Next Article in Journal
Oncolytic Therapies for Glioblastoma: Advances, Challenges, and Future Perspectives
Previous Article in Journal
Recent Advances in PET and Radioligand Therapy for Lung Cancer: FDG and FAP
Previous Article in Special Issue
Novel Molecular Methods in Soft Tissue Sarcomas: From Diagnostics to Theragnostics
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Cancers
Volume 17
Issue 15
10.3390/cancers17152551
cancers-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Integration of an OS-Based Machine Learning Score (AS Score) and Immunoscore as Ancillary Tools for Predicting Immunotherapy Response in Sarcomas

by
Isidro Machado
1,2,3,4,*,†,
Raquel López-Reig
5,6,†,
Eduardo Giner
7,
Antonio Fernández-Serra
5,6,
Celia Requena
8,
Beatriz Llombart
8,
Francisco Giner
3,7,
Julia Cruz
1,
Victor Traves
1,
Javier Lavernia
9,
Antonio Llombart-Bosch
3 and
José Antonio López Guerrero
6,7,10
1
Pathology Department, Instituto Valenciano de Oncología, Calle Gregorio Gea 31, 4to Piso, 46009 Valencia, Spain
2
Patologika Laboratory, Hospital QuirónSalud, 46010 Valencia, Spain
3
Pathology Department, University of Valencia, 46010 Valencia, Spain
4
Centro de Investigación Biomédica en Red de Cáncer (CIBERONC), 28029 Madrid, Spain
5
Molecular Biology Department, Instituto Valenciano de Oncología, 46009 Valencia, Spain
6
Joint Cancer Research Unit, IVO-CIPF, 46009 Valencia, Spain
7
Pathology Department, Hospital Universitario y Politécnico “La Fe”, 46009 Valencia, Spain
8
Dermatology Department, Instituto Valenciano de Oncología, 46009 Valencia, Spain
9
Oncology Department, Instituto Valenciano de Oncología, 46009 Valencia, Spain
10
Pathology Department, Catholic University of Valencia, 46001 Valencia, Spain
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.
Cancers 2025, 17(15), 2551; https://doi.org/10.3390/cancers17152551
Submission received: 11 June 2025 / Revised: 24 July 2025 / Accepted: 30 July 2025 / Published: 1 August 2025
(This article belongs to the Special Issue Genomics and Transcriptomics in Sarcoma)
Browse Figures
Review Reports Versions Notes

Simple Summary

Angiosarcomas (ASs) are a highly aggressive subset of sarcomas. This study aimed to develop and validate an immune-related prognostic model, termed the AS score, using two independent sarcoma cohorts. The model was initially established using a previously characterized cohort of 25 ASs. Validation was performed using transcriptomic data from TCGA Sarcoma cohort. The AS score effectively stratified AS cases into two prognostically distinct groups. Cases with a high AS score exhibited significantly poorer outcomes. Validation in TCGA sarcoma cohort confirmed the prognostic utility of both the AS score and the Immunoscore. Further survival analyses integrated these scores into four categories. Notably, among tumors with a high Immunoscore, the AS score significantly distinguished outcomes (p < 0.0001), identifying a subset with poor prognosis but potential sensitivity to immunotherapy. The combined use of the AS score and Immunoscore provides significant prognostic value in sarcoma, indicating that patients with high scores in both metrics may be suitable candidates for immunotherapy.

Abstract

Background: Angiosarcomas (ASs) represent a heterogeneous and highly aggressive subset of tumors that respond poorly to systemic treatments and are associated with short progression-free survival (PFS) and overall survival (OS). The aim of this study was to develop and validate an immune-related prognostic model—termed the AS score—using data from two independent sarcoma cohorts. Methods: A prognostic model was developed using a previously characterized cohort of 25 angiosarcoma samples. Candidate genes were identified via the Maxstat algorithm (Maxstat v0.7-25 for R), combined with log-rank testing. The AS score was then computed by weighing normalized gene expression levels according to Cox regression coefficients. For external validation, transcriptomic data from TCGA Sarcoma cohort (n = 253) were analyzed. The Immunoscore—which reflects the tumor immune microenvironment—was inferred using the ESTIMATE package (v1.0.13) in R. All statistical analyses were performed in RStudio (v 4.0.3). Results: Four genes—IGF1R, MAP2K1, SERPINE1, and TCF12—were ultimately selected to construct the prognostic model. The resulting AS score enabled the classification of angiosarcoma cases into two prognostically distinct groups (p = 0.00012). Cases with high AS score values, which included both cutaneous and non-cutaneous forms, exhibited significantly poorer outcomes, whereas cases with low AS scores were predominantly cutaneous. A significant association was observed between the AS score and the Immunoscore (p = 0.025), with higher Immunoscore values found in high-AS score tumors. Validation using TCGA sarcoma cohort confirmed the prognostic value of both the AS score (p = 0.0066) and the Immunoscore (p = 0.0029), with a strong correlation between their continuous values (p = 2.9 × 10−8). Further survival analysis, integrating categorized scores into four groups, demonstrated robust prognostic significance (p = 0.00021). Notably, in tumors with a low Immunoscore, AS score stratification was not prognostic. In contrast, among cases with a high Immunoscore, the AS score effectively distinguished outcomes (p < 0.0001), identifying a subgroup with poor prognosis but potential sensitivity to immunotherapy. Conclusions: This combined classification using the AS score and Immunoscore has prognostic relevance in sarcoma, suggesting that angiosarcomas with an immunologically active microenvironment (high Immunoscore) and poor prognosis (high AS score) may be prime candidates for immunotherapy and this approach warrants prospective validation.

