Abstract Introduction Cellular differentiation and lineage commitment are known to be associated with differences in DNA methylation. Leiomyosarcoma (LMS) is a tumor thought to originate from smooth muscle cells in the walls of vessels in the soft tissue (STLMS) or from the uterine myometrium (ULMS). Here, we identify the methylation signatures of normal smooth muscle cells from blood vessels and the uterine wall and compare these with those found in STLMS and ULMS. We hypothesized that these methylation signatures could be used to assign a smooth muscle subtype of origin to individual leiomyosarcomas, and that tumors of different origin would show biological differences with potential therapeutic relevance. Methods To define methylation profiles for smooth muscle from vessel walls versus those found in myometrium, EPIC methylation profiling was performed on DNA from 49 formalin-fixed paraffin-embedded (FFPE) normal smooth muscle samples. A supervised machine learning algorithm (Random Forest) was used to distinguish the methylation patterns of normal smooth muscle cells in vessel walls from those in the myometrium. The resulting classifier was applied to methylation data on 67 cases of LMS with corresponding bulk RNAseq data to identify which tumors showed a methylation signature most consistent with either blood vessel wall (LMS vessel ) or myometrial smooth muscle (LMS wall ). A custom signature matrix derived from scRNAseq data from 6 samples of LMS was used in CIBERSORTx analysis to compare the cellular composition of LMS cases with a vessel or uterine wall methylation signature. Results A high degree of correlation was found between the known site of origin for LMS (STLMS vs ULMS) and the methylation signature derived from different types of normal smooth muscle. LMS wall tumors compared to LMS vessel tumors had significantly higher activation of the PD-1 checkpoint pathway in RNAseq analysis. Digital flow cytometry by CIBERSORTx analysis showed an increased expression of transcriptomic signatures of several immune cell subtypes in LMS vessel tumors. Conclusion Using a supervised machine learning approach we classified LMS samples as either showing a high similarity in methylation patterns to normal smooth muscle cells of either the vessel wall or the myometrium. We found a correlation between LMS showing either a “vessel” or “muscle wall” methylation signature and their site of origin, but notably we also identified some exceptions. When classified based on their methylation signature LMS wall and LMS vessel differed in their PD-1 pathway activation and in their predicted immune cell populations, suggesting potential implications for immunotherapeutic approaches.

Prognostic performance of mitotic index, DL score, and MJ risk models in cohorts C1 to C3.

Kaplan-Meier curves for recurrence-free survival (RFS) and overall survival (OS) in the subgroup of C2 and C3 patients with high risk score according to pathological Miettienen-Joensuu scoring system and Imatinib sensitive mutations depending on the deep Miettinen-Joensuu models. Kaplan-Meier curves for RFS (A) and OS (B) depending on the deep Miettinen-Joensuu model employing the C2 DL Score in this subgroup of the C2 subcohort. Kaplan-Meier curves for RFS (C) and OS (D) depending on the deep Miettinen-Joensuu model employing the C3 DL Score in this subgroup of the C3 subcohort. *: p < 0.05; ***: p < 0.001. Tests are log-rank tests.

Background:Gastrointestinal stromal tumor (GIST) is the most common gastrointestinal mesenchymal tumor, driven by tyrosine-protein kinase KIT and platelet-derived growth factor receptor A (PDGFRA) mutations. Specific variants, such as KIT exon 11 deletions, carry prognostic and therapeutic implications, whereas wild-type (WT) variants derive limited benefit from tyrosine kinase inhibitors (TKIs). Given the limited reproducibility of established clinicopathological risk models, deep learning (DL) applied to whole-slide images (WSIs) emerged as a promising tool for molecular classification and prognostic assessment. Patients and methods:We analyzed 8398 GIST cases from 21 centers in 7 countries, including 7238 with molecular data and 2638 with clinical follow-up. DL models were trained on WSIs to predict mutations, treatment sensitivity, and recurrence-free survival (RFS). Results:DL predicted mutational status in GIST from WSIs, with area under the curve (AUC) of 0.87 for KIT, 0.96 for PDGFRA. High performance was observed for subtypes, including KIT exon 11 del-inss 557-558 (0.67) and PDGFRA exon 18 D842V (0.93). For therapeutic categories, performance reached 0.84 for avapritinib sensitivity, 0.81 for imatinib sensitivity. DL models predicted RFS, with hazard-ratios (HR) of 8.44 (95%CI 6.14-11.61) in the overall cohort and 4.74 (95%CI 3.34-6.74) in patients receiving adjuvant therapy. Prognostic performance was comparable to pathology-based scores, with highest discrimination in the overall cohort and in patients without adjuvant therapy (9.44, 95%CI (5.87-15.20)). Conclusion:DL applied to WSIs enables prediction of molecular alterations, treatment sensitivity, and RFS in GIST, performing comparably to established risk scores across international cohorts, providing a baseline for future multimodal predictors.

