main(['--method=aladin', '--dataset=gpcr', '--cvs=1', '--specify-arg=1', '--method-opt=hpLearning=1'])   
	main(['--method=netlaprls', '--dataset=gpcr', '--cvs=1', '--specify-arg=1'])
main(['--method=wnngip', '--dataset=gpcr', '--cvs=1', '--specify-arg=1'])
main(['--method=blmnii', '--dataset=gpcr', '--cvs=1', '--specify-arg=1'])

The following examples performes prediction of top-20 new interactions within the GPCR dataset (hyperparameter search is included)
main(['--method=aladin', '--dataset=gpcr', '--predict-num=20', '--specify-arg=1', '--method-opt=hpLearning=1'])
main(['--method=netlaprls', '--dataset=gpcr', '--predict-num=20', '--specify-arg=1']) 
main(['--method=wnngip', '--dataset=gpcr', '--predict-num=20', '--specify-arg=1']) 
main(['--method=blmnii', '--dataset=gpcr', '--predict-num=20', '--specify-arg=1']) 

---

Hyperparameter search

By default, hyperparameters are learned via grid search for all evaluated methods on the train set.

To change searched hyperparameters, update cv_eval.py: 
blmnii_cv_eval() for blmnii,
aladin_cv_eval() for aladin,
wnngip_cv_eval() for wnngip,
netlaprls_cv_eval() for netlaprls,

---

Evaluation

The results can be analysed via result_sign_analysis.py.