1. Introduction

Sarcomas are a diverse and heterogeneous group of malignant tumors of mesenchymal origin, collectively accounting for approximately 1% of adult solid tumors [1,2]. With over 70 histological subtypes identified, they exhibit a wide spectrum of clinical behaviors and biological characteristics, many of which are highly aggressive and resistant to systemic therapy [1,2]. Among them, angiosarcomas (ASs) represent one of the most aggressive and heterogeneous subtypes, often associated with poor progression-free survival (PFS) and poor overall survival (OS) [3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. ASs represent between 1 and 2% of all soft tissue sarcomas and are typically found in older individuals (60–80 years old) [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. The most common location (50–60%) is the skin/subcutaneous tissue, with more than half occurring on the scalp and face [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. Mammary cases are much less frequent (5–10%), usually following radiotherapy for breast cancer [2,9,13]. Visceral cases (<5%), such as hepatic or splenic, are rare [2,3,4,5,7]. ASs associated with chronic lymphedema (Stewart–Treves syndrome) are also uncommon [2,3,5,8,9,12].
ASs are broadly categorized into primary and secondary forms [2,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20]. The secondary form often arises in the context of prior radiotherapy or chronic lymphedema and is characterized by distinct molecular profiles, such as MYC amplification [2,4,5,6,7,8,9,10,11,12,13,14,15,16]. Despite advances in cytotoxic therapies and surgical approaches [21,22,23,24], advanced/metastatic ASs remain largely refractory to most conventional treatments, underscoring the urgent need for alternative therapeutic strategies—particularly in metastatic/recurrent AS as well as in unresectable AS.
Immunotherapy, particularly immune checkpoint inhibitors (ICIs), has emerged as a promising therapeutic approach for several malignancies that have previously been resistant to treatment [25,26,27,28,29,30,31,32,33,34]. While ICIs, either alone or in combination with polychemotherapy, are currently approved for certain tumors in the neoadjuvant setting and as first-line systemic therapy for advanced or metastatic disease, their role in angiosarcomas and other sarcoma subtypes is not yet fully understood [25,26,27,28,29,30,31,32,33,34]. Recent studies suggest that a subset of ASs—especially those located in the head and neck region or lacking MYC amplification—may possess a more immunogenic tumor microenvironment (TME), potentially making them more responsive to immunotherapy [10,12,16,17,19,20,27,28]. In this context, the integration of immune-related biomarkers, such as the Immunoscore, has been proposed as a method to assess the immunologic activity of the TME [10,12,16,17,19,20].
Building upon our previous work characterizing the immune microenvironment in ASs through transcriptomic profiling and immunohistochemistry [12], we have developed and validated an immune-related prognostic model—the AS score—based on the expression of selected immune-associated genes. The validation step of the prognostic model was performed on the soft tissue sarcoma cohort from The Cancer Genome Atlas (TCGA) [35]. Additionally, we explored the integration of the AS score with the Immunoscore to refine risk stratification and evaluate their potential as ancillary tools for identifying patients who may benefit from immunotherapy in the context of sarcomas.

2. Material and Methods

2.1. Data Acquisition and Cohort Description

2.1.1. Training Cohort/Series

This study was designed as a validation analysis of an immune-related prognostic model previously developed using transcriptomic data from a series of 25 formalin-fixed, paraffin-embedded (FFPE) angiosarcoma (AS) tissue samples (Table 1) [12].
Formalin-fixed, paraffin-embedded (FFPE) tissue samples, encompassing both cutaneous and non-cutaneous angiosarcomas (including soft tissue and visceral forms; see Figure 1), were collected from patients diagnosed between 2000 and 2015 and retrieved from the pathology archives of multiple hospitals in our region. Patient age, sex, tumor size, medical history, anatomical location, treatment modalities (surgery, chemotherapy, radiotherapy), type of surgical procedure, and clinical outcomes (alive or deceased due to disease progression) were abstracted from medical records.
This study received approval from the Institutional Review Board and was conducted in accordance with the ethical guidelines of the participating institution and the principles outlined in the Declaration of Helsinki (1975, as amended in 2008). This cohort was utilized to develop the prognostic model, referred to as the “AS score”.

2.1.2. Validation Cohort/Series

To assess the prognostic applicability of the developed “AS score” in a broader sarcoma context, the model was applied to the soft tissue sarcoma cohort from The Cancer Genome Atlas (TCGA), which includes 253 adult cases (Supplementary Table S1) comprising dedifferentiated liposarcoma, leiomyosarcoma, undifferentiated pleomorphic sarcoma, myxofibrosarcoma, malignant peripheral nerve sheath tumor, and synovial sarcoma [35].
Given the absence of a dedicated angiosarcoma series within TCGA, the entire sarcoma cohort was used as a surrogate validation set.
Transcriptomic and clinical data were accessed through the firebrowse platform (http://firebrowse.org/) publicly available matrices.

2.2. Immuno-Related Gene Expression Data (HTG EdgeSeq Precision Immuno-Oncology Assay)

Gene expression was quantified using the HTG EdgeSeq Precision Immuno-Oncology Assay (HTG Molecular Diagnostics, Tucson, AZ, USA) [12], which profiles the immune contexture within tumor microenvironments. This targeted panel measures the expression of 1392 immune-related mRNAs across multiple tumor types. Expression values were normalized both within and between samples: we applied median-of-ratios normalization via DESeq2 (v1.42.1) [36] to reduce technical variability and emphasize biologically meaningful differences.