OBJECTIVES:To report the outcomes in adult patients with advanced pleomorphic rhabdomyosarcoma (P-RMS) treated with systemic therapy. METHODS:This global, multicenter, retrospective study conducted within the Pushing Ultra-Rare Sarcomas Towards Hope consortium (PUSH) included patients > 40 years with histologically confirmed advanced P-RMS, treated with at least one line of systemic therapy between 2013 and 2023. The primary endpoint was progression-free survival from first diagnosis of advanced disease, and from systemic treatment start (PFS-1 and PFS-2). Secondary endpoints included overall response rate (ORR), overall survival from first diagnosis of advanced disease and from treatment start (OS-1 and OS-2), and treatment-specific outcomes. RESULTS:Seventy-seven patients were included from 21 sarcoma reference centers. At a median follow-up of 44 months (IQR: 17.0-74.8), 49 (64%) patients had died and 48 (62%) had progressed. The median OS-1 and PFS-1 were 13.6 (95% confidence interval (CI): 9.4-22.5) and 5.4 (95% CI: 4.2-7.3) months, respectively. Two- and three-year OS-1 were 32.5% and 30.3%. Anthracycline-based regimens (n = 42) achieved a 50% ORR, with mPFS-2 and mOS-2 of 5.2 and 19.2 months; gemcitabine-based regimens (n = 15) a 42% ORR, with mPFS-2 and mOS-2 of 3.7 and 7.8 months; pazopanib (n = 6) a 33% ORR, with mPFS-2 and mOS-2 of 2.4 and 4.2 months; PD-1 inhibitors (n = 2) induced one response lasting 53 months. CONCLUSIONS:This series of advanced P-RMS treated with systemic agents, the largest available to date, showed meaningful activity of anthracycline- and gemcitabine-based regimens, and anecdotal responses to pazopanib and PD-1 inhibitors. Further prospective validation is planned.

DL predicts treatment sensitivity and supports clinical decision-making in GIST. A, Standard clinical workflow from initial diagnosis to treatment selection in patients with GIST. B, Proposed integration of the DL model into the diagnostic–therapeutic pathway, enabling early triage of patients for further molecular analysis or tailored treatment selection. C, AUC with 95% CI for DL-based prediction of treatment sensitivity, reported for the internal validation cohort (pink), external validation cohort (green), and an external biopsy-only validation cohort (light green). D, Confusion matrices for avapritinib and imatinib sensitivity, shown at the optimal threshold by Youden’s index (green) and at the threshold yielding the highest F1 score (pink).

Predictive performance of the DL models for selected molecular mutations in GIST.

BackgroundImproving osteosarcoma treatment beyond conventional (neo)adjuvant chemotherapy and resection remains challenging. An urgent need for novel therapeutic options, particularly personalized and targeted approaches, has emerged due to high inter-patient molecular heterogeneity. A lack of representative in vitro and in vivo models impedes therapeutic development, therefore we aimed to create 3D in vitro long term culture models directly from patient material.MethodsTumour cells from seven osteosarcoma patients were propagated in monolayer or collagen hydrogels, while whole-exome sequencing of corresponding primary tumour tissue was performed to identify potential drug targets. Established cultures were subsequently used to assess efficacy of the identified personalized treatment options.ResultsThree out of seven hydrogel cultures harbored the same genetic alterations as the corresponding primary tumours, but only one culture (L6565) showed viable cells after cryopreservation in combination with long term expansion. Our findings demonstrate feasibility of establishing long-term patient-derived osteosarcoma cultures with a success rate of 14%. This single patient line was used to evaluate genome-informed therapy and to compare cell culture models of increasing complexity. L6565 exhibited homozygous CDKN2A loss with retained Rb expression, rendering tumour cells sensitive to CDK4/CDK6 inhibition via palbociclib. Tumour heterogeneity was reflected in advanced culture methods producing more variability in treatment response.ConclusionThese results highlight the potential of genome-informed therapies in osteosarcoma and the importance of refining culture techniques to enhance translational research and therapeutic outcomes.

Sensitivity and specificity for treatment-sensitivity prediction.

