France Rose
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
France Rose has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Katarzyna Bozek
Data science of Bioimages Lab, Faculty of Medicine and University Hospital Cologne, Center for Molecular Medicine Cologne (CMMC), University of Cologne, Cologne, Germany
3 shared publications - Liam O'SHaughnessy
Department of Physics and Astronomy, VU University Amsterdam, Amsterdam, The Netherlands
3 shared publications Monika Michaluk
Faculty of Mathematics, Informatics and Mechanics, University of Warsaw, Warsaw, Poland
3 shared publicationsTimon Blindauer
Institute for Biomedical Informatics, Faculty of Medicine and University Hospital Cologne, Cologne, Germany
3 shared publicationsBogna M. Ignatowska-Jankowska
Neuronal Rhythms in Movement Unit, Okinawa Institute of Science and Technology, Onna-son, Japan
2 shared publications- Greg J. Stephens
Department of Physics and Astronomy, VU University Amsterdam, Amsterdam, The Netherlands
2 shared publications
Latest Publications
- 3 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Analysis of phenotypic and spatial cellular heterogeneity from large scale microscopy data
Analyse de l'hétérogénéité phénotypique et spatiale des cellules à partir de grandes données de microscopie: La robotique et l’automatisation des microscopes ont ouvert la voie aux cribles cellulaires à haut contenu : des marqueurs fluorescents ciblant l’ADN ou d’autres composants sont utilisés pour imager des centaines de milliers de cellules dans différentes conditions. Il a été montré que les cribles cellulaires sont efficaces pour découvrir des médicaments de nouvelles classes thérapeutiques, cad ceux qui agissent sur une nouvelle cible. Les cribles permettent d’identifier des composés prometteurs et de les caractériser en leur associant des annotations fonctionnelles, comme leur cible moléculaire ou leur mécanisme d’action (MOA).J’ai étudié l'hétérogénéité des réponses cellulaires à différents niveaux et comment cette hétérogénéité phénotypique peut être exploitée pour mieux caractériser les composés. Au premier niveau, j’ai étudié l'hétérogénéité entre patients. Nous avons montré qu’utiliser différentes lignées cellulaires dérivées de patients augmente la probabilité de prédire la cible moléculaire du composé testé. Le second niveau correspond à la diversité des réponses cellulaires de la même lignée cellulaire soumise au même traitement. Des méthodes de clustering appropriées peuvent être utilisées pour clarifier cette complexité et pour grouper les cellules en sous-populations. Les proportions de chaque sous-population par traitement permettent de prédire le bon MOA. Le troisième niveau regarde comment les sous-populations cellulaires sont organisées spatialement. J’ai trouvé que les cellules voisines s’influencent les unes les autres et affichent un phénotype similaire plus fréquemment qu’attendu par chance. Ces résultats obtenus sur une centaine de traitements montrent que des cellules génétiquement identiques ne sont pas identiques et indépendantes mais sont à l’origine d’une hétérogénéité spatiale par le lignage cellulaire et les interactions. En utilisant l’information spatiale ainsi que l'hétérogénéité phénotypique, les méthodes à noyaux de graphes améliorent la classification en MOA sous certaines conditions. Parallèlement, comme l’analyse spatiale peut s’appliquer à n’importe quelle image de microscopie, j’ai développé une librairie d’analyse Python, PySpacell, pour étudier l’aléatoire spatial de marqueurs quantitatifs et qualitatifs.
