Transformer-based tool recommendation system in Galaxy

Abstract

Galaxy is a web-based open-source platform for scientific analyses. Researchers use thousands of high-quality tools and workflows for their respective analyses. Tool recommender system predicts a collection of tools that can be used to extend an analysis. In this work, a tool recommender system is developed by training a Transformer-based neural network on workflows available on Galaxy Europe. Compared to the existing tool recommender system on Galaxy Europe that trains a recurrent neural network, the transformer-based neural network achieves two times faster convergence, has a four times lower model usage (model loading + prediction) time and shows a better generalisation that goes beyond training workflows. The scripts to create the recommendation model are available under MIT licence at https://github.com/anuprulez/galaxy_tool_recommendation_transformers.