AiiDAlab: on the route to accelerate science

TL;DR

The paper presents AiiDAlab, a browser-based GUI that sits atop the AiiDA workflow engine to automate and orchestrate large-scale simulations while preserving full provenance. It documents maturation from computational materials science to cross-disciplinary use, including quantum chemistry, atmospheric modeling, battery testing, and experimental data analysis at large facilities. Key contributions include streamlined onboarding, HPC access, ELN/LIMS integration with semantic annotation for interoperability, and education-focused deployments such as teaching labs and demos. The resulting platform enables researchers to focus on science rather than infrastructure, with scales like $O(10^5)$ LPDM footprints and integrated data workflows advancing open and reproducible science.

Abstract

With the availability of ever-increasing computational capabilities, robust and automated research workflows are essential to enable and facilitate the execution and orchestration of large numbers of interdependent simulations in supercomputer facilities. However, the execution of these workflows still typically requires technical expertise in setting up calculation inputs, interpreting outputs, and handling the complexity of parallel code execution on remote machines. To address these challenges, the AiiDAlab platform was developed, making complex computational workflows accessible through an intuitive user interface that runs in a web browser. Here, we discuss how AiiDAlab has matured over the past few years, shifting its focus from computational materials science to become a powerful platform that accelerates scientific discovery across multiple disciplines. Thanks to its design, AiiDAlab allows scientists to focus on their research rather than on computational details and challenges, while keeping automatically track of the full simulation provenance via the underlying AiiDA engine and thus ensuring reproducibility. In particular, we discuss its adoption into quantum chemistry, atmospheric modeling, battery research, and even experimental data analysis at large-scale facilities, while also being actively used in educational settings. Driven by user feedback, significant effort has been made to simplify user onboarding, streamline access to computational resources, and provide robust mechanisms to work with large datasets. Furthermore, AiiDAlab is being integrated with electronic laboratory notebooks (ELNs), reinforcing adherence to the FAIR principles and supporting researchers in data-centric scientific disciplines in easily generating reproducible Open Research Data (ORD).

Figures (6)

• Figure 1: Top: The AiiDAlab-FLEXPART interface to run FLEXPART simulations for greenhouse gas monitoring sites. Bottom: example concentration footprints for the Beromünster tall-tower site in Switzerland, given for four different hourly time intervals. The concentration footprints indicate where the air sampled at the monitoring site was previously in contact with the surface and was potentially impacted by greenhouse gas emissions.
• Figure 2: Example workflow in the AiiDAlab-AtmoSpec app for a 3-hydroxypropanol molecule. First, a) molecular conformers are autogenerated from their SMILES code using a standard GUI component (a SmilesWidget). In the next step, b) the simulation parameters (such as a DFT functional and basis set) are specified, with pre-determined default values provided. Once the calculation is complete, c) the UV/vis spectrum is plotted in the spectrum widget, with the option to decompose the spectrum into the contributions of individual conformers.
• Figure 3: The AiiDAlab-Aurora graphical user interface showing a) the sample creation widget for importing the output files from the cell assembly robot, b) the protocol creation widget for defining the cycling conditions, c) the experiment building wizard, and d) the results widgets here showing a multi-cell analysis view.
• Figure 4: Screenshots of the integration between AiiDAlab and the openBIS ELN-LIMS, for a use case where a) a molecule is extracted from the ELN inventory and b) imported in the AiiDAlab submission interface for STM simulations. c) Results of the simulations (i.e., the STM images) are shown in AiiDAlab. d) Once exported to the openBIS ELN, results can be displayed directly in the ELN using a dedicated viewer for microscopy images. e) A chatbot can be used to query the database, e.g., to retrieve available simulations and experiments related to a specific molecule.
• Figure 5: CAMEA data pipeline. The data of the CAMEA experiment are first pushed from local to a shared NFS storage, then accessed for post-process analysis via the AiiDAlab PSI deployment and the LNS app. CAMEA users will see experimental data "live" in their AiiDAlab deployment, eliminating the need to manually transfer it periodically from the storage attached to the CAMEA hardware.