Foundation Models for Data-driven Human Cell Simulation
Title: Foundation Models for Data-driven Human Cell Simulation
DNr: Berzelius-2024-230
Project Type: LiU Berzelius
Principal Investigator: Wei Ouyang <>
Affiliation: Kungliga Tekniska högskolan
Duration: 2024-08-01 – 2025-02-01
Classification: 10610


In the realm of system biology, the advent of whole-cell modeling heralds a new era of comprehensive, quantitative perspectives on cellular behavior, with profound implications for synthetic biology, medicine, and broader life science applications. The capacity to perform in-silico experiments promises to revolutionize our approach to biological research, enabling the exploration of cell dynamics and interactions in unprecedented detail. However, the complexity of modeling the entirety of a human cell remains a formidable challenge, often hindered by limitations in our current understanding of intricate biological systems. Supported by the KAW Data-Driven Life Science Fellows program, our newly formed research group AICell Lab ( at SciLifeLab aims to rise to this grand challenge by harnessing the potential of contemporary advancements in multi-omics data generation and artificial intelligence. We propose a novel approach to human cell simulation that leverages state-of-the-art deep learning techniques, including convolutional neural networks, transformers, and diffusion models. Guided by the success of tools like AlphaFold, we plan to dissect and analyze extant multi-omics datasets by building foundation models for human cell modeling. A key part of our project involves the integration of massive volumes of newly produced live cell, multiplexed images. By interweaving this data with our AI-driven analysis, we anticipate generating a robust and dynamic model of cellular behavior. Our strategy aligns with the emerging paradigm of "Foundation Models," aiming to provide a generative and predictive framework that can simulate human cellular responses and interactions with high fidelity. The goal of our project extends beyond the theoretical. We envision our human cell simulator as an innovative tool for practical application, empowering researchers to conduct complex virtual experiments that would otherwise be impossible or impractical. Such capability is poised to catalyze a leap forward in our understanding of human biology, providing transformative insights in areas such as disease modeling, drug discovery, and personalized medicine. The pursuit of this ambitious goal necessitates access to advanced computational resources, in particular, GPU clusters. The intensive computational demands of training and refining our sophisticated AI models underline the necessity for such resources. With the right support, we believe our project could make a substantive contribution to the field of system biology and beyond, illuminating the intricacies of human cell behavior and paving the way for the future of in-silico experimentation.