bims-malgli 2022-12-11 papers

bims-malgli

Biomed News

on Biology of malignant gliomas

Issue of 2022‒12‒11
eleven papers selected by
Oltea Sampetrean
Keio University

Aporphine and isoquinoline derivatives block glioblastoma cell stemness and enhance temozolomide cytotoxicity.
Identification, validation and biological characterization of novel Glioblastoma Tumour Microenvironment subtypes: Implications for precision immunotherapy.
Use of multimodality imaging, histology, and treatment feasibility to characterize a transgenic Rag2-null rat model of glioblastoma.
The landscape of tumor cell states and spatial organization in H3-K27M mutant diffuse midline glioma across age and location.
Glioblastoma cells have increased capacity to repair radiation-induced DNA damage after migration to the olfactory bulb.
Oncogene expression from extrachromosomal DNA is driven by copy number amplification and does not require spatial clustering in glioblastoma stem cells.
Interfer(on)ing with Zika virus.
Defining the cell of origin for diffuse midline gliomas.
K27M in canonical and noncanonical H3 variants occurs in distinct oligodendroglial cell lineages in brain midline gliomas.
Examining the cooperation between extrachromosomal DNA circles.
Federated learning enables big data for rare cancer boundary detection.

Sci Rep. 2022 Dec 07. 12(1): 21113

Aporphine and isoquinoline derivatives block glioblastoma cell stemness and enhance temozolomide cytotoxicity.

Dorival Mendes Rodrigues-Junior, Cristiano Raminelli, Haifa Hassanie, Gustavo Henrique Goulart Trossini, Givago Prado Perecim, Laia Caja, Aristidis Moustakas, André Luiz Vettore.

Glioblastoma (GBM) is the most aggressive and common primary malignant brain tumor with limited available therapeutic approaches. Despite improvements in therapeutic options for GBM patients, efforts to develop new successful strategies remain as major unmet medical needs. Based on the cytotoxic properties of aporphine compounds, we evaluated the biological effect of 12 compounds obtained through total synthesis of ( ±)-apomorphine hydrochloride (APO) against GBM cells. The compounds 2,2,2-trifluoro-1-(1-methylene-3,4-dihydroisoquinolin-2(1H)-yl)ethenone (A5) and ( ±)-1-(10,11-dimethoxy-6a,7-dihydro-4H-dibenzo[de,g]quinolin-6(5H)-yl)ethenone (C1) reduced the viability of GBM cells, with 50% inhibitory concentration ranging from 18 to 48 μM in patient-derived GBM cultures. Our data show that APO, A5 or C1 modulate the expression of DNA damage and apoptotic markers, impair 3D-gliomasphere growth and reduce the expression of stemness markers. Potential activity and protein targets of A5, C1 or APO were predicted in silico based on PASS and SEA software. Dopamine receptors (DRD1 and 5), CYP2B6, CYP2C9 and ABCB1, whose transcripts were differentially expressed in the GBM cells, were among the potential A5 or C1 target proteins. Docking analyses (HQSAR and 3D-QSAR) were performed to characterize possible interactions of ABCB1 and CYP2C9 with the compounds. Notably, A5 or C1 treatment, but not temozolomide (TMZ), reduced significantly the levels of extracellular ATP, suggesting ABCB1 negative regulation, which was correlated with stronger cytotoxicity induced by the combination of TMZ with A5 or C1 on GBM cells. Hence, our data reveal a potential therapeutic application of A5 and C1 as cytotoxic agents against GBM cells and predicted molecular networks that can be further exploited to characterize the pharmacological effects of these isoquinoline-containing substances.

DOI: https://doi.org/10.1038/s41598-022-25534-2
Ann Oncol. 2022 Nov 29. pii: S0923-7534(22)04743-3. [Epub ahead of print]

Identification, validation and biological characterization of novel Glioblastoma Tumour Microenvironment subtypes: Implications for precision immunotherapy.