This nonrandomized clinical trial aims to determine the efficacy and toxicity profile of a reduced preoperative radiotherapy dose in patients with myxoid liposarcoma. QuestionIs a dose reduction of preoperative radiotherapy in patients with myxoid liposarcoma oncologically safe with long-term follow up?FindingsIn this phase 2 nonrandomized clinical trial of 90 patients with myxoid liposarcoma of the trunk or extremity, a reduced-dose preoperative radiotherapy regimen of 36 Gy in 18 daily fractions demonstrated 97.4% local recurrence-free survival at 5 years. Among 87 patients who underwent surgery, wound complication and grade 2 or higher toxic effects occurred in 14 patients (16%) and 16 patients (18%), respectively.MeaningThese updated data are compelling, and a reduced dose can be considered in shared decision-making with the patient. ImportanceProspective data from 2 phase 2 trials showed favorable wound complication rates and promising local control after a reduced preoperative radiotherapy dose for myxoid liposarcoma (MLS). However, long-term follow-up data are currently lacking.ObjectiveTo determine the efficacy and toxicity profile of a reduced preoperative radiotherapy dose in patients with MLS with long-term follow-up.Design, Setting, and ParticipantsThe Dose Reduction of Preoperative Radiotherapy in Myxoid Liposarcoma (DOREMY) trial is a prospective, single-group, phase 2 nonrandomized clinical trial conducted in 9 tertiary sarcoma centers in Europe and the US. Eligible patients were adults with biopsy-proven and translocation-confirmed localized MLS of the trunk or extremity who were enrolled from November 24, 2010, to May 14, 2020. Data were analyzed from January to December 2025.InterventionPreoperative radiotherapy to a reduced dose of 36 Gy in once-daily 2-Gy fractions followed by resection.Main Outcomes and MeasuresLong-term local recurrence-free survival, progression-free survival, disease-specific survival, overall survival, and late toxic effects.ResultsNinety patients (mean [SD] age, 47 [13.1] years; 50 [56%] male) were included and followed up for a median (IQR) of 66.4 (48.8-87.5) months. Preoperative radiotherapy was delivered according to protocol in all patients. Surgery was not performed in 3 patients (3%) due to intercurrent metastatic disease. Local recurrence-free survival, progression-free survival, disease-specific survival, and overall survival rates at 5 years were 97.4% (95% CI, 93.9%-100%), 81.0% (95% CI, 72.6%-89.4%), 89.5% (95% CI, 82.6%-96.4%), and 88.5% (95% CI, 81.2%-95.8%), respectively. In total, 18 patients (21%) experienced a wound complication, and 14 (16%) required intervention. Any grade 2 or grade 3 late toxic effects were seen among 13 patients (15%) and 3 patients (3%), respectively.Conclusions and relevanceThis long-term analysis of the DOREMY nonrandomized clinical trial demonstrated excellent local control and a favorable toxicity profile following dose reduction of preoperative radiotherapy in patients with MLS. These compelling phase 2 findings support adoption of this regimen as an appropriate treatment option through shared decision-making with the patient, given the impracticality of conducting a phase 3 trial for a rare cancer.Trial RegistrationClinicalTrials.gov Identifier: NCT02106312

Abstract Gastrointestinal stromal tumor (GIST) is the most common gastrointestinal mesenchymal tumor, driven by tyrosine–protein kinase (KIT) and platelet-derived growth factor receptor A (PDGFRA) mutations. Specific variants, such as KIT exon 11 deletions, carry prognostic and therapeutic implications, whereas wild-type variants derive limited benefit from tyrosine kinase inhibitors. Given the limited reproducibility of established clinicopathologic risk models, deep learning (DL) applied to whole-slide images (WSI) emerged as a promising tool for molecular classification and prognostic assessment. We analyzed 8398 GIST cases from 21 centers in seven countries, including 7,238 with molecular data and 2,638 with clinical follow-up. DL models were trained on WSIs to predict mutations, treatment sensitivity, and recurrence-free survival (RFS). DL predicted mutational status in GIST from WSIs, with area under the curve of 0.87 for KIT and 0.96 for PDGFRA, and high performance was observed for subtypes, including KIT exon 11 del–inss 557 to 558 (0.67) and PDGFRA exon 18 D842V (0.93). For therapeutic categories, performance reached 0.84 for avapritinib sensitivity and 0.81 for imatinib sensitivity. DL models predicted RFS, with hazard ratios of 8.44 in the overall cohort and 4.74 in patients receiving adjuvant therapy. Prognostic performance was comparable with pathology-based scores, with highest discrimination in the overall cohort and in patients without adjuvant therapy. DL applied to WSIs enables prediction of molecular alterations, treatment sensitivity, and RFS in GIST, performing comparably with established risk scores across international cohorts, providing a baseline for future multimodal predictors. Significance: Deep learning on histology predicts KIT and PDGFRA mutations and stratifies recurrence-free survival in a large international cohort of gastrointestinal stromal tumors from multiple centers.

