OncoImmune Machine-Learning Model Predicts Immune Response and Prognosis in Leiomyosarcoma

OncoImmune Machine-Learning Model Predicts Immune Response and Prognosis in Leiomyosarcoma

September 19, 2025 Author Summary 0

Understanding the Challenge of Leiomyosarcoma

Leiomyosarcoma (LMS) is a malignant tumor that develops from smooth muscle and can occur in the uterus, gastrointestinal tract, or other soft tissues. It represents one of the most common adult soft-tissue sarcomas. Patients with advanced or unresectable LMS face limited treatment options. Standard first-line therapy often results in only about five months of progression-free survival and 14–16 months of overall survival.

Although immunotherapy has transformed outcomes for some cancers, only a small proportion of LMS patients benefit. This limited success highlights the need for new biomarkers that can capture the complexity of the tumor immune microenvironment (TIME). Within the TIME, monocytes—immune cells that can differentiate into dendritic cells or macrophages—play diverse roles in shaping tumor responses. Their behavior is influenced by signals such as nutrient levels, oxygen, and tumor-derived factors. Understanding how monocyte differentiation links to immune activity in LMS could open new directions for treatment.

A New Integrative Approach

The study combined single-cell RNA sequencing of monocyte-to-dendritic cell differentiation with bulk RNA sequencing data from LMS tumors and normal tissues. By overlapping these data, the researchers defined a set of 311 “OncoImmune” genes enriched for cytokine and immune-response pathways. These genes became the foundation for exploring how immune activity states connect with tumor behavior in LMS.

Stratifying Immune States

OncoImmune genes revealed distinct immune activation patterns in LMS, showing that monocyte differentiation states could reflect the immune environment of tumors. This connection suggests that such gene signatures might serve as biomarkers to guide clinical decision-making or identify new therapeutic targets.

A Seven-Gene Risk Model

From the OncoImmune hub genes, the team developed a machine-learning prognostic model. The model, based on just seven genes, stratified patients into high- and low-risk groups and showed moderate accuracy in predicting long-term outcomes. The authors describe this as a proof-of-concept tool with practical utility for personalized risk assessment in LMS.

Linking Genomics to Immunity

A key discovery was the relationship between ATRX mutation, CCDC69 expression, and mast-cell function. Tumors with ATRX mutations showed reduced levels of CCDC69, which in turn was associated with higher risk and impaired mast-cell activity in the tumor microenvironment. Mast cells, long recognized for their roles in immunity, appear here as important players in shaping LMS progression.

The ATRX–CCDC69–Mast-Cell Axis

Taken together, the findings support a working model in which ATRX mutation suppresses CCDC69, leading to impaired mast-cell function and worse patient outcomes. The study proposes this ATRX–CCDC69–mast-cell axis as a potential immunological and prognostic marker in LMS.

Implications for Research and Practice

  • Risk stratification: The seven-gene model offers a compact tool for separating LMS patients into prognostic groups, potentially aiding trial design or patient monitoring.

  • Biