CellDJBench: Benchmark Datasets for Data-Driven Biological Fluid Simulation

Qianru Zhang, Haixin Wang, Yudong Yan, Chenglei Yu, Xiangyu Kuang, Xiao Luo, Yuansheng Cao, Siu Ming Yiu, Tailin Wu

Keywords: Cell; biological dynamic simulation

Abstract: Biological fluid simulation is a critical tool for comprehending the intricate and complex fluid dynamics that occur within biological systems. Recently, data-driven techniques have emerged as a promising avenue to enhance the accuracy and efficiency of biological fluid simulations. However, the community encounters two challenges. (1) Existing biological datasets only capture static snapshots, lacking the ability to capture dynamic biological processes. (2) These datasets are limited in scale due to the demanding experimental conditions. To address these challenges, this paper introduces four comprehensive large-scale datasets: Tension, Wets, CellDivision and Jellyfish, containing a wealth of biological dynamics and pushing the boundary of data-driven methods. These datasets have been meticulously designed to encompass a wide array of biological fluid dynamics scenarios. By incorporating physical modeling techniques such as phase-field method, these datasets provide a standardized evaluation framework for data-driven approaches. They empower researchers to objectively assess and compare different methodologies, fostering advancements in the field of biological fluid simulation. Furthermore, the availability of these benchmark datasets facilitates reproducibility and enhances the comparability of results across studies, promoting knowledge sharing and collaboration within the research community. Researchers can build upon existing models, leading to cumulative progress in the development of accurate and efficient data-driven models for simulating complex fluid dynamics within biological systems. We offer benchmark code and \model\ dataset link through the following link: \href{https://anonymous.4open.science/r/BioJCell--9E53/README.md}{https://anonymous.4open.science/r/ CellDJBench}.

Primary Area: datasets and benchmarks

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 2030