Chondrosarcomas are malignant cartilage-forming bone tumors with heterogeneous behavior, making prognostication and clinical management challenging. Histological grading is the primary tool for predicting clinical outcomes in conventional chondrosarcoma. However, its high interobserver variability limits reliable distinction between low- and high-risk patients, potentially leading to suboptimal clinical management. Given the increasing use of DNA methylation profiling as a valuable tool in surgical pathology for tumor classification, we investigated its prognostic value in central conventional chondrosarcoma. We generated methylation data from 69 primary central conventional chondrosarcomas profiled with Illumina's Human MethylationEPIC Array (850 k sites), and identified methylation sites individually linked to patient outcome. A LASSO Cox regression model was applied to these methylation sites to identify an optimal set of eight informative sites. With the regression coefficients of these methylation sites, we constructed a risk score that predicts central conventional Chondrosarcoma Risk Outcome from Methylation (CHROME). CHROME stratifies patients into High or Low risk groups for disease recurrence, onset of metastasis and disease-specific mortality. Survival analysis in an independent validation cohort (n = 68) demonstrated strong discriminatory performance of CHROME, with complete separation of outcomes and no adverse events observed in the Low risk group. Overview of the clinico-pathological information showed that a low grade (ACT/G1) case with an event was correctly assigned to the High risk group, while nine high grade cases without events were classified as Low risk. CHROME provides accurate risk stratification in central conventional chondrosarcoma and may represent a valuable tool for improving its prognostication and clinical management.

Sensitivity, specificity, and F1-score for molecular mutation prediction.

Importance:Chordoma is a rare malignant bone tumor with high local recurrence, metastatic spread in 40% to 60% of patients over the disease course, and significant morbidity. Because of its rarity, anatomical complexity, and prolonged natural history, high-quality evidence to guide management is limited. International consensus guidelines for localized chordoma were first published in 2015; however, advances in pathology, imaging, surgery, radiotherapy, and supportive care since then necessitate updated multidisciplinary recommendations. Objective:To update and expand the 2015 consensus recommendations on the diagnosis, treatment, and follow-up of pediatric and adult patients with primary, localized chordoma. Evidence Review:In June 2025, a meeting of the Global Chordoma Consensus Group was held in Milan, Italy, that included experts from all relevant specialties as well as patient representatives. A comprehensive literature review guided structured discussions on the management of primary localized disease. Levels of evidence and grades of recommendation were assigned. Findings:A total of 305 articles were included in the literature review. Management strategies were stratified by anatomical site (skull base, mobile spine, and sacrum). The central principal of care was treatment at experienced, multidisciplinary centers, with maximally safe surgery followed by high-dose, highly conformal radiotherapy. Guidance was provided on diagnosis, surgical approaches, and radiotherapy planning for each anatomical site. Systemic therapy options; long-term, risk-adapted follow-up; and supportive, palliative, and rehabilitative care were also addressed. Conclusions and Relevance:This global consensus statement provided updated multidisciplinary guidance for the management of primary, localized chordoma. It aimed to harmonize clinical practice, support shared decision-making, and identify priorities for future collaborative research in this rare and challenging disease.

Importance:Prospective data from 2 phase 2 trials showed favorable wound complication rates and promising local control after a reduced preoperative radiotherapy dose for myxoid liposarcoma (MLS). However, long-term follow-up data are currently lacking. Objective:To determine the efficacy and toxicity profile of a reduced preoperative radiotherapy dose in patients with MLS with long-term follow-up. Design, Setting, and Participants:The Dose Reduction of Preoperative Radiotherapy in Myxoid Liposarcoma (DOREMY) trial is a prospective, single-group, phase 2 nonrandomized clinical trial conducted in 9 tertiary sarcoma centers in Europe and the US. Eligible patients were adults with biopsy-proven and translocation-confirmed localized MLS of the trunk or extremity who were enrolled from November 24, 2010, to May 14, 2020. Data were analyzed from January to December 2025. Intervention:Preoperative radiotherapy to a reduced dose of 36 Gy in once-daily 2-Gy fractions followed by resection. Main Outcomes and Measures:Long-term local recurrence-free survival, progression-free survival, disease-specific survival, overall survival, and late toxic effects. Results:Ninety patients (mean [SD] age, 47 [13.1] years; 50 [56%] male) were included and followed up for a median (IQR) of 66.4 (48.8-87.5) months. Preoperative radiotherapy was delivered according to protocol in all patients. Surgery was not performed in 3 patients (3%) due to intercurrent metastatic disease. Local recurrence-free survival, progression-free survival, disease-specific survival, and overall survival rates at 5 years were 97.4% (95% CI, 93.9%-100%), 81.0% (95% CI, 72.6%-89.4%), 89.5% (95% CI, 82.6%-96.4%), and 88.5% (95% CI, 81.2%-95.8%), respectively. In total, 18 patients (21%) experienced a wound complication, and 14 (16%) required intervention. Any grade 2 or grade 3 late toxic effects were seen among 13 patients (15%) and 3 patients (3%), respectively. Conclusions and relevance:This long-term analysis of the DOREMY nonrandomized clinical trial demonstrated excellent local control and a favorable toxicity profile following dose reduction of preoperative radiotherapy in patients with MLS. These compelling phase 2 findings support adoption of this regimen as an appropriate treatment option through shared decision-making with the patient, given the impracticality of conducting a phase 3 trial for a rare cancer. Trial Registration:ClinicalTrials.gov Identifier: NCT02106312.

