Bio

Moritz Schaefer builds AI systems that let scientists and clinicians interrogate biomedical data in plain language, and uses them to understand why cancer patients stop responding to therapy. He is a postdoctoral scholar at Stanford University and a Cancer Research Institute Immuno-Informatics Postdoctoral Fellow, working with Zinaida Good (Immunology and Rheumatology; Computational Medicine) and Jure Leskovec (Computer Science), and affiliated with the Center for Cancer Cell Therapy at the Stanford Cancer Institute.

He trained first as a computer scientist (B.Sc. and M.Sc., TU Berlin) and then as a biologist (Dr. sc. ETH Zurich in Systems Biology, with Constance Ciaudo), generating his own wet-lab data alongside the methods he built for it. As a postdoctoral researcher with Christoph Bock at the Medical University of Vienna and CeMM, he developed CellWhisperer, a model that links transcriptomes to the text describing them and lets biologists explore single-cell datasets by asking questions instead of writing code (Nature Biotechnology, 2025). The work was covered by The Scientist and recognized with Researcher of the Month at the Medical University of Vienna; he described it in an invited Tools of the Trade article in Nature Reviews Genetics.

At Stanford he extends this from cells to tissues, and from exploring data to explaining it. SpatialWhisperer annotates cells in histopathology images through language (ICML 2026), and AI agents reason across imaging, genomics and clinical records to propose why CAR T cell therapy fails in some patients with large B cell lymphoma and not others. His earlier SimulateGPT showed that large language models can act as biomedical simulators (Computers in Biology and Medicine, 2024). His models, training pipelines and benchmarks are released openly.

Website: https://www.moritzs.de
Google Scholar: https://scholar.google.com/citations?user=9S9SsvEAAAAJ
GitHub: https://github.com/moritzschaefer

Honors & Awards

  • Immuno-Informatics Postdoctoral Fellowship (CRI17992), Cancer Research Institute (2026 - 2029)
  • Researcher of the Month, Medical University of Vienna (2026)
  • Deutschlandstipendium National Scholarship, German Federal Ministry of Education and Research / General Electric Germany (2016 - 2018)
  • Deutschlandstipendium National Scholarship, German Federal Ministry of Education and Research / Carmeq GmbH (2015 - 2016)
  • Ferry Porsche Prize for outstanding achievement in mathematics and physics, Ferry Porsche Foundation (2011)

Boards, Advisory Committees, Professional Organizations

  • Reviewer, Nature Methods; Bioinformatics (2026 - Present)
  • Area Chair, AIDaR Workshop (AI Data Readiness for Scientific Discovery), NeurIPS (2026 - 2026)
  • Reviewer, ICML; NeurIPS workshops (FM4LS, ICBINB-BIO) (2025 - Present)
  • Member, EU-LIFE Postdoc Working Group (2024 - 2025)
  • Postdoc Representative, CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences (2023 - 2025)
  • Reviewer, ELLIS PhD and Postdoc Program application screening (2022 - 2024)

Professional Education

  • Doctor of Philosophy, Technische Universitat Berlin (2021)
  • Master of Science, Technische Universitat Berlin (2018)
  • Bachelor of Science, Technische Universitat Berlin (2016)
  • Doctor of Philosophy, Eidgenossische Technische Hochschule (ETH Zurich) (2022)
  • Dr. sc. ETH Zurich, ETH Zurich, Switzerland, Biology - Systems Biology (2022)
  • M.Sc., Technical University of Berlin, Germany, Computer Science (2018)
  • B.Sc., Technical University of Berlin, Germany, Computer Science (2016)

Stanford Advisors

Contact

  • Academic
    moritzs@stanford.edu
    University - Scholar Department: School of Medicine Position: Postdoctoral Scholar

Additional Info

Links

Current Research and Scholarly Interests

I build AI systems that can be asked a question about biomedical data in ordinary language and that show their work well enough to be checked. Two lines of research feed into each other: multimodal models that connect measurements of cells and tissues to the language describing them, and AI agents that use those models to carry an analysis through from raw data to a mechanistic claim.

Language-multimodal models for cells and tissues. CellWhisperer (Nature Biotechnology, 2025) embeds one million RNA-sequencing profiles together with AI-curated text descriptions, so a single-cell dataset can be interrogated by asking rather than by writing code; in a head-to-head comparison it reached the same conclusions as a conventional analysis roughly four times faster. SpatialWhisperer (ICML 2026) carries the approach into histopathology, annotating individual cells in tissue images through language, and SpotWhisperer links what a histopathology image shows to the molecular state underneath it. Models, training pipelines and benchmarks are released openly.

Agentic AI for therapy resistance. CAR T cell therapy cures some patients with large B cell lymphoma and fails in others, and the evidence that would explain the difference is scattered across pathology images, genomics and the clinical record. I develop AI agents that assemble that evidence across modalities and propose resistance mechanisms specific enough to be tested at the bench. This follows from SimulateGPT (Computers in Biology and Medicine, 2024), which showed that a language model prompted as a simulator outperforms direct inference at predicting gene essentiality and patient survival.

Underneath both lines is the same commitment: an AI system that a biologist or clinician cannot interrogate, and whose reasoning they cannot inspect, is not usable for a decision that matters. Building models that explain themselves in language is the most direct route to that, and it is what my future lab will be built around.