K White, K Connor, M Meylan, A Bougoüin, M Salvucci, F Bielle, A C O'Farrell, K Sweeney, L Weng, G Bergers, P Dicker, D M Ashley, E S Lipp, J T Low, J Zhao, P Wen, R Prins, M Verreault, A Idbaih, A Biswas, J H M Prehn, D Lambrechts, I Arijs, F Lodi, G Dilcan, M Lamfers, S Leenstra, F Fabro, I Ntafoulis, J M Kros, J Cryan, F Brett, E Quissac, A Beausang, S MacNally, P O'Halloran, J Clerkin, O Bacon, A Kremer, R Tching Chi Yen, F S Varn, R G W Verhaak, C Sautès-Fridman, W H Fridman, A T Byrne.

  BACKGROUND: New precision medicine therapies are urgently required for glioblastoma (GBM). However, to date, efforts to subtype patients based on molecular profiles, have failed to direct treatment strategies. We hypothesized that interrogation of the GBM tumor microenvironment (TME) and identification of novel TME specific subtypes could inform new precision immunotherapy treatment strategies.METHODS: A refined and validated microenvironment cell population (MCP)-counter method was applied to >800 GBM patient tumours (GBM-MCP-counter). Specifically, partition arounds medoids (PAM) clustering of GBM-MCP-counter scores in the GLIOTRAIN discovery cohort identified 3 novel patient clusters, uniquely characterised by TME-composition, functional orientation markers and immune checkpoint proteins. Validation was performed in three independent GBM-RNA-seq datasets. Neo-antigen, mutational, and gene ontology analysis identified mutations and uniquely altered pathways across subtypes. The longitudinal GLASS cohort and three immunotherapy clinical trial cohorts (treatment with neoadjuvant/adjuvant anti-PD1 or PSVRIPO) were further interrogated to assess subtype alterations between primary and recurrent tumours, and to assess the utility of TME classifiers as immunotherapy biomarkers.
RESULTS: TMEHigh tumours (30%) displayed elevated lymphocyte, myeloid cell immune checkpoint, PDCD1(PD1) and CTLA4 transcripts. TMEHigh/mesenchymal+ patients featured tertiary lymphoid structures (TLS). TMEMed (46%) tumours were enriched for endothelial cell gene expression profiles and displayed heterogeneous immune populations. TMELow (24%) tumours were manifest as an 'immune-desert' group. TME-subtype transitions upon recurrence were identified in the longitudinal GLASS cohort. Assessment of GBM immunotherapy trial datasets revealed that TMEHigh patients receiving neo-adjuvant anti-PD1 had significantly increased OS (P=0.04). Moreover, TMEHigh patients treated with adjuvant anti-PD1 or oncolytic virus (PVSRIPO), showed a trend towards improved survival.
CONCLUSIONS: We have established a novel TME-based classification system for application in intracranial malignancies. TME-subtypes represent canonical "termini a quo" (starting points) to support an improved precision immunotherapy treatment approach.

Keywords:  IDH-wt Glioblastoma; Immunotherapy; Precision therapy; Subtypes; Tumour Microenvironment

DOI:  https://doi.org/10.1016/j.annonc.2022.11.008
Front Oncol. 2022 ;12 939260

Use of multimodality imaging, histology, and treatment feasibility to characterize a transgenic Rag2-null rat model of glioblastoma.

Luke R Jackson, Megan R Masi, Bryce M Selman, George E Sandusky, Hamideh Zarrinmayeh, Sudip K Das, Surendra Maharjan, Nian Wang, Qi-Huang Zheng, Karen E Pollok, Scott E Snyder, Phillip Zhe Sun, Gary D Hutchins, Elizabeth R Butch, Michael C Veronesi.