Predictive performance of the DL models for treatment-sensitivity categories in GIST.

Introduction Synthetic lethal interactions with IDH1 and IDH2 (IDH) mutations were identified in non-endogenous IDH mutant (IDHMUT) AML and glioma models, but are absent in endogenous IDHMUT chondrosarcoma cell lines. The translation into successful clinical applications has remained challenging, implying artificially created models do not fully recapitulate endogenous IDHMUT tumour biology. The aim of this study was to elucidate if the model system is indeed an important factor to consider when studying therapeutic vulnerabilities in IDHMUT tumours. Methods Vector-based and CRISPR-Cas9 approaches were used to introduce or revert the IDH1 mutation in chondrosarcoma cell lines. These isogenic cell line pairs were used to examine the presence of known therapeutic vulnerabilities and their underlying biological mechanisms. Results Vector-based IDHMUT chondrosarcoma models showed the previously reported synthetic lethal interactions, but these treatment sensitivities were absent in the CRISPR-edited models. Interestingly, not all vector-based IDHMUT cell lines displayed the same therapeutic vulnerabilities. Differences in treatment response were associated with multiple factors, including IDHMUT protein expression and D-2-HG levels, in line with the fact that therapeutic vulnerabilities could be induced in the CRISPR-edited models by enhancing D-2-HG levels. Conclusion Our findings demonstrate that synthetic lethal interactions observed in vector-based models are often a consequence of IDHMUT protein overexpression and supra-physiological D-2-HG levels. These results highlight that relying on artificially created IDHMUT models may lead to the identification of therapeutic vulnerabilities that are not present in IDHMUT tumours, potentially explaining the poor translation of preclinical findings to clinical trials.

Chordomas are ultra-rare malignancies of the axial skeleton with limited treatment options. Some tumors exhibit immunogenic features and respond to checkpoint blockade, but the biological determinants remain unclear. To investigate chordoma immunogenicity, we integrated spatial transcriptomic and metabolomic profiling. GeoMx digital spatial profiling was performed on 15 chordomas, analyzing tumor and stromal compartments and comparing inflamed versus non-inflamed tumors. Consecutive sections underwent MALDI mass spectrometry imaging. Key findings were validated by immunohistochemistry. To link transcriptional states with interferon-γ, chordoma cell lines were treated with interferon-γ. Spatial profiling revealed transcriptomic differences between inflamed and non-inflamed tumors across both compartments, and immune infiltration correlated with a broader gene expression repertoire in cancer cells. MSI identified distinct metabolic signatures, including glycogen accumulation in inflamed tumors. However, interferon-γ treatment did not induce TBXT (brachyury) expression in chordoma cells. Together, these data reveal strong associations between transcriptional states, metabolic profiles, and immune infiltration in chordomas.

Statistical pipeline for recurrence-free survival (RFS) analysis. (A) Data partitioning. Boxes with light dashed contours corresponded to the patients with available DL Scores while boxes with solid thick contours to the patient included in the survival analysis. (B) Methodological approach performed in C1, C2 and C3 to obtain deep learning (DL) scores and to develop survival models in the Training cohorts using the Cox proportional hazard (CPH) algorithm and to compare them. Other abbreviations: ext.val: external validation, int. val: internal validation, M.J.: Miettinen-Joensuu. *Covariables were: age, sex, adjuvant TKI therapy and mutational status for C1; and age, sex and mutational status for C2 and C3.