Lab Affiliations

All Publications

  • Multimodal learning enables chat-based exploration of single-cell data NATURE BIOTECHNOLOGY Schaefer, M., Peneder, P., Malzl, D., Lombardo, S., Peycheva, M., Burton, J., Hakobyan, A., Sharma, V., Krausgruber, T., Sin, C., Menche, J., Tomazou, E. M., Bock, C. 2025

    Abstract

    Single-cell sequencing characterizes biological samples at unprecedented scale and detail, but data interpretation remains challenging. Here, we present CellWhisperer, an artificial intelligence (AI) model and software tool for chat-based interrogation of gene expression. We establish a multimodal embedding of transcriptomes and their textual annotations, using contrastive learning on 1 million RNA sequencing profiles with AI-curated descriptions. This embedding informs a large language model that answers user-provided questions about cells and genes in natural-language chats. We benchmark CellWhisperer's performance for zero-shot prediction of cell types and other biological annotations and demonstrate its use for biological discovery in a meta-analysis of human embryonic development. We integrate a CellWhisperer chat box with the CELLxGENE browser, allowing users to interactively explore gene expression through a combined graphical and chat interface. In summary, CellWhisperer leverages large community-scale data repositories to connect transcriptomes and text, thereby enabling interactive exploration of single-cell RNA-sequencing data with natural-language chats.

    View details for DOI 10.1038/s41587-025-02857-9

    View details for Web of Science ID 001611684000001

    View details for PubMedID 41219484

    View details for PubMedCentralID 7453005

  • Systematic discovery of CRISPR-boosted CAR T cell immunotherapies NATURE Datlinger, P., Pankevich, E. V., Arnold, C. D., Pranckevicius, N., Lin, J., Romanovskaia, D., Schaefer, M., Piras, F., Orts, A., Nemc, A., Biesaga, P. N., Chan, M., Neuwirth, T., Artemov, A. V., Li, W., Ladstatter, S., Krausgruber, T., Bock, C. 2025

    Abstract

    Chimeric antigen receptor (CAR) T cell therapy has shown remarkable success in treating blood cancers, but CAR T cell dysfunction remains a common cause of treatment failure1. Here we present CELLFIE, a CRISPR screening platform for enhancing CAR T cells across multiple clinical objectives. We performed genome-wide screens in human primary CAR T cells, with readouts capturing key aspects of T cell biology, including proliferation, target cell recognition, activation, apoptosis and fratricide, and exhaustion. Screening hits were prioritized using a new in vivo CROP-seq2 method in a xenograft model of human leukaemia, establishing several gene knockouts that boost CAR T cell efficacy. Most notably, we discovered that RHOG knockout is a potent and unexpected CAR T cell enhancer, both individually and together with FAS knockout, which was validated across multiple in vivo models, CAR designs and sample donors, and in patient-derived cells. Demonstrating the versatility of the CELLFIE platform, we also conducted combinatorial CRISPR screens to identify synergistic gene pairs and saturation base-editing screens to characterize RHOG variants. In summary, we discovered, validated and biologically characterized CRISPR-boosted CAR T cells that outperform standard CAR T cells in widely used benchmarks, establishing a foundational resource for optimizing cell-based immunotherapies.

    View details for DOI 10.1038/s41586-025-09507-9

    View details for Web of Science ID 001577752400001

    View details for PubMedID 40993398

    View details for PubMedCentralID 10965011

  • GPT-4 as a biomedical simulator. Computers in biology and medicine Schaefer, M., Reichl, S., Ter Horst, R., Nicolas, A. M., Krausgruber, T., Piras, F., Stepper, P., Bock, C., Samwald, M. 2024; 178: 108796

    Abstract

    BACKGROUND: Computational simulation of biological processes can be a valuable tool for accelerating biomedical research, but usually requires extensive domain knowledge and manual adaptation. Large language models (LLMs) such as GPT-4 have proven surprisingly successful for a wide range of tasks. This study provides proof-of-concept for the use of GPT-4 as a versatile simulator of biological systems.METHODS: We introduce SimulateGPT, a proof-of-concept for knowledge-driven simulation across levels of biological organization through structured prompting of GPT-4. We benchmarked our approach against direct GPT-4 inference in blinded qualitative evaluations by domain experts in four scenarios and in two quantitative scenarios with experimental ground truth. The qualitative scenarios included mouse experiments with known outcomes and treatment decision support in sepsis. The quantitative scenarios included prediction of gene essentiality in cancer cells and progression-free survival in cancer patients.RESULTS: In qualitative experiments, biomedical scientists rated SimulateGPT's predictions favorably over direct GPT-4 inference. In quantitative experiments, SimulateGPT substantially improved classification accuracy for predicting the essentiality of individual genes and increased correlation coefficients and precision in the regression task of predicting progression-free survival.CONCLUSION: This proof-of-concept study suggests that LLMs may enable a new class of biomedical simulators. Such text-based simulations appear well suited for modeling and understanding complex living systems that are difficult to describe with physics-based first-principles simulations, but for which extensive knowledge is available as written text. Finally, we propose several directions for further development of LLM-based biomedical simulators, including augmentation through web search retrieval, integrated mathematical modeling, and fine-tuning on experimental data.

    View details for DOI 10.1016/j.compbiomed.2024.108796

    View details for PubMedID 38909448

https://orcid.org/0000-0001-6489-1947