  Many drugs that show potential in animal models of glioblastoma (GBM) fail to translate to the clinic, contributing to a paucity of new therapeutic options. In addition, animal model development often includes histologic assessment, but multiparametric/multimodality imaging is rarely included despite increasing utilization in patient cancer management. This study developed an intracranial recurrent, drug-resistant, human-derived glioblastoma tumor in Sprague-Dawley Rag2-Rag2 tm1Hera knockout rat and was characterized both histologically and using multiparametric/multimodality neuroimaging. Hybrid 18F-fluoroethyltyrosine positron emission tomography and magnetic resonance imaging, including chemical exchange saturation transfer (18F-FET PET/CEST MRI), was performed for full tumor viability determination and characterization. Histological analysis demonstrated human-like GBM features of the intracranially implanted tumor, with rapid tumor cell proliferation (Ki67 positivity: 30.5 ± 7.8%) and neovascular heterogeneity (von Willebrand factor VIII:1.8 to 5.0% positivity). Early serial MRI followed by simultaneous 18F-FET PET/CEST MRI demonstrated consistent, predictable tumor growth, with exponential tumor growth most evident between days 35 and 49 post-implantation. In a second, larger cohort of rats, 18F-FET PET/CEST MRI was performed in mature tumors (day 49 post-implantation) for biomarker determination, followed by evaluation of single and combination therapy as part of the model development and validation. The mean percentage of the injected dose per mL of 18F-FET PET correlated with the mean %CEST (r = 0.67, P < 0.05), but there was also a qualitative difference in hot spot location within the tumor, indicating complementary information regarding the tumor cell demand for amino acids and tumor intracellular mobile phase protein levels. Finally, the use of this glioblastoma animal model for therapy assessment was validated by its increased overall survival after treatment with combination therapy (temozolomide and idasanutlin) (P < 0.001). Our findings hold promise for a more accurate tumor viability determination and novel therapy assessment in vivo in a recently developed, reproducible, intracranial, PDX GBM.

Keywords:  CEST MRI; amide proton transfer (APT) imaging; amino acid PET; glioblastoma; hybrid PET/MRI; magnetic resonance imaging (MRI); positron emission tomography (PET)

DOI:  https://doi.org/10.3389/fonc.2022.939260
Nat Genet. 2022 Dec;54(12): 1881-1894

The landscape of tumor cell states and spatial organization in H3-K27M mutant diffuse midline glioma across age and location.

Ilon Liu, Li Jiang, Erik R Samuelsson, Sergio Marco Salas, Alexander Beck, Olivia A Hack, Daeun Jeong, McKenzie L Shaw, Bernhard Englinger, Jenna LaBelle, Hafsa M Mire, Sibylle Madlener, Lisa Mayr, Michael A Quezada, Maria Trissal, Eshini Panditharatna, Kati J Ernst, Jayne Vogelzang, Taylor A Gatesman, Matthew E Halbert, Hana Palova, Petra Pokorna, Jaroslav Sterba, Ondrej Slaby, Rene Geyeregger, Aaron Diaz, Izac J Findlay, Matthew D Dun, Adam Resnick, Mario L Suvà, David T W Jones, Sameer Agnihotri, Jessica Svedlund, Carl Koschmann, Christine Haberler, Thomas Czech, Irene Slavc, Jennifer A Cotter, Keith L Ligon, Sanda Alexandrescu, W K Alfred Yung, Isabel Arrillaga-Romany, Johannes Gojo, Michelle Monje, Mats Nilsson, Mariella G Filbin.