2.3. Development and Validation of Prognostic AS Score

The AS score was based on the expression data of candidates’ immune-related genes from 25 AS cases [12]. A multi-step analytical approach was employed to build the model. Initially, the prognostic significance of each of the 1392 genes under consideration was assessed using the Maxstat algorithm (R package Maxstat, v0.7-25), which determines the optimal cut-off based on log-rank test statistics for overall survival, thereby allowing the identification of genes with the strongest univariate associations with patient outcomes. Following the screening, the ten genes showing the highest statistical significance, as indicated by the lowest p-values, were selected for further analysis (Supplementary Table S2). These ten top-ranked candidates were subsequently entered into a multivariate Cox regression to evaluate their independent prognostic value and assign their weights (Figure 2). As a result of this analysis, four genes—IGF1R, MAP2K1, SERPINE1, and TCF12—were identified (Supplementary Table S3). Based on these four genes, the prognostic risk score, referred to as the AS score, was developed. The AS score was defined as the sum of the normalized expression values of the four genes, each weighted by its corresponding regression coefficient derived from the multivariate Cox analysis
AS score = (IGF1R × −1.22281) + (MAP2K1 × −0.63790) + (SERPINE1 × 1.00977) + (TCF12 × 0.61161)
A cut-off value of −1.9525 was established based on the value that best stratified the population according to OS. This value was used to dichotomize patients into high- and low-risk groups.
The developed model was applied to TCGA sarcoma cohort re-establishing the cut-off value to assess its prognostic value in an independent population. The new cut-off was 0.85. Gene expression data was retrieved as read counts and median normalized with DESeq2 [36] to compare this data with our series. Clinical metadata included histological subtype, tumor location, patient sex, and overall survival information. The AS score was calculated for all TCGA samples. To assess immune cell infiltration, the ESTIMATE R package (v1.0.13) [37] was used. Immunoscore values were normalized and dichotomized into high and low groups based on the most significative cut-off value regarding OS by the Maxstat algorithm.

2.4. Statistical Analysis

Nonparametric tests, including the Wilcoxon rank-sum and Kruskal–Wallis tests, were used to analyze continuous variables. Survival outcomes were assessed using Kaplan–Meier estimates, with group differences evaluated via the log-rank test. Multivariate analyses were performed with Cox proportional-hazards regression models. A two-sided p-value < 0.05 was considered statistically significant in all analyses. The primary time-to-event endpoint was overall survival (OS), defined as the time from diagnosis to death. All statistical analyses were conducted in RStudio version 4.3.3.

3. Results

3.1. Development of the AS Score in the AS Series

The total number of interrogated immune-related genes was categorized using the optimal cutoff value determined by the Maxstat algorithm. Genes were then selected using the log-rank test, applying a stringent p-value threshold of 0.002 to reduce the number of features used as input for the Cox regression model. The normalized expression levels of the 10 most significant genes were used to train the Cox regression model. An overall survival (OS) score was calculated using the coefficients of the significant parameters and their normalized expression levels, as follows
Score = IGF1R* × −1.22281 + MAP2K1* × −0.63790 + SERPINE1* × 1.00977 + TCF12* × 0.61161
Samples with high scores, including both cutaneous and non-cutaneous localizations, exhibited significantly poorer prognosis (p = 0.00012) compared to low-score samples, which were predominantly cutaneous (Figure 3).
To further validate the prognostic value of the AS score, we conducted both univariate and multivariate Cox regression analyses, incorporating common clinicopathological parameters alongside the AS score (Supplementary Table S4). In the univariate analysis, the primary tumor site (cutaneous vs. non-cutaneous; p = 0.0012), surgical margin status (p = 0.0049), and chemotherapy treatment (p = 0.0003) were significantly associated with overall survival (OS). However, in the multivariate model, only the AS score remained an independent predictor of poor prognosis (HR = 7.0; 95% CI: 2.4–21; p = 0.000444), while the other clinical variables lost statistical significance. These findings support the robustness of the AS score as a prognostic biomarker that provides independent predictive value beyond conventional clinicopathological features.

3.2. Validation of the AS Score in TCGA Sarcoma Cohort

The OS-based immune prognostic score (AS score) was applied to TCGA sarcoma dataset, which includes 253 soft tissue sarcoma cases. Patients were stratified into high- and low-risk groups based on a pre-established cut-off value of −1.9525. The AS score showed a significant correlation with overall survival, effectively distinguishing two prognostic groups with distinct clinical outcomes (p = 0.0006; Figure 4A). Patients in the high-risk group had shorter survival times compared to those in the low-risk group. These findings replicate the results previously observed in the angiosarcoma cohort, further validating the AS score as a robust and independent prognostic tool across a broader sarcoma population.

3.3. Prognostic Relevance of the Immunoscore

The Immunoscore, calculated using the ESTIMATE algorithm, was also evaluated in TCGA sarcoma cohort. Patients with a high Immunoscore demonstrated significantly improved overall survival (p = 0.0029), supporting the role of immune infiltration as a favorable prognostic factor. Notably, a positive correlation was observed between the Immunoscore and the AS score when analyzed as continuous variables (p = 2.9 × 10−8), suggesting that tumors with greater immune infiltration also tended to exhibit higher AS scores (Figure 4B). These findings indicate that immune-hot, yet high-risk tumors may constitute a distinct subgroup characterized by both aggressive behavior and potentially actionable immunogenicity.

