Title:	Alpha Cell: Whole cell modeling at scale
DNr:	Berzelius-2026-182
Project Type:	LiU Berzelius
Principal Investigator:	Jan Ellenberg <jan.ellenberg@ki.se>
Affiliation:	Karolinska Institutet
Duration:	2026-06-26 – 2027-01-01
Classification:	10210
Homepage:	https://www.scilifelab.se/alpha-cell/
Keywords:

Abstract

Alpha Cell is a new SciLifeLab strategic program funded by the Knut and Alice Wallenberg Foundation, and is led by Jan Ellenberg, Director of SciLifeLab and Mathias Ulhen, Director of the Human Protein Atlas (HPA), together with a collaborative, multi-institutional constellation of SciLifeLab researchers and infrastructure units. The Alpha Cell program aims to build the first molecular-level AI model of the human cell that can predict how a cell carries out its function. Data for the program will be ingested from the Human Protein Atlas’ inventory of the human proteome, complementary data resources including the Mitocheck, OpenCell and CellMap databases and newly generated using the most advanced imaging technologies, including 3D super-resolution microscopy and correlative 3D volume and tomographic cryo-electron microscopy. The program is planned over five strategic phases: 1. Human Protein Atlas: Build a foundational model of protein distribution. 2. Space: Build a molecular resolution 3D reference data set for the human cell and generate a foundational spatial AI cell model. 3. Time: Build a single molecule precision temporal reference data set for the human cell and generate a foundational temporal AI cell model. 4. Future: Build a molecular-space-time foundation AI cell model to predict future cell states. 5. Control: Translational application of Alpha Cell to predict the key molecular handles that gives us control to maintain health and recognize and prevent disease. The current application will support Phase 1 where we will leverage the HPA and other open source data sets to create a foundational generative AI model to predict the metabolic functions of human cells. This phase serves as a proof-of-concept, using high-quality cellular and subcellular multi-omic data to identify which molecular information is most predictive of cell behavior and where critical biological data gaps still exist.