Histone 3 lysine27-to-methionine (H3-K27M) mutations most frequently occur in diffuse midline gliomas (DMGs) of the childhood pons but are also increasingly recognized in adults. Their potential heterogeneity at different ages and midline locations is vastly understudied. Here, through dissecting the single-cell transcriptomic, epigenomic and spatial architectures of a comprehensive cohort of patient H3-K27M DMGs, we delineate how age and anatomical location shape glioma cell-intrinsic and -extrinsic features in light of the shared driver mutation. We show that stem-like oligodendroglial precursor-like cells, present across all clinico-anatomical groups, display varying levels of maturation dependent on location. We reveal a previously underappreciated relationship between mesenchymal cancer cell states and age, linked to age-dependent differences in the immune microenvironment. Further, we resolve the spatial organization of H3-K27M DMG cell populations and identify a mitotic oligodendroglial-lineage niche. Collectively, our study provides a powerful framework for rational modeling and therapeutic interventions.

DOI: https://doi.org/10.1038/s41588-022-01236-3
Cancer Cell Int. 2022 Dec 08. 22(1): 389

Glioblastoma cells have increased capacity to repair radiation-induced DNA damage after migration to the olfactory bulb.

Charlotte Degorre, Ian C Sutton, Stacey L Lehman, Uma T Shankavaram, Kevin Camphausen, Philip J Tofilon.

  BACKGROUND: The invasive nature of GBM combined with the diversity of brain microenvironments creates the potential for a topographic heterogeneity in GBM radioresponse. Investigating the mechanisms responsible for a microenvironment-induced differential GBM response to radiation may provide insights into the molecules and processes mediating GBM radioresistance.METHODS: Using a model system in which human GBM stem-like cells implanted into the right striatum of nude mice migrate throughout the right hemisphere (RH) to the olfactory bulb (OB), the radiation-induced DNA damage response was evaluated in each location according to γH2AX and 53BP1 foci and cell cycle phase distribution as determined by flow cytometry and immunohistochemistry. RNAseq was used to compare transcriptomes of tumor cells growing in the OB and the RH. Protein expression and neuron-tumor interaction were defined by immunohistochemistry and confocal microscopy.
RESULTS: After irradiation, there was a more rapid dispersal of γH2AX and 53BP1 foci in the OB versus in the RH, indicative of increased double strand break repair capacity in the OB and consistent with the OB providing a radioprotective niche. With respect to the cell cycle, by 6 h after irradiation there was a significant loss of mitotic tumor cells in both locations suggesting a similar activation of the G2/M checkpoint. However, by 24 h post-irradiation there was an accumulation of G2 phase cells in the OB, which continued out to at least 96 h. Transcriptome analysis showed that tumor cells in the OB had higher expression levels of DNA repair genes involved in non-homologous end joining and genes related to the spindle assembly checkpoint. Tumor cells in the OB were also found to have an increased frequency of soma-soma contact with neurons.
CONCLUSION: GBM cells that have migrated to the OB have an increased capacity to repair radiation-induced double strand breaks and altered cell cycle regulation. These results correspond to an upregulation of genes involved in DNA damage repair and cell cycle control. Because the murine OB provides a source of radioresistant tumor cells not evident in other experimental systems, it may serve as a model for investigating the mechanisms mediating GBM radioresistance.

Keywords:  Glioblastoma; Microenvironment; Radioresistance

DOI:  https://doi.org/10.1186/s12935-022-02819-0
Elife. 2022 Dec 07. pii: e80207. [Epub ahead of print]11

Oncogene expression from extrachromosomal DNA is driven by copy number amplification and does not require spatial clustering in glioblastoma stem cells.

Karin Purshouse, Elias T Friman, Shelagh Boyle, Pooran Singh Dewari, Vivien Grant, Alhafidz Hamdan, Gillian M Morrison, Paul M Brennan, Sjoerd V Beentjes, Steven M Pollard, Wendy A Bickmore.

