A Deep Learning Framework for Prediction of the Mechanism of Action”,
Document Type
Article
Publication Date
6-1-2021
Publication Title
International Journal of Computer Applications
Volume
183
Issue
12
First page number:
1
Last page number:
7
Abstract
This paper aims to apply deep learning algorithms to advance a new drug’s mechanism of action (MoA) prediction. Since one drug can have one or more MoAs, algorithms must be developed to perform multi-label classification problems. This paper puts forward a deep learning framework, MoA Net, which ensembles one residual network and five convolutional neural networks to predict MoA targets. To find optimal parameter sets, the authors implements Bayesian tuning techniques on each sub network of MoA Net. The study uses logarithmic loss function to evaluate the model’s performance. Results show successful MoA target prediction in the dataset provided by the LISH and LINCS Connectivity Map.
Keywords
Deep Learning, prediction, MOA
Disciplines
Chemicals and Drugs | Genetics and Genomics
Repository Citation
Dai, J.,
Latifi, S.
(2021).
A Deep Learning Framework for Prediction of the Mechanism of Action”,.
International Journal of Computer Applications, 183(12),
1-7.
http://dx.doi.org/10.5120/ijca2021921382