3.4. Combined Stratification by AS Score and Immunoscore

To further investigate the prognostic landscape, patients were stratified based on their AS score and Immunoscore. Prognostic associations with overall survival were assessed independently for each score, as well as in combination. Patients were categorized into four groups based on their AS score and Immunoscore status: Group 1: High AS score/High Immunoscore; Group 2: High AS score/Low Immunoscore; Group 3: Low AS score/High Immunoscore and Group 4: Low AS score/Low Immunoscore.
The combination of both classifiers demonstrated significant prognostic stratification (p = 0.00021; Figure 5A). Notably, within the low Immunoscore cohort, the AS score did not significantly stratify patients. In contrast, among tumors with a high Immunoscore, the AS score showed strong prognostic power (p < 0.0001; Figure 5B). These findings highlight the potential utility of a combined immune-risk model to enhance patient selection in future immunotherapy-based clinical trials, with patients in Group 1 being eligible candidates for immunotherapy.
Furthermore, multivariate Cox regression analyses—including both the AS score and Immunoscore as covariates—confirmed that each provides independent prognostic information. This supports the hypothesis that the AS score and Immunoscore reflect distinct, yet potentially complementary, biological processes relevant to tumor progression and immune contexture (Figure 6).
To rigorously assess whether the AS score offers prognostic value independent of established clinicopathological variables, we conducted an extended multivariate Cox regression analysis in TCGA sarcoma cohort, adjusting for age, gender, tumor size, mitotic rate, radiotherapy, and surgical margin status (Supplementary Table S5). Significantly, both the AS score (HR = 2.47; 95% CI: 1.42–4.30; p = 0.0014) and the Immunoscore (HR = 0.42; 95% CI: 0.25–0.70; p = 0.0009) retained statistical significance in this comprehensive model, whereas none of the clinical variables, including tumor size, mitotic rate, or surgical margin status did. These findings confirm that the AS score is an independent prognostic marker, offering robust, biology-driven prognostic information that complements traditional clinical features in soft tissue sarcomas.

4. Discussion

In a recently published study, we reported a series of cutaneous and non-cutaneous angiosarcomas, which were stratified into tumors with a ‘hot’ immune microenvironment and those with a weaker immune response [12]. Hypothetically, these ‘hot’ tumors could benefit to some extent from immunotherapy, although other genetic, histological, and clinical factors may also influence the response to such treatment [24,25,26,27,28,29,30,31,32,33,34]. While some angiosarcomas may be susceptible to immunotherapy, to the best of our knowledge, no studies have stratified these neoplasms using scores based on immunity-associated genes and their correlation with overall survival. Furthermore, the use of immune microenvironment-related scores linked to overall survival has not been extensively explored in other types of sarcomas.
An AS score, based on the expression of four immune-related genes, was defined in the clinical context of angiosarcoma (AS) in the present study. Score-based stratification identifies a subgroup of cutaneous ASs with poorer prognosis, which correlates with a higher Immunoscore. Our findings align with previous observations in angiosarcomas, where tumors exhibiting greater immune activity—i.e., a ‘hot’ microenvironment—particularly those lacking MYC amplification or located in the head and neck region, were associated with more favorable outcomes [10,12,16,17,19,20,27,28]. In this study, we extend these insights to a heterogeneous cohort of soft tissue sarcomas [35], highlighting that immune-related transcriptomic features may have broad prognostic significance across different histologic subtypes. The robustness of the AS score in this independent cohort further supports its predictive value beyond vascular sarcomas.
In addition, we independently validated the developed immune-related prognostic model (AS score), within a broader soft tissue sarcoma cohort from The Cancer Genome Atlas (TCGA) [36]. The model allows the successful stratification of patients into high- and low-risk groups, demonstrating statistically significant differences in overall survival. These findings reinforce the prognostic value of the AS score and support its potential applicability beyond angiosarcoma. However, future studies should validate each score separately across diverse histological subtypes and consider recalibrating classification thresholds to account for tumor-specific biological contexts.
In parallel, we evaluated the Immunoscore derived from gene expression signatures indicative of immune cell infiltration. A higher Immunoscore was significantly associated with improved survival, reinforcing the concept that the tumor immune microenvironment plays a critical role in modulating sarcoma progression [25,26,27,28,29,30,31,32,33,34]. Interestingly, we observed a significant positive correlation between the Immunoscore and the AS score, suggesting that some tumors exhibit high immunogenicity despite being classified as high-risk by the prognostic model. These “immune-hot and high-risk” tumors may represent a biologically distinct subset characterized by both aggressive behavior and potential sensitivity to immunomodulatory therapies.
To further refine prognostication, we combined the AS score and the Immunoscore, generating four risk groups based on both immune activity and a transcriptomic-based AS score. This integrated approach improved survival stratification and identified specific subgroups with markedly different outcomes. Notably, the prognostic impact of the AS score was preserved in tumors with high immune infiltration, while it was diminished in immune-cold tumors. These results suggest that combining immune-related transcriptomic features with estimates of immune infiltration may enhance risk prediction and guide patient selection for future immunotherapy-based strategies.
This study has some limitations. The retrospective nature of the analysis, combined with the lack of a dedicated angiosarcoma cohort in TCGA, limits the specificity of the validation. Furthermore, the AS score was applied without cohort-specific recalibration, although supporting robustness may impact performance in histologically diverse sarcomas. In addition, we used a conventional statistical method, even though there are various advanced deep learning techniques [38,39,40] that could provide more detailed insights into the underlying immune mechanisms of certain biological processes and complement our research. Given the potential value of applying deep learning models, it would be interesting to employ them in future studies to validate our findings using prospective, independent angiosarcoma cohorts and other sarcoma types. Further clinical correlations are needed to confirm the model’s utility in guiding therapeutic decisions.

5. Conclusions

In conclusion, our findings validate the AS score as a prognostic tool for soft tissue sarcomas and demonstrate that its combination with the Immunoscore refines risk stratification. The combination of the AS score, based on the expression of four immune-related genes, together with Immunoscore-based classification presented a prognostic impact in sarcoma patients. Due to its hypothesized hot microenvironment (high Immunoscore) and worse prognosis (high AS score), these findings could constitute a window of opportunity for immunotherapy in this subgroup. These scores could constitute a valuable ancillary tool as selection criteria for clinical trials based on immunotherapy, particularly in the context of AS. However, the results should be validated in a wider series of angiosarcomas.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cancers17152551/s1. Table S1: Main clinicopathological characteristics of the TCGA sarcoma cohort (n = 253). Table S2: The 10 most significant genes evaluated by the maxstat algorithm. Table S3: Significant Gene-performance in cox regression analysis. Table S4: Univariate and multivariate Cox regression analysis of clinicopathological parameters and AS score in the angiosarcoma cohort. * The analysis was performed based on the available clinical data. Table S5: Univariate and multivariate Cox regression analysis of clinicopathological parameters, AS score, and Immunoscore in the TCGA sarcoma cohort. * The analysis was conducted using all available clinical information.