  Extrachromosomal DNA (ecDNA) are frequently observed in human cancers and are responsible for high levels of oncogene expression. In glioblastoma (GBM), ecDNA copy number correlates with poor prognosis. It is hypothesized that their copy number, size, and chromatin accessibility facilitate clustering of ecDNA and colocalization with transcriptional hubs, and that this underpins their elevated transcriptional activity. Here, we use super-resolution imaging and quantitative image analysis to evaluate GBM stem cells harbouring distinct ecDNA species (EGFR, CDK4, PDGFRA). We find no evidence that ecDNA routinely cluster with one another or closely interact with transcriptional hubs. Cells with EGFR-containing ecDNA have increased EGFR transcriptional output, but transcription per gene copy is similar in ecDNA compared to the endogenous chromosomal locus. These data suggest that it is the increased copy number of oncogene-harbouring ecDNA that primarily drives high levels of oncogene transcription, rather than specific interactions of ecDNA with each other or with high concentrations of the transcriptional machinery.

Keywords:  EGFR; chromosomes; extrachromosomal DNA; gene expression; human; oncogenes; super-resolution imaging; transcription

DOI:  https://doi.org/10.7554/eLife.80207
Neuron. 2022 Dec 07. pii: S0896-6273(22)00960-6. [Epub ahead of print]110(23): 3853-3854

Interfer(on)ing with Zika virus.

Alvaro G Alvarado, Harley I Kornblum.

In this issue of Neuron, Bulstrode et al.1 demonstrate that glioblastoma slice cultures, unlike neural progenitors, are refractory to Zika virus infection. The anti-infective mechanism is myeloid-lineage cell-secreted interferon beta. These studies have implications for therapeutics in both glioblastoma and Zika virus infections.

DOI: https://doi.org/10.1016/j.neuron.2022.10.027
Nat Genet. 2022 Dec;54(12): 1770-1771

Defining the cell of origin for diffuse midline gliomas.

Xiao-Nan Li.

DOI: https://doi.org/10.1038/s41588-022-01231-8
Nat Genet. 2022 Dec;54(12): 1865-1880

K27M in canonical and noncanonical H3 variants occurs in distinct oligodendroglial cell lineages in brain midline gliomas.

Selin Jessa, Abdulshakour Mohammadnia, Ashot S Harutyunyan, Maud Hulswit, Srinidhi Varadharajan, Hussein Lakkis, Nisha Kabir, Zahedeh Bashardanesh, Steven Hébert, Damien Faury, Maria C Vladoiu, Samantha Worme, Marie Coutelier, Brian Krug, Augusto Faria Andrade, Manav Pathania, Andrea Bajic, Alexander G Weil, Benjamin Ellezam, Jeffrey Atkinson, Roy W R Dudley, Jean-Pierre Farmer, Sebastien Perreault, Benjamin A Garcia, Valérie Larouche, Mathieu Blanchette, Livia Garzia, Aparna Bhaduri, Keith L Ligon, Pratiti Bandopadhayay, Michael D Taylor, Stephen C Mack, Nada Jabado, Claudia L Kleinman.

Canonical (H3.1/H3.2) and noncanonical (H3.3) histone 3 K27M-mutant gliomas have unique spatiotemporal distributions, partner alterations and molecular profiles. The contribution of the cell of origin to these differences has been challenging to uncouple from the oncogenic reprogramming induced by the mutation. Here, we perform an integrated analysis of 116 tumors, including single-cell transcriptome and chromatin accessibility, 3D chromatin architecture and epigenomic profiles, and show that K27M-mutant gliomas faithfully maintain chromatin configuration at developmental genes consistent with anatomically distinct oligodendrocyte precursor cells (OPCs). H3.3K27M thalamic gliomas map to prosomere 2-derived lineages. In turn, H3.1K27M ACVR1-mutant pontine gliomas uniformly mirror early ventral NKX6-1+/SHH-dependent brainstem OPCs, whereas H3.3K27M gliomas frequently resemble dorsal PAX3+/BMP-dependent progenitors. Our data suggest a context-specific vulnerability in H3.1K27M-mutant SHH-dependent ventral OPCs, which rely on acquisition of ACVR1 mutations to drive aberrant BMP signaling required for oncogenesis. The unifying action of K27M mutations is to restrict H3K27me3 at PRC2 landing sites, whereas other epigenetic changes are mainly contingent on the cell of origin chromatin state and cycling rate.

