RT Journal Article
SR Electronic
T1 Tool recommender system in Galaxy using deep learning
JF bioRxiv
FD Cold Spring Harbor Laboratory
SP 838599
DO 10.1101/838599
A1 Anup Kumar
A1 Björn Grüning
A1 Rolf Backofen
YR 2019
UL http://biorxiv.org/content/early/2019/11/12/838599.abstract
AB Galaxy is a web-based and open-source scientific data-processing platform. Researchers compose pipelines in Galaxy to analyse scientific data. These pipelines, also known as workflows, can be complex and difficult to create from thousands of tools, especially for researchers new to Galaxy. To make creating workflows easier, faster and less error-prone, a predictive system is developed to recommend tools facilitating further analysis. A model is created to recommend tools by analysing workflows, composed by researchers on the European Galaxy server, using a deep learning approach. The higher-order dependencies in workflows, represented as directed acyclic graphs, are learned by training a gated recurrent units (GRU) neural network, a variant of a recurrent neural network (RNN). The weights of tools used in the neural network training are derived from their usage frequencies over a period of time. The hyper-parameters of the neural network are optimised using Bayesian optimisation. An accuracy of 97% in predicting tools is achieved by the model for precision@1, precision@2 and precision@3 metrics. It is accessed by a Galaxy API to recommend tools in real-time. Multiple user interface (UI) integrations on the server communicate with this API to apprise researchers of these recommended tools interactively.Contact kumara{at}informatik.uni-freiburg.degruening{at}informatik.uni-freiburg.debackofen{at}informatik.uni-freiburg.de