Type: Virtual Research Environment (VRE)

Building a HADDOCK3 service making use of EOSC cloud resources

About

Computational structural biology has provided valuable insights into many research fields into the complex and intricate network of interactions between macromolecules. Our software, HADDOCK, is contributing to shedding light on these interactions, with over 63,000 users worldwide, mainly through its WeNMR webserver implementation. Its operation mode is geared towards HTC (grid/local) resources.

The latest version of HADDOCK, called HADDOCK3, co-developed within the context of the BioExcel European Centre of Excellence for Computational Biomolecular Research, features a complete rewrite of HADDOCK into a modular structure that provides users with great flexibility to define custom workflows, thereby expanding HADDOCK’s capabilities and usability. However, this new version is not suited for HTC/grid operation and would be better suited for HTC cloud. As such, it currently lacks a web service to make it easily available to end users.

Within EOSC Data Commons, we will develop an execution machinery for HADDOCK3 workloads, utilising EOSC cloud resources and creating a web interface to HADDOCK3 to promote its use worldwide. In addition to HADDOCK3-related work, the use case will also develop a web service for DeepRank, an AI-based software developed in Utrecht for the analysis of biomolecular complexes.

Use Case Status Before Joining EOSC Data Commons

The HADDOCK server has been utilising European HTC resources since 2010 and operates under the European Open Science Cloud (EOSC). Besides the application software, the service also provides automated pre- and post-processing, compute, temporary storage and job scheduling and monitoring for running the application, so that researchers do not need to worry about application porting and procuring the necessary compute infrastructure. HADDOCK has been and continues to be used to address highly relevant scientific questions in the life sciences domain. Its user community comprises almost 70,000 registered users from 177 different countries, sending over 13 million HTC jobs and consuming around 5,000 CPU years annually on the EGI/EOSC HTC resources (grid computing). The current web portal runs HADDOCK version 2.5, which uses the DIRAC service in the background to distribute jobs.

Objectives in the Project

To facilitate data usage, exploitation and reproducibility.
To accelerate research with automated orchestration and deployment of analysis/modelling workflows
To provide the next-generation HADDOCK (version 3) cloud-enabled service, offering HADDOCK3 as a web service will require a new way of interacting with the EOSC cloud resources and new middleware to manage workloads (including automatic VM management). Alternatively, we will consider building a DIRAC interaction mechanism in HADDOCK3 to allow it to use the current grid HTC resources.
To facilitate the use of DeepRank, a web service will be developed using both local resources at Utrecht University and EOSC cloud resources (and possibly GPU resources). As for HADDOCK3, this will require new middleware to manage workloads (including automatic VM management).

Integration with EOSC Data Commons Services and Components

Expected Results

New HADDOCK3 and DeepRank web services, fully cloud-enabled and allowing for the serving of a large community of users.

Technical Integration Plan

Timeline

By November 2025

Development of the full stack apps for HADDOCK3/DeepRank services, and demonstration of execution only on local resources.

By June 2026

Initial integration and testing of HADDOCK3/DeepRank with the EOSC Data Player.

By January 2028

Complete integration of HADDOCK3/DeepRank with the EOSC Data Player.