DOI: https://doi.org/10.1038/s41588-022-01205-w
Elife. 2022 Dec 07. pii: e84639. [Epub ahead of print]11

Examining the cooperation between extrachromosomal DNA circles.

Jiahui Zhang, Cheng-Zhong Zhang.

  In a departure from previous findings, new results suggest that free-floating pieces of DNA which carry additional copies of cancer-driving genes do not tend to cluster or have increased transcription.

Keywords:  cancer; chromosomes; extrachromosomal DNA; gene expression; glioblastoma; human; oncogenes; super-resolution imaging; transcription factors hub

DOI:  https://doi.org/10.7554/eLife.84639
Nat Commun. 2022 Dec 05. 13(1): 7346

Federated learning enables big data for rare cancer boundary detection.

Sarthak Pati, Ujjwal Baid, Brandon Edwards, Micah Sheller, Shih-Han Wang, G Anthony Reina, Patrick Foley, Alexey Gruzdev, Deepthi Karkada, Christos Davatzikos, Chiharu Sako, Satyam Ghodasara, Michel Bilello, Suyash Mohan, Philipp Vollmuth, Gianluca Brugnara, Chandrakanth J Preetha, Felix Sahm, Klaus Maier-Hein, Maximilian Zenk, Martin Bendszus, Wolfgang Wick, Evan Calabrese, Jeffrey Rudie, Javier Villanueva-Meyer, Soonmee Cha, Madhura Ingalhalikar, Manali Jadhav, Umang Pandey, Jitender Saini, John Garrett, Matthew Larson, Robert Jeraj, Stuart Currie, Russell Frood, Kavi Fatania, Raymond Y Huang, Ken Chang, Carmen Balaña Quintero, Jaume Capellades, Josep Puig, Johannes Trenkler, Josef Pichler, Georg Necker, Andreas Haunschmidt, Stephan Meckel, Gaurav Shukla, Spencer Liem, Gregory S Alexander, Joseph Lombardo, Joshua D Palmer, Adam E Flanders, Adam P Dicker, Haris I Sair, Craig K Jones, Archana Venkataraman, Meirui Jiang, Tiffany Y So, Cheng Chen, Pheng Ann Heng, Qi Dou, Michal Kozubek, Filip Lux, Jan Michálek, Petr Matula, Miloš Keřkovský, Tereza Kopřivová, Marek Dostál, Václav Vybíhal, Michael A Vogelbaum, J Ross Mitchell, Joaquim Farinhas, Joseph A Maldjian, Chandan Ganesh Bangalore Yogananda, Marco C Pinho, Divya Reddy, James Holcomb, Benjamin C Wagner, Benjamin M Ellingson, Timothy F Cloughesy, Catalina Raymond, Talia Oughourlian, Akifumi Hagiwara, Chencai Wang, Minh-Son To, Sargam Bhardwaj, Chee Chong, Marc Agzarian, Alexandre Xavier Falcão, Samuel B Martins, Bernardo C A Teixeira, Flávia Sprenger, David Menotti, Diego R Lucio, Pamela LaMontagne, Daniel Marcus, Benedikt Wiestler, Florian Kofler, Ivan Ezhov, Marie Metz, Rajan Jain, Matthew Lee, Yvonne W Lui, Richard McKinley, Johannes Slotboom, Piotr Radojewski, Raphael Meier, Roland Wiest, Derrick Murcia, Eric Fu, Rourke Haas, John Thompson, David Ryan Ormond, Chaitra Badve, Andrew E Sloan, Vachan Vadmal, Kristin Waite, Rivka R Colen, Linmin Pei, Murat Ak, Ashok