- 3 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Analysis of phenotypic and spatial cellular heterogeneity from large scale microscopy data
Analyse de l'hétérogénéité phénotypique et spatiale des cellules à partir de grandes données de microscopie: La robotique et l’automatisation des microscopes ont ouvert la voie aux cribles cellulaires à haut contenu : des marqueurs fluorescents ciblant l’ADN ou d’autres composants sont utilisés pour imager des centaines de milliers de cellules dans différentes conditions. Il a été montré que les cribles cellulaires sont efficaces pour découvrir des médicaments de nouvelles classes thérapeutiques, cad ceux qui agissent sur une nouvelle cible. Les cribles permettent d’identifier des composés prometteurs et de les caractériser en leur associant des annotations fonctionnelles, comme leur cible moléculaire ou leur mécanisme d’action (MOA).J’ai étudié l'hétérogénéité des réponses cellulaires à différents niveaux et comment cette hétérogénéité phénotypique peut être exploitée pour mieux caractériser les composés. Au premier niveau, j’ai étudié l'hétérogénéité entre patients. Nous avons montré qu’utiliser différentes lignées cellulaires dérivées de patients augmente la probabilité de prédire la cible moléculaire du composé testé. Le second niveau correspond à la diversité des réponses cellulaires de la même lignée cellulaire soumise au même traitement. Des méthodes de clustering appropriées peuvent être utilisées pour clarifier cette complexité et pour grouper les cellules en sous-populations. Les proportions de chaque sous-population par traitement permettent de prédire le bon MOA. Le troisième niveau regarde comment les sous-populations cellulaires sont organisées spatialement. J’ai trouvé que les cellules voisines s’influencent les unes les autres et affichent un phénotype similaire plus fréquemment qu’attendu par chance. Ces résultats obtenus sur une centaine de traitements montrent que des cellules génétiquement identiques ne sont pas identiques et indépendantes mais sont à l’origine d’une hétérogénéité spatiale par le lignage cellulaire et les interactions. En utilisant l’information spatiale ainsi que l'hétérogénéité phénotypique, les méthodes à noyaux de graphes améliorent la classification en MOA sous certaines conditions. Parallèlement, comme l’analyse spatiale peut s’appliquer à n’importe quelle image de microscopie, j’ai développé une librairie d’analyse Python, PySpacell, pour étudier l’aléatoire spatial de marqueurs quantitatifs et qualitatifs.
- Open Access
- 4 December 2025
- Published by Springer Nature in Nature Methods
- Vol. 23 (1) , 236-247
Deep Imputation for Skeleton data (DISK) for behavioral science
Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. While animal behavior experiments are expensive and complex, tracking errors sometimes make large portions of the experimental data unusable. Here our deep learning method, Deep Imputation for Skeleton data (DISK), uncovers dependencies between keypoints and their dynamics to impute missing tracking data without the help of any manual annotations. We demonstrate the utility and performance of DISK on seven animal skeletons including multi-animal setups. The imputed recordings allow us to detect more episodes of motion, such as steps, and obtain more statistically robust results when comparing these episodes between experimental conditions. In addition, by learning to impute the missing content, DISK learns meaningful representations of the data capturing, for example, underlying actions. This stand-alone imputation package, available at https://github.com/bozeklab/DISK.git/ , is applicable to outputs of tracking methods (marker-based or markerless) and allows for varied types of downstream analysis.
- 4 December 2025
- Published by Springer Nature in Nature Methods
- Vol. 23 (1) , 236-247
Open AccessDeep Imputation for Skeleton data (DISK) for behavioral science
Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. While animal behavior experiments are expensive and complex, tracking errors sometimes make large portions of the experimental data unusable. Here our deep learning method, Deep Imputation for Skeleton data (DISK), uncovers dependencies between keypoints and their dynamics to impute missing tracking data without the help of any manual annotations. We demonstrate the utility and performance of DISK on seven animal skeletons including multi-animal setups. The imputed recordings allow us to detect more episodes of motion, such as steps, and obtain more statistically robust results when comparing these episodes between experimental conditions. In addition, by learning to impute the missing content, DISK learns meaningful representations of the data capturing, for example, underlying actions. This stand-alone imputation package, available at https://github.com/bozeklab/DISK.git/ , is applicable to outputs of tracking methods (marker-based or markerless) and allows for varied types of downstream analysis.
- Open Access
- 8 September 2025
- Published by Springer Nature in BMC Research Notes
- Vol. 18 (1) , 1-4
Imaging mass cytometry dataset of small-cell lung cancer tumors and tumor microenvironments
Objectives: Small cell lung cancer (SCLC) accounts for approximately 15% of lung tumors and is marked by aggressive growth and early metastatic spread. In this study, we used two SCLC mouse models with differing tumor mutation burdens (TMB). To investigate tumor composition, spatial architecture, and interactions with the surrounding microenvironment, we acquired multiplexed images of mouse lung tumors using imaging mass cytometry (IMC). These data build upon a previously published characterization of the mouse model. Data description: After tumor detection, mice were assigned to one of five treatment groups. Lung tumor tissues were imaged with a 37-marker IMC panel designed to identify major cell types—tumor, immune, and structural—as well as their functional states. When possible, each tumor was sampled both at its center and border regions. Tumor masks in the form of binary images are provided to delineate tumor areas. Additional metadata include tumor onset and endpoint dates to support downstream correlation or predictive analyses based on the image data. This dataset offers a valuable resource for studying the histological and cellular complexity of SCLC in a genetically controlled mouse model across multiple therapeutic conditions.