Author Contributions

Conceptualization, I.M., R.L.-R., A.F.-S., A.L.-B. and J.A.L.G.; methodology, I.M., R.L.-R. and A.F.-S.; software, R.L.-R. and A.F.-S.; validation, R.L.-R. and A.F.-S.; formal analysis, R.L.-R. and A.F.-S.; investigation, I.M., R.L.-R., A.F.-S., C.R., B.L., F.G., J.C., V.T., J.A.L.G. and A.L.-B.; resources, C.R., B.L., F.G., J.C., V.T., J.L., A.L.-B. and J.A.L.G.; data curation, R.L.-R. and A.F.-S.; writing—original draft preparation, I.M. and E.G.; writing—review and editing, I.M., R.L.-R., A.F.-S., C.R., B.L., F.G., J.C., V.T., J.L., A.L.-B. and J.A.L.G.; visualization, I.M., R.L.-R., A.F.-S., C.R., B.L., F.G., J.C., V.T., J.L., A.L.-B. and J.A.L.G.; supervision, A.L.-B. and J.A.L.G.; project administration, A.L.-B. and J.A.L.G. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

The patients’ consent and approval from the Institutional Research Ethics Committee of IVO (06-21) were obtained for research purposes. Ethic Committee Name: IVO; Approval Code: (06-21); Approval Date: 10/01/2024.

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study.

Data Availability Statement

The datasets generated and/or analyzed during this study are publicly available upon reasonable request.

Acknowledgments

We sincerely thank public databases such as TCGA for providing this platform and the investigators for sharing their important datasets.

