Tool recommender system in Galaxy using deep learning

Abstract

Background 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 help researchers with creating workflows, a system is developed to recommend tools that can facilitate further data analysis. Findings A model is developed to recommend tools using a deep learning approach by analysing workflows composed by researchers on the European Galaxy server. The higher-order dependencies in workflows, represented as directed acyclic graphs, are learned by training a gated recurrent units neural network, a variant of a recurrent neural network. In the neural network training, the weights of tools used are derived from their usage frequencies over time and the sequences of tools are uniformly sampled from training data. Hyperparameters of the neural network are optimized using Bayesian optimization. Mean accuracy of 98% in recommending tools is achieved for the top-1 metric. Conclusions The model is accessed by a Galaxy API to provide researchers with recommended tools in an interactive manner using multiple user interface integrations on the European Galaxy server. High-quality and highly used tools are shown at the top of the recommendations. The scripts and data to create the recommendation system are available under MIT license at https://github.com/anuprulez/galaxy_tool_recommendation.

The role of metadata in reproducible computational research

“…116,117 In the context of workflows, certain tool combinations tend to be chained in predictable usage patterns driven by application; these patterns can be mined for tool recommender software used in workbenches. 118 CodeMeta 119 prescribes JSON-LD (JSON for Linked Data) standards for code metadata markup. While CodeMeta is not itself an ontology, it leverages Schema.org ontologies to provide language-agnostic means of describing software as well as "crosswalks" to translate manifests from various software repositories, registries, and archives into CodeMeta (Box 3).…”

Synthesis of Scientific Workflows: Theory and Practice of an Instance-Aware Approach

“…This simplification of the problem, when compared to the general program synthesis, allows researchers to optimise the synthesis approach and provide more lightweight solutions. For example, systems, such as Wings [46] and the tool recommender system in Galaxy [87] focus on finding individual tools to fill in the gaps in existing workflows. These types of approaches synthesise a workflow by instantiating the missing steps within the provided workflow structure.…”

“…At each step of the workflow construction process, only services which are compatible and likely to be useful are displayed. Similarly, the tool recommender system in Galaxy [87] focuses on finding suitable tools to aid the user while developing a workflow. Galaxy is a popular web-based platform for high-throughput sequencing data and other big data analyses in bioinformatics.…”

“…The presented approaches are often hard to compare as they focus on a different part of the workflow life-cycle [92], as well as different domains. Some systems [34,87] provide assistance while choosing concrete workflow steps, other [46,74] aid in setting up the "abstract" workflows, before they can be implemented, while a few [13,138] automate the whole process omitting the workflow altogether. Similarly, the theoretical approaches (e.g., [15,23,61]), presented in Chapter 3, aim to solve complex and universal synthesis problems, assuming that the required implementations and/or domain annotations are available.…”

“…However, such annotations are hard to find for an arbitrary scientific workflow. That is why many of the current approaches that aim to automate workflow composition restrict the problem and focus on either well-defined and curated domains [145] or the automation of individual workflows steps, instead of the workflow as a whole [46,87].…”