Towards the automatic integration of tools in Galaxy: the anvi'o suite and a new AI-driven automation pipeline

Towards the automatic integration of tools in Galaxy: the anvi’o suite and a new AI-driven automation pipeline Fabio Cumbo1, Jayadev Joshi1, Daniel Blankenberg1,2,* 1 Center for Computational Life Sciences, Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA 2 Department of Molecular Medicine, Cleveland Clinic Lerner College of Medicine, Case Western Reserve University, Cleveland, OH, USA * Correspondent Author: [email protected] Abstract The Galaxy platform has emerged as a leading framework for accessible, reproducible, and scalable data analysis in the life sciences. However, integrating new tools into Galaxy remains challenging due to the complexity of manually defining XML-based wrappers. This challenge is especially pronounced for large and modular tool suites such as anvi’o, a comprehensive framework for the analysis and visualization of ‘omics data. In this work, we present two complementary efforts to simplify and accelerate the integration of tools into Galaxy: (i) the automatic integration of the anvi’o Python-based suite into the Galaxy ecosystem and (ii) the development of an AI-driven pipeline to automate wrapper generation. First, we worked on integrating the anvi’o suite into Galaxy. With its over 76-thousands downloads solely from Conda, anvi’o is a widely used framework for the analysis and visualization of metagenomics, pangenomics, and microbial population data. Its extensive modular design makes it a powerful tool for 'omics data analysis, but also introduces significant complexity. We developed a structured pipeline to process all of the >100 anvi’o modules and subroutines, and create XML wrappers ready to be integrated into the Galaxy environment. This work expands Galaxy’s capabilities by enabling researchers to incorporate anvi’o’s powerful analytical tools into their workflows, and it is the first step towards a general purpose framework for the automatic integration of Python-based tools in Galaxy. In parallel, we developed an AI-driven pipeline to automate the generation of Galaxy wrappers. It is based on DeepSeek R1, a state-of-the-art large language model, which we fine-tuned using a curated dataset of existing Galaxy wrappers from the official ToolShed repository. The model was trained to understand the structure of Galaxy wrapper XML files, including input and output definitions, parameter handling, and execution logic. By providing the model with a tool’s command-line interface, input arguments, and expected outputs, our pipeline can produce a fully defined XML wrapper. It is worth to note, however, that this is in its very early stage of development and far from the end goal of producing fully functional wrappers, indeed requiring manual adjustments in most of the cases, but still drastically reducing the time and technical knowledge required to define wrappers from scratch. With that said, our preliminary results clearly highlight this as the right path for the first AI-driven framework for the automatic integration of tools in Galaxy regardless of their programming language and technical development aspects. By combining these efforts, we address one of the key challenges that always affected the Galaxy community whose solution would benefit technical and non-technical developers: the need for automated software solutions for the integration of tools into the Galaxy platform. Presented at the Galaxy and Bioconductor Community Conference (GBCC2025), New York, June 2025