Histopathology-based protein multiplex generation using deep learning

Multiplexed protein imaging offers valuable insights into interactions between tumours and their surrounding tumour microenvironment, but its widespread use is limited by cost, time and tissue availability. Here we present HistoPlexer, a deep learning framework that generates spatially resolved protein multiplexes directly from standard haematoxylin and eosin (H&E) histopathology images. HistoPlexer jointly predicts multiple tumour and immune markers using a conditional generative adversarial architecture with custom loss functions designed to ensure pixel- and embedding-level similarity while mitigating slice-to-slice variations. A comprehensive evaluation of metastatic melanoma samples demonstrates that HistoPlexer-generated protein maps closely resemble real maps, as validated by expert assessment. They preserve crucial biological relationships by capturing spatial co-localization patterns among proteins. The spatial distribution of immune infiltration from HistoPlexer-generated protein multiplex enables stratification of tumours into immune subtypes. In an independent cohort, integration of HistoPlexer-derived features into predictive models enhances performance in survival prediction and immune subtype classification compared to models using H&E features alone. To assess broader applicability, we benchmarked HistoPlexer on publicly available pixel-aligned datasets from different cancer types. In all settings, HistoPlexer consistently outperformed baseline methods, demonstrating robustness across diverse tissue types and imaging conditions. By enabling whole-slide protein multiplex generation from routine H&E images, HistoPlexer offers a cost- and time-efficient approach to tumour microenvironment characterization with strong potential to advance precision oncology.

Authors

Institutions

Publication Details

Journal
Nature Machine Intelligence
Published
2025-08-04
DOI
https://doi.org/10.1038/s42256-025-01074-y
Citations
6
Primary Topic
Cell Image Analysis Techniques
Type
article
Field-Weighted Citation Impact
9.80
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Histopathology-based protein multiplex generation using deep learning

Rebekka Wegmann, René Holtackers, Viktor Hendrik Koelzer, Faisal Alquaddoomi et al.
6 citations
Nature Machine Intelligence
Cell Image Analysis Techniques
9.80
article

Histopathology-based protein multiplex generation using deep learning

Rebekka Wegmann, René Holtackers, Viktor Hendrik Koelzer, Faisal Alquaddoomi, Laurie Prélot, Marina Bacac, Jonas Albinus, Audrey van Drogen, André Fedier, Marina Tusup, Marcus Vetter, Sujana Sivapatham, Franziska Singer, Stefanie Engler, Ricardo Coelho, Markus Gabriel Manz, Patrick G. A. Pedrioli, Bernd Bodenmiller, Lars Bosshard, Andreas N. WICKI, Boqi Chen, Linda Grob, Natalia Chicherova, Daniel Baumhoer, Michael Weller, Tobias Schär, Beatrice Beck‐Schimmer, Gabriela Senti, Stefan Nicolet, Petra Catalina Schwalie, Christian P. Kunze, Kjong-Van Lehmann, Martin Mehnert, Fabian Wendt, Cinzia Esposito, Sonali Andani, Lara Bernasconi, Holger Moch, Bernard Friedrich Hild, Flavio C. Lombardo, Katja Eschbach, Lucas Pelkmans, Martin Erkens, Severin Schwan, Gregor Zuend, Alice K. Jacobs, Bettina Sobottka, Vipin T. Sreedharan, Shuqing Yu, Abdullah Kahraman, Detlef Günther, Nicola Miglino, Gunnar Rätsch, Niko Beerenwinkel, Michael Prummer, Sandra Goetze, Stefan G. Stark, Jack Kuipers, Tatjana Vlajnic, Esther Danenberg, Ruben Casanova, Francis Jacob, Per-Olof Attinger, Johanna Ziegler, Tamara Huesser, Berend Snijder, Anja Frei, Julian M. Metzler, Bruno S. Frey, Gabriele Gut, Ximena Bonilla, Wenguang Shao, Salvatore Piscuoglio, Ilaria Alborelli, Simone E. Muenst, Mattheus H. E. Wildschut, Mónica Núñez López, Norbert Wey, Christian Beisel, Tinu M. Thomas, Viola A Heinzelmann-Schwarz, Rudolf Aebersold, Ramona Schlenker, Charlotte K.Y. Ng, Alexandre P. A. Theocharides, Anne Bertolini, Christian Rommel, Martin Zoche, Maya D’Costa, Alva Rani James, Ulrike Lischetti, Anja Irmisch, Alexander Immer, Werner Kuebler, Stéphane Chevrier, André Kahles, Bernd Wollscheid, Mustafa A. Tuncel, Philip M. Jermann, Sandra Weber, Emanuela S. Milani, Natascha Santacroce, Julien Mena, Markus Tolnay, Natalie R. Davidson, Pirmin Haeuptle, Nora C. Toussaint, Vinko Toševski, Jacobo Sarabia del Castillo, William P. Weber, Sebastian Lugert, Tanmay Tanna, Daniel J. Stekhoven, Natalie Rimmer, Monica-Andreea Baciu-Drăgan, Reinhard Dummer, Philipp Markolin, Christian Kurzeder, Byron Calgua, Joanna Ficek-Pascual, Simon Heinke, Pedro Ferreira, Katharina Jahn, Mitchell P. Levesque, Marta Nowak, Riccardo Murri, Gerd Maass, Marc Zimmermann, Tim M. Jaeger, Mathilde Ritter, Silvana I. Albert, Mara L. Rosano-González
article en
6 citations

Abstract

Multiplexed protein imaging offers valuable insights into interactions between tumours and their surrounding tumour microenvironment, but its widespread use is limited by cost, time and tissue availability. Here we present HistoPlexer, a deep learning framework that generates spatially resolved protein multiplexes directly from standard haematoxylin and eosin (H&E) histopathology images. HistoPlexer jointly predicts multiple tumour and immune markers using a conditional generative adversarial architecture with custom loss functions designed to ensure pixel- and embedding-level similarity while mitigating slice-to-slice variations. A comprehensive evaluation of metastatic melanoma samples demonstrates that HistoPlexer-generated protein maps closely resemble real maps, as validated by expert assessment. They preserve crucial biological relationships by capturing spatial co-localization patterns among proteins. The spatial distribution of immune infiltration from HistoPlexer-generated protein multiplex enables stratification of tumours into immune subtypes. In an independent cohort, integration of HistoPlexer-derived features into predictive models enhances performance in survival prediction and immune subtype classification compared to models using H&E features alone. To assess broader applicability, we benchmarked HistoPlexer on publicly available pixel-aligned datasets from different cancer types. In all settings, HistoPlexer consistently outperformed baseline methods, demonstrating robustness across diverse tissue types and imaging conditions. By enabling whole-slide protein multiplex generation from routine H&E images, HistoPlexer offers a cost- and time-efficient approach to tumour microenvironment characterization with strong potential to advance precision oncology.

Nature Machine IntelligenceVol. 7(8)
Roche (Switzerland) (CH), University of Bern (CH), SIB Swiss Institute of Bioinformatics (CH), University of Basel (CH), University of Zurich (CH), University Hospital of Basel (CH), ETH Zurich (CH), Institute for Biomedical Engineering (CH), Roche Pharma AG (Germany) (DE), Integrated Oncology (United States) (US), University Hospital of Zurich (CH), Zurich University of Teacher Education (CH), Center for Pediatric Endocrinology Zurich (CH), Swiss Data Science Center (CH), Max Planck Institute for Intelligent Systems (DE), University Hospital Cologne (DE), RWTH Aachen University (DE), FH Aachen (DE)
Openalex Percentile: Top 2%
Cell Image Analysis Techniques
9.80
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.