- 8 September 2025
- Published by Springer Nature in BMC Research Notes
- Vol. 18 (1) , 1-4
Open AccessImaging mass cytometry dataset of small-cell lung cancer tumors and tumor microenvironments
Objectives: Small cell lung cancer (SCLC) accounts for approximately 15% of lung tumors and is marked by aggressive growth and early metastatic spread. In this study, we used two SCLC mouse models with differing tumor mutation burdens (TMB). To investigate tumor composition, spatial architecture, and interactions with the surrounding microenvironment, we acquired multiplexed images of mouse lung tumors using imaging mass cytometry (IMC). These data build upon a previously published characterization of the mouse model. Data description: After tumor detection, mice were assigned to one of five treatment groups. Lung tumor tissues were imaged with a 37-marker IMC panel designed to identify major cell types—tumor, immune, and structural—as well as their functional states. When possible, each tumor was sampled both at its center and border regions. Tumor masks in the form of binary images are provided to delineate tumor areas. Additional metadata include tumor onset and endpoint dates to support downstream correlation or predictive analyses based on the image data. This dataset offers a valuable resource for studying the histological and cellular complexity of SCLC in a genetically controlled mouse model across multiple therapeutic conditions.
- 19 May 2025
- Published by EMBL-EBI
Imaging Mass Cytometry Dataset of Small-Cell Lung Cancer tumors and tumor environments
BioStudies – one package for all the data supporting a study
- 19 May 2025
- Published by EMBL-EBI
Imaging Mass Cytometry Dataset of Small-Cell Lung Cancer tumors and tumor environments
BioStudies – one package for all the data supporting a study
- Open Access
- 20 May 2024
- Published by Springer Nature in Research Square
Deep Imputation for Skeleton Data (DISK) for Behavioral Science
Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. However, these methods are not perfect and contain missing data. Our method, Deep Imputation for Skeleton data (DISK), leverages deep learning algorithms to learn dependencies between keypoints and their dynamics to impute missing tracking data. We developed an unsupervised training scheme, which does not rely on manual annotations, and tested several neural network architectures for the imputation task. We found that transformer outperforms other architectures including graph convolutional networks that were developed specifically for skeleton-based action recognition. We demonstrate the usability and performance of our imputation method on seven different animal skeletons including two multi-animal set-ups. With an optional estimated imputation error, DISK enables behavior scientists to assess the reliability of the imputed data. The imputed recordings allow to detect more episodes of motion, such as steps, and to obtain more statistically robust results when comparing these episodes between experimental conditions. While animal behavior experiments are expensive and complex, tracking errors make sometimes large portions of the experimental data unusable. DISK allows for filling in the missing information and for taking full advantage of the rich behavioral data. This stand-alone imputation package, freely available at https://github.com/bozeklab/DISK.git, is applicable to results of any tracking method (marker-based or markerless) and allows for any type of downstream analysis.
- 20 May 2024
- Published by Springer Nature in Research Square
Open AccessDeep Imputation for Skeleton Data (DISK) for Behavioral Science
Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. However, these methods are not perfect and contain missing data. Our method, Deep Imputation for Skeleton data (DISK), leverages deep learning algorithms to learn dependencies between keypoints and their dynamics to impute missing tracking data. We developed an unsupervised training scheme, which does not rely on manual annotations, and tested several neural network architectures for the imputation task. We found that transformer outperforms other architectures including graph convolutional networks that were developed specifically for skeleton-based action recognition. We demonstrate the usability and performance of our imputation method on seven different animal skeletons including two multi-animal set-ups. With an optional estimated imputation error, DISK enables behavior scientists to assess the reliability of the imputed data. The imputed recordings allow to detect more episodes of motion, such as steps, and to obtain more statistically robust results when comparing these episodes between experimental conditions. While animal behavior experiments are expensive and complex, tracking errors make sometimes large portions of the experimental data unusable. DISK allows for filling in the missing information and for taking full advantage of the rich behavioral data. This stand-alone imputation package, freely available at https://github.com/bozeklab/DISK.git, is applicable to results of any tracking method (marker-based or markerless) and allows for any type of downstream analysis.
- Open Access
Deep Imputation for Skeleton Data (DISK) for Behavioral Science
Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. However, these methods are not perfect and contain missing data. Our method, Deep Imputation for Skeleton data (DISK), leverages deep learning algorithms to learn dependencies between keypoints and their dynamics to impute missing tracking data. We developed an unsupervised training scheme, which does not rely on manual annotations, and tested several neural network architectures for the imputation task. We found that transformer outperforms other architectures including graph convolutional networks that were developed specifically for skeleton-based action recognition. We demonstrate the usability and performance of our imputation method on seven different animal skeletons including two multi-animal set-ups. With an optional estimated imputation error, DISK enables behavior scientists to assess the reliability of the imputed data. The imputed recordings allow to detect more episodes of motion, such as steps, and to obtain more statistically robust results when comparing these episodes between experimental conditions. While animal behavior experiments are expensive and complex, tracking errors make sometimes large portions of the experimental data unusable. DISK allows for filling in the missing information and for taking full advantage of the rich behavioral data. This stand-alone imputation package, freely available at https://github.com/bozeklab/DISK.git, is applicable to results of any tracking method (marker-based or markerless) and allows for any type of downstream analysis.
