Researchers propose AI-driven 'digital organism' to simulate drug effects across biological scales

Researchers proposing an AI-driven 'digital organism' in Nature Medicine on August 13, 2026, aim to simulate drug and genetic effects across biological scales. Described as a 'flight simulator for medicine,' the system connects specialized models for DNA, proteins, and cells. While GenBio AI plans to release the first implementation, the Chan Zuckerberg Biohub is building a competing open platform. The technology faces challenges like biased training data but aims to eliminate weak drug candidates before clinical trials.

AI-driven virtual cell concept

• ▪Scientists describe the proposed AI-driven digital organism, published in Nature Medicine on August 13, 2026, as a "flight simulator for medicine" to run biological experiments virtually.
• ▪A paper published in Nature Medicine on August 13, 2026, proposed an AI-driven "digital organism" to simulate how drugs or genetic changes affect molecules, cells, and humans.

Multi-scale biological simulation architecture

• ▪The virtual cell system proposed by researcher Eric Xing and co-authors aims to integrate sequence, structure, imaging, and interaction data to predict the dynamic effects of biological interventions.
• ▪Traditional biology AI models are limited to specialized tasks like predicting protein shapes, whereas the proposed digital organism simulator published in Nature Medicine on August 13, 2026, aims to connect models across DNA, proteins, cells, and tissues.

Drug screening risk prediction

• ▪Key challenges in developing virtual cells include biased or incomplete training data, varying patient responses, and AI's tendency to confuse correlation with causation.
• ▪The virtual cell simulator is designed to streamline biology by helping researchers screen drug candidates earlier and eliminate weaker ideas before animal or clinical trials.

GenBio AI implementation efforts

• ▪Eric Xing, a co-author of the Nature Medicine paper published on August 13, 2026, also serves as the president of the Mohamed bin Zayed University of AI and a professor at Carnegie Mellon University.
• ▪GenBio AI, co-founded by chief scientist Eric Xing, plans to release what it describes as the world's first implementation of a virtual cell.

Biohub virtual cell competition

• ▪The Chan Zuckerberg Biohub is developing an open platform of AI models to predict cell behavior across tissues, individuals, and health conditions.
• ▪Eric Xing characterized the Chan Zuckerberg Biohub's virtual-cell program as a friendly competitor that relies more heavily on language-model-style sequence models than GenBio AI's multi-scale approach.

Foundation models for genomics

• ▪Single-cell transcriptomics foundation models, including scGPT and scLong, have been designed to capture gene context and analyze multi-omics data using generative AI.
• ▪Researchers have developed genomic foundation models, such as the Nucleotide Transformer and Evo, to perform sequence modeling and design from molecular to genome scales.