Srinivasan, J Rajiv Bapuraj, Arvind Rao, Nicholas Wang, Ota Yoshiaki, Toshio Moritani, Sevcan Turk, Joonsang Lee, Snehal Prabhudesai, Fanny Morón, Jacob Mandel, Konstantinos Kamnitsas, Ben Glocker, Luke V M Dixon, Matthew Williams, Peter Zampakis, Vasileios Panagiotopoulos, Panagiotis Tsiganos, Sotiris Alexiou, Ilias Haliassos, Evangelia I Zacharaki, Konstantinos Moustakas, Christina Kalogeropoulou, Dimitrios M Kardamakis, Yoon Seong Choi, Seung-Koo Lee, Jong Hee Chang, Sung Soo Ahn, Bing Luo, Laila Poisson, Ning Wen, Pallavi Tiwari, Ruchika Verma, Rohan Bareja, Ipsa Yadav, Jonathan Chen, Neeraj Kumar, Marion Smits, Sebastian R van der Voort, Ahmed Alafandi, Fatih Incekara, Maarten M J Wijnenga, Georgios Kapsas, Renske Gahrmann, Joost W Schouten, Hendrikus J Dubbink, Arnaud J P E Vincent, Martin J van den Bent, Pim J French, Stefan Klein, Yading Yuan, Sonam Sharma, Tzu-Chi Tseng, Saba Adabi, Simone P Niclou, Olivier Keunen, Ann-Christin Hau, Martin Vallières, David Fortin, Martin Lepage, Bennett Landman, Karthik Ramadass, Kaiwen Xu, Silky Chotai, Lola B Chambless, Akshitkumar Mistry, Reid C Thompson, Yuriy Gusev, Krithika Bhuvaneshwar, Anousheh Sayah, Camelia Bencheqroun, Anas Belouali, Subha Madhavan, Thomas C Booth, Alysha Chelliah, Marc Modat, Haris Shuaib, Carmen Dragos, Aly Abayazeed, Kenneth Kolodziej, Michael Hill, Ahmed Abbassy, Shady Gamal, Mahmoud Mekhaimar, Mohamed Qayati, Mauricio Reyes, Ji Eun Park, Jihye Yun, Ho Sung Kim, Abhishek Mahajan, Mark Muzi, Sean Benson, Regina G H Beets-Tan, Jonas Teuwen, Alejandro Herrera-Trujillo, Maria Trujillo, William Escobar, Ana Abello, Jose Bernal, Jhon Gómez, Joseph Choi, Stephen Baek, Yusung Kim, Heba Ismael, Bryan Allen, John M Buatti, Aikaterini Kotrotsou, Hongwei Li, Tobias Weiss, Michael Weller, Andrea Bink, Bertrand Pouymayou, Hassan F Shaykh, Joel Saltz, Prateek Prasanna, Sampurna Shrestha, Kartik M Mani, David Payne, Tahsin Kurc, Enrique Pelaez, Heydy Franco-Maldonado, Francis Loayza, Sebastian Quevedo, Pamela Guevara, Esteban Torche, Cristobal Mendoza, Franco Vera, Elvis Ríos, Eduardo López, Sergio A Velastin, Godwin Ogbole, Mayowa Soneye, Dotun Oyekunle, Olubunmi Odafe-Oyibotha, Babatunde Osobu, Mustapha Shu'aibu, Adeleye Dorcas, Farouk Dako, Amber L Simpson, Mohammad Hamghalam, Jacob J Peoples, Ricky Hu, Anh Tran, Danielle Cutler, Fabio Y Moraes, Michael A Boss, James Gimpel, Deepak Kattil Veettil, Kendall Schmidt, Brian Bialecki, Sailaja Marella, Cynthia Price, Lisa Cimino, Charles Apgar, Prashant Shah, Bjoern Menze, Jill S Barnholtz-Sloan, Jason Martin, Spyridon Bakas.

Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.

DOI: https://doi.org/10.1038/s41467-022-33407-5