Open AccessDeep Imputation for Skeleton Data (DISK) for Behavioral Science
Pose estimation methods and motion capture systems have opened doors to quantitative measurements of animal kinematics. However, these methods are not perfect and contain missing data. Our method, Deep Imputation for Skeleton data (DISK), leverages deep learning algorithms to learn dependencies between keypoints and their dynamics to impute missing tracking data. We developed an unsupervised training scheme, which does not rely on manual annotations, and tested several neural network architectures for the imputation task. We found that transformer outperforms other architectures including graph convolutional networks that were developed specifically for skeleton-based action recognition. We demonstrate the usability and performance of our imputation method on seven different animal skeletons including two multi-animal set-ups. With an optional estimated imputation error, DISK enables behavior scientists to assess the reliability of the imputed data. The imputed recordings allow to detect more episodes of motion, such as steps, and to obtain more statistically robust results when comparing these episodes between experimental conditions. While animal behavior experiments are expensive and complex, tracking errors make sometimes large portions of the experimental data unusable. DISK allows for filling in the missing information and for taking full advantage of the rich behavioral data. This stand-alone imputation package, freely available at https://github.com/bozeklab/DISK.git, is applicable to results of any tracking method (marker-based or markerless) and allows for any type of downstream analysis.
- 8 October 2021
- Published by Springer Nature in Human-Intelligent Systems Integration
- Vol. 3 (4) , 263-275
Enhancing human-machine teaming for medical prognosis through neural ordinary differential equations (NODEs)
Machine learning (ML) has recently been demonstrated to rival expert-level human accuracy in prediction and detection tasks in a variety of domains, including medicine. Despite these impressive findings, however, a key barrier to the full realization of ML’s potential in medical prognoses is technology acceptance. Recent efforts to produce explainable AI (XAI) have made progress in improving the interpretability of some ML models, but these efforts suffer from limitations intrinsic to their design: they work best at identifying why a system fails, but do poorly at explaining when and why a model’s prediction is correct. We posit that the acceptability of ML predictions in expert domains is limited by two key factors: the machine’s horizon of prediction that extends beyond human capability, and the inability for machine predictions to incorporate human intuition into their models. We propose the use of a novel ML architecture, Neural Ordinary Differential Equations (NODEs) to enhance human understanding and encourage acceptability. Our approach prioritizes human cognitive intuition at the center of the algorithm design, and offers a distribution of predictions rather than single outputs. We explain how this approach may significantly improve human-machine collaboration in prediction tasks in expert domains such as medical prognoses. We propose a model and demonstrate, by expanding a concrete example from the literature, how our model advances the vision of future hybrid human-AI systems.
- 8 October 2021
- Published by Springer Nature in Human-Intelligent Systems Integration
- Vol. 3 (4) , 263-275
Enhancing human-machine teaming for medical prognosis through neural ordinary differential equations (NODEs)
Machine learning (ML) has recently been demonstrated to rival expert-level human accuracy in prediction and detection tasks in a variety of domains, including medicine. Despite these impressive findings, however, a key barrier to the full realization of ML’s potential in medical prognoses is technology acceptance. Recent efforts to produce explainable AI (XAI) have made progress in improving the interpretability of some ML models, but these efforts suffer from limitations intrinsic to their design: they work best at identifying why a system fails, but do poorly at explaining when and why a model’s prediction is correct. We posit that the acceptability of ML predictions in expert domains is limited by two key factors: the machine’s horizon of prediction that extends beyond human capability, and the inability for machine predictions to incorporate human intuition into their models. We propose the use of a novel ML architecture, Neural Ordinary Differential Equations (NODEs) to enhance human understanding and encourage acceptability. Our approach prioritizes human cognitive intuition at the center of the algorithm design, and offers a distribution of predictions rather than single outputs. We explain how this approach may significantly improve human-machine collaboration in prediction tasks in expert domains such as medical prognoses. We propose a model and demonstrate, by expanding a concrete example from the literature, how our model advances the vision of future hybrid human-AI systems.