24

Jilong Liu, Jianwen Zhou, Zhaoqiang Li, Qingyu Xiao, Kaixin Yang, Ziru Chen, Lu Yao, Hengshen Liu, Yibo Miao, Junyu Luo, Yiping Chen, Ruifeng jing, Yixue Li

Guangzhou Laboratory, Guangzhou 510005, Guangdong Province, China

Bioinformatics exploratory analysis heavily utilizes interactive environments (e.g., Jupyter Lab). However, transitioning ad-hoc analyses into reusable, scalable applications remains a significant challenge, hindering reproducibility and workflow integration. To bridge this gap, we introduce the Interactive Execution Service (IES), a novel paradigm extending the GA4GH Task Execution Service (TES) standard. IES establishes a continuous analytical lifecycle, seamlessly connecting interactive exploration with production-grade batch execution. The IES framework defines two core states: Interactive Analysis Instances (Instances) and Packaged Interactive Applications (Apps). An IES Instance provides a cloud-native (Kubernetes) environment integrated with object storage and supports popular IDEs (Jupyter, Code-Server) for exploration. Crucially, users can declaratively package an entire analytical context—code, environment, and runtime specifications—directly within the Instance into a reusable IES App. This transformation uses an intuitive graphical interface, requiring no manual scripting or workflow language expertise (e.g., WDL). Users simply define TES-compatible inputs, outputs, and commands. Apps can also be restored into active instances for iterative refinement. The resultant IES App is a versioned, shareable TES task, ready for immediate batch orchestration via existing TES engines. Simultaneously, it serves as a first-class workflow component, integrable into pipelines built with standards like WDL. By unifying interactive and batch modalities under GA4GH TES, IES eliminates the traditional disconnect between research exploration and production deployment. This paradigm shift accelerates research translation, substantially enhances reproducibility, and fosters truly integrated, efficient bioinformatics workflows through formalized application encapsulation from the interactive outset.