Conflicts of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

  1. Fletcher, C.D.M.; Bridge, J.A.; Hogendoorn, P.C.W. (Eds.) World Health Organization Classification of Tumours of Soft Tissue and Bone, 4th ed.; IARC Press: Lyon, France, 2013. [Google Scholar]
  2. WHO Classification of Tumours Editorial Board (Ed.) World Health Organization Classification of Soft Tissue and Bone Tumours, 5th ed.; IARC Press: Lyon, France, 2020. [Google Scholar]
  3. Fury, M.G.; Antonescu, C.R.; Van Zee, K.J.; Brennan, M.F.; Maki, R.G. A 14-year retrospective review of angiosarcoma: Clinical characteristics, prognostic factors, and treatment outcomes with surgery and chemotherapy. Cancer J. 2005, 11, 241–247. [Google Scholar] [CrossRef]
  4. Fayette, J.; Martin, E.; Piperno-Neumann, S.; Le Cesne, A.; Robert, C.; Bonvalot, S.; Ranchère, D.; Pouillart, P.; Coindre, J.M.; Blay, J.Y. Angiosarcomas, a heterogeneous group of sarcomas with specific behavior depending on primary site: A retrospective study of 161 cases. Ann. Oncol. 2007, 18, 2030–2036. [Google Scholar] [CrossRef] [PubMed]
  5. Antonescu, C. Malignant vascular tumors—An update. Mod. Pathol. 2014, 27, S30–S38. [Google Scholar] [CrossRef]
  6. van IJzendoorn, D.G.P.; Bovée, J.V.M.G. Vascular Tumors of Bone: The Evolvement of a Classification Based on Molecular Developments. Surg. Pathol. Clin. 2017, 10, 621–635. [Google Scholar] [CrossRef] [PubMed]
  7. Ramakrishnan, N.; Mokhtari, R.; Charville, G.W.; Bui, N.; Ganjoo, K. Cutaneous Angiosarcoma of the Head and Neck-A Retrospective Analysis of 47 Patients. Cancers 2022, 14, 3841. [Google Scholar] [CrossRef] [PubMed]
  8. Requena, C.; Sendra, E.; Llombart, B.; Sanmartín, O.; Guillén, C.; Lavernia, J.; Traves, V.; Cruz, J. Cutaneous Angiosarcoma: Clinical and Pathology Study of 16 Cases. Actas Dermosifiliogr. 2017, 108, 457–465. [Google Scholar] [CrossRef]
  9. Shustef, E.; Kazlouskaya, V.; Prieto, V.G.; Ivan, D.; Aung, P.P. Cutaneous angiosarcoma: A current update. J. Clin. Pathol. 2017, 70, 917–925. [Google Scholar] [CrossRef]
  10. Abebe, K.; Wulff, I.; Thorsen, K.V.; Wagenblast, A.L.; Schmidt, G.; Jensen, D.H.; Holm, C.E.; Petersen, M.M.; Loya, A.C.; Mentzel, T.; et al. Clinicopathological characteristics, long-term prognosis and follow-up recommendations of primary and secondary cutaneous angiosarcoma: A Danish nationwide population-based cohort study. Eur. J. Surg. Oncol. 2025, 51, 109680. [Google Scholar] [CrossRef]
  11. Kuba, M.G.; Dermawan, J.K.; Xu, B.; Singer, S.; Plitas, G.; Tap, W.D.; D’Angelo, S.P.; Rosenbaum, E.; Brogi, E.; Antonescu, E.; et al. Histopathologic Grading Is of Prognostic Significance in Primary Angiosarcoma of Breast: Proposal of a Simplified 2-tier Grading System. Am. J. Surg. Pathol. 2023, 47, 307–317. [Google Scholar] [CrossRef]
  12. Machado, I.; Requena, C.; López-Reig, R.; Fernández-Serra, A.; Giner, F.; Cruz, J.; Traves, V.; Lavernia, J.; Claramunt, R.; Llombart, B.; et al. Tumor Microenvironment and Its Clinicopathologic and Prognostic Association in Cutaneous and Noncutaneous Angiosarcomas. Am. J. Clin. Pathol. 2023, 160, 18–34. [Google Scholar] [CrossRef]
  13. Guo, T.; Zhang, L.; Chang, N.E.; Singer, S.; Maki, R.G.; Antonescu, C.R. Consistent MYC and FLT4 gene amplification in radiation-induced angiosarcoma but not in other radiation-associated atypical vascular lesions. Genes Chromosomes Cancer 2011, 50, 25–33. [Google Scholar] [CrossRef]
  14. Behjati, S.; Tarpey, P.S.; Sheldon, H.; Martincorena, I.; Van Loo, P.; Gundem, G.; Wedge, D.C.; Ramakrishna, M.; Cooke, S.; Pillay, N.; et al. Recurrent PTPRB and PLCG1 mutations in angiosarcoma. Nat. Genet. 2014, 46, 376–379. [Google Scholar] [CrossRef]
  15. Huang, S.C.; Zhang, L.; Sung, Y.S.; Chen, C.L.; Kao, Y.C.; Agaram, N.P.; Singer, S.; Tap, W.D.; D’Angelo, S.; Antonescu, C.R. Recurrent CIC Gene Abnormalities in Angiosarcomas: A Molecular Study of 120 Cases With Concurrent Investigation of PLCG1, KDR, MYC, and FLT4 Gene Alterations. Am. J. Surg. Pathol. 2016, 40, 645–655. [Google Scholar] [CrossRef]
  16. Manner, J.; Radlwimmer, B.; Hohenberger, P.; Mössinger, K.; Küffer, S.; Sauer, C.; Belharazem, D.; Zettl, A.; Coindre, J.M.; Hallermann, C.; et al. MYC high level gene amplification is a distinctive feature of angiosarcomas after irradiation or chronic lymphedema. Am. J. Pathol. 2010, 176, 34–39. [Google Scholar] [CrossRef] [PubMed]
  17. Ko, T.K.; Guo, Z.; Kannan, B.; Lim, B.Y.; Lee, J.Y.; Li, Z.; Lee, E.C.Y.; Teh, B.T.; Chan, J.Y. Multiomics characterization of breast angiosarcoma from an Asian cohort reveals enrichment for angiogenesis signaling pathway and tumor-infiltrating macrophages. Front. Immunol. 2025, 15, 1515935. [Google Scholar] [CrossRef]
  18. Painter, C.A.; Jain, E.; Tomson, B.N.; Dunphy, M.; Stoddard, R.E.; Thomas, B.S.; Damon, A.L.; Shah, S.; Kim, D.; Gómez Tejeda Zañudo, J.; et al. The Angiosarcoma Project: Enabling genomic and clinical discoveries in a rare cancer through patient-partnered research. Nat. Med. 2020, 26, 181–187. [Google Scholar] [CrossRef]
  19. Espejo-Freire, A.P.; Elliott, A.; Rosenberg, A.; Costa, P.A.; Barreto-Coelho, P.; Jonczak, E.; D’Amato, G.; Subhawong, T.; Arshad, J.; Diaz-Perez, J.A.; et al. Genomic Landscape of Angiosarcoma: A Targeted and Immunotherapy Biomarker Analysis. Cancers 2021, 13, 4816. [Google Scholar] [CrossRef] [PubMed]
  20. Chan, J.Y.; Lim, J.Q.; Yeong, J.; Ravi, V.; Guan, P.; Boot, A.; Tay, T.K.Y.; Selvarajan, S.; Md Nasir, N.D.; Loh, J.H.; et al. Multiomic analysis and immunoprofiling reveal distinct subtypes of human angiosarcoma. J. Clin. Invest. 2020, 130, 5833–5846. [Google Scholar] [CrossRef] [PubMed]
  21. Young, R.J.; Natukunda, A.; Litière, S.; Woll, P.J.; Wardelmann, E.; van der Graaf, W.T. First-line anthracycline-based chemotherapy for angiosarcoma and other soft tissue sarcoma subtypes: Pooled analysis of eleven European Organisation for Research and Treatment of Cancer Soft Tissue and Bone Sarcoma Group trials. Eur. J. Cancer 2014, 50, 3178–3186. [Google Scholar] [CrossRef]
  22. Pasquier, E.; André, N.; Street, J.; Chougule, A.; Rekhi, B.; Ghosh, J.; Philip, D.S.J.; Meurer, M.; MacKenzie, K.L.; Kavallaris, M.; et al. Effective Management of Advanced Angiosarcoma by the Synergistic Combination of Propranolol and Vinblastine-based Metronomic Chemotherapy: A Bench to Bedside Study. EBioMedicine. 2016, 6, 87–95. [Google Scholar] [CrossRef]
  23. Bi, S.; Zhong, A.; Yin, X.; Li, J.; Cen, Y.; Chen, J. Management of Cutaneous Angiosarcoma: An Update Review. Curr. Treat. Options Oncol. 2022, 23, 137–154. [Google Scholar] [CrossRef]
  24. Cristescu, A.N.; Dumitrescu, I.; Tudose, I.; Beiu, C.; Dumitrescu, A.M.; Popa, L.G. Personalized Treatment for Scalp Angiosarcoma. J. Clin. Med. 2025, 14, 1278. [Google Scholar] [CrossRef] [PubMed]
  25. Assi, T.; Ibrahim, R.; Ibrahim, T.; Khoury, R.; Cesne, A.L. Ipilimumab and nivolumab: The ‘new kid on the block’ in advanced angiosarcoma. Immunotherapy 2023, 15, 1089–1091. [Google Scholar] [CrossRef] [PubMed]
  26. Chen, T.W. Angiosarcoma: Role of Immunotherapy. Curr. Treat. Options Oncol. 2025, 26, 242–250. [Google Scholar] [CrossRef]
  27. Guan, L.; Palmeri, M.; Groisberg, R. Cutaneous angiosarcoma: A review of current evidence for treatment with checkpoint inhibitors. Front. Med. 2023, 10, 1090168. [Google Scholar] [CrossRef]
  28. Rosenbaum, E.; Antonescu, C.R.; Smith, S.; Bradic, M.; Kashani, D.; Richards, A.L.; Donoghue, M.; Kelly, C.M.; Nacev, B.; Chan, J.E.; et al. Clinical, genomic, and transcriptomic correlates of response to immune checkpoint blockade-based therapy in a cohort of patients with angiosarcoma treated at a single center. J. Immunother. Cancer 2022, 10, e004149. [Google Scholar] [CrossRef] [PubMed]
  29. Deshpande, A.; Munoz, J.; Kelemen, K.; Dabak, V.; Hanbali, A.; Kurzrock, R. Images in Immunotherapy and Precision Oncology: Angiosarcoma of the Spleen and Liver. J. Immunother. Precis. Oncol. 2022, 6, 56–58. [Google Scholar] [CrossRef]
  30. Tawbi, H.A.; Burgess, M.; Bolejack, V.; Van Tine, B.A.; Schuetze, S.M.; Hu, J.; D’Angelo, S.; Attia, S.; Riedel, R.F.; Priebat, D.A.; et al. Pembrolizumab in advanced soft-tissue sarcoma and bone sarcoma (SARC028): A multicentre, two-cohort, single-arm, open-label, phase 2 trial. Lancet Oncol. 2017, 18, 1493–1501. [Google Scholar] [CrossRef]
  31. Botti, G.; Scognamiglio, G.; Marra, L.; Pizzolorusso, A.; Di Bonito, M.; De Cecio, R.; Cantile, M.; De Chiara, A. Programmed Death Ligand 1 (PD-L1) Expression in Primary Angiosarcoma. J. Cancer 2017, 8, 3166–3172. [Google Scholar] [CrossRef]
  32. D’Angelo, S.P.; Shoushtari, A.N.; Agaram, N.P.; Kuk, D.; Qin, L.X.; Carvajal, R.D.; Dickson, M.A.; Gounder, M.; Keohan, M.L.; Schwartz, G.K.; et al. Prevalence of tumor-infiltrating lymphocytes and PD-L1 expression in the soft tissue sarcoma microenvironment. Hum. Pathol. 2015, 46, 357–365. [Google Scholar] [CrossRef]
  33. Fujii, H.; Arakawa, A.; Utsumi, D.; Sumiyoshi, S.; Yamamoto, Y.; Kitoh, A.; Ono, M.; Matsumura, Y.; Kato, M.; Konishi, K.; et al. CD8+ tumor-infiltrating lymphocytes at primary sites as a possible prognostic factor of cutaneous angiosarcoma. Int. J. Cancer 2014, 134, 2393–2402. [Google Scholar] [CrossRef]
  34. Florou, V.; Rosenberg, A.E.; Wieder, E.; Komanduri, K.V.; Kolonias, D.; Uduman, M.; Castle, J.C.; Buell, J.S.; Trent, J.C.; Wilky, B.A. Angiosarcoma patients treated with immune checkpoint inhibitors: A case series of seven patients from a single institution. J. Immunother. Cancer 2019, 7, 213. [Google Scholar] [CrossRef] [PubMed]
  35. The Cancer Genome Atlas Research Network. Comprehensive and Integrated Genomic Characterization of Adult Soft Tissue Sarcomas. Cell 2017, 2, 950–965.e28. [Google Scholar] [CrossRef] [PubMed]
  36. Love, M.I.; Huber, W.; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014, 15, 550. [Google Scholar] [CrossRef]
  37. Yoshihara, K.; Kim, H.; Verhaak, R.G. Estimate: Estimate of stromal and immune cells in malignant tumor tissues from expression data. R. Package Version 2016, 1, r21. [Google Scholar]
  38. Salimi, M.; Houshi, S.; Gholamrezanezhad, A.; Vadipour, P.; Seifi, S. Radiomics-based machine learning in prediction of response to neoadjuvant chemotherapy in osteosarcoma: A systematic review and meta-analysis. Clin. Imaging. 2025, 123, 110494. [Google Scholar] [CrossRef]
  39. Sabeghi, P.; Kinkar, K.K.; Castaneda, G.D.R.; Eibschutz, L.S.; Fields, B.K.K.; Varghese, B.A.; Patel, D.B.; Gholamrezanezhad, A. Artificial intelligence and machine learning applications for the imaging of bone and soft tissue tumors. Front. Radiol. 2024, 5, 1332535. [Google Scholar] [CrossRef]
  40. Schmitz, F.; Sedaghat, S. Diagnostic Value of Magnetic Resonance Imaging Radiomics and Machine-learning in Grading Soft Tissue Sarcoma: A Mini-review on the Current State. Acad Radiol. 2025, 32, 311–315. [Google Scholar] [CrossRef] [PubMed]
Cancers 17 02551 g001
Figure 1. (A) Cutaneous angiosarcoma of the scalp; (B) Primary cutaneous angiosarcoma with prominent lymphoid infiltration (hematoxylin and eosin [H&E], 20×); (C) Primary breast angiosarcoma with hemorrhagic areas; (D) Primary breast angiosarcoma showing extensive vascular formation, endothelial atypia, and sparse lymphoid infiltration (H&E, 40×); (E) Lymphedema-associated cutaneous angiosarcoma; (F) Cutaneous angiosarcoma with spindle cell proliferation and minimal lymphoid infiltration (H&E, 40×); (G) Primary intestinal angiosarcoma with a fleshy appearance and hemorrhagic areas; (H) Primary intestinal angiosarcoma with pleomorphic morphology, mitotic figures, and marked nuclear atypia (H&E, 40×); (I) MYC amplification in secondary cutaneous angiosarcoma, 100×.
Figure 1. (A) Cutaneous angiosarcoma of the scalp; (B) Primary cutaneous angiosarcoma with prominent lymphoid infiltration (hematoxylin and eosin [H&E], 20×); (C) Primary breast angiosarcoma with hemorrhagic areas; (D) Primary breast angiosarcoma showing extensive vascular formation, endothelial atypia, and sparse lymphoid infiltration (H&E, 40×); (E) Lymphedema-associated cutaneous angiosarcoma; (F) Cutaneous angiosarcoma with spindle cell proliferation and minimal lymphoid infiltration (H&E, 40×); (G) Primary intestinal angiosarcoma with a fleshy appearance and hemorrhagic areas; (H) Primary intestinal angiosarcoma with pleomorphic morphology, mitotic figures, and marked nuclear atypia (H&E, 40×); (I) MYC amplification in secondary cutaneous angiosarcoma, 100×.
Cancers 17 02551 g001
Cancers 17 02551 g002
Figure 2. Immuno-based AS score performance. Model weights from selected parameters.
Figure 2. Immuno-based AS score performance. Model weights from selected parameters.
Cancers 17 02551 g002
Cancers 17 02551 g003
Figure 3. Immuno-based performance. The clinical impact of model-based stratification.
Figure 3. Immuno-based performance. The clinical impact of model-based stratification.
Cancers 17 02551 g003
Cancers 17 02551 g004
Figure 4. (A) Prognostic impact on the overall survival of the AS score; (B) Distribution of the Immunoscore between AS score-based groups.
Figure 4. (A) Prognostic impact on the overall survival of the AS score; (B) Distribution of the Immunoscore between AS score-based groups.
Cancers 17 02551 g004
Cancers 17 02551 g005
Figure 5. Impact of the AS score and Immunoscore combined classification by a long-rank test regarding: (A) four prognostic groups and (B) high Immunoscore groups.
Figure 5. Impact of the AS score and Immunoscore combined classification by a long-rank test regarding: (A) four prognostic groups and (B) high Immunoscore groups.
Cancers 17 02551 g005
Cancers 17 02551 g006
Figure 6. Multivariate Cox regression analysis demonstrating the independent prognostic value of the AS score and the Immunoscore. ** p-value < 0.01 and *** p-value < 0.001.
Figure 6. Multivariate Cox regression analysis demonstrating the independent prognostic value of the AS score and the Immunoscore. ** p-value < 0.01 and *** p-value < 0.001.
Cancers 17 02551 g006
Table 1. Main clinicopathological characteristics of the cohort (n = 25).
Table 1. Main clinicopathological characteristics of the cohort (n = 25).
Patient CharacteristicsTumor Characteristics
Age at diagnosis (years) Tumor size (cm)
  Median71  Median8
  Range31–83  Range2–50
Gender Mitotic rate
  Male10 (40)  Median20
  Female15 (60)  Range0.0–37.0
Primary Site Margins
  Cutaneous17 (68)  Negative21 (84)
  non-cutaneous soft tissue2 (8)  Positive4 (16)
  non-cutaneous visceral6 (24)Surgery
Follow-Up and OutcomesYes22 (88)
Overall survival No3 (12)
  N25Chemotherapy treatment
  Events20Yes9 (36)
  Median (range)14 (1–114) monthsNo16 (64)
Radiotherapy treatment
Yes2 (8)
No23 (92)
MYC (FISH)
Non-amplified10 (40)
Amplified7 (28)
NP8 (32)
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Machado, I.; López-Reig, R.; Giner, E.; Fernández-Serra, A.; Requena, C.; Llombart, B.; Giner, F.; Cruz, J.; Traves, V.; Lavernia, J.; et al. Integration of an OS-Based Machine Learning Score (AS Score) and Immunoscore as Ancillary Tools for Predicting Immunotherapy Response in Sarcomas. Cancers 2025, 17, 2551. https://doi.org/10.3390/cancers17152551

AMA Style

Machado I, López-Reig R, Giner E, Fernández-Serra A, Requena C, Llombart B, Giner F, Cruz J, Traves V, Lavernia J, et al. Integration of an OS-Based Machine Learning Score (AS Score) and Immunoscore as Ancillary Tools for Predicting Immunotherapy Response in Sarcomas. Cancers. 2025; 17(15):2551. https://doi.org/10.3390/cancers17152551

Chicago/Turabian Style

Machado, Isidro, Raquel López-Reig, Eduardo Giner, Antonio Fernández-Serra, Celia Requena, Beatriz Llombart, Francisco Giner, Julia Cruz, Victor Traves, Javier Lavernia, and et al. 2025. "Integration of an OS-Based Machine Learning Score (AS Score) and Immunoscore as Ancillary Tools for Predicting Immunotherapy Response in Sarcomas" Cancers 17, no. 15: 2551. https://doi.org/10.3390/cancers17152551

APA Style

Machado, I., López-Reig, R., Giner, E., Fernández-Serra, A., Requena, C., Llombart, B., Giner, F., Cruz, J., Traves, V., Lavernia, J., Llombart-Bosch, A., & López Guerrero, J. A. (2025). Integration of an OS-Based Machine Learning Score (AS Score) and Immunoscore as Ancillary Tools for Predicting Immunotherapy Response in Sarcomas. Cancers, 17(15), 2551. https://doi.org/10.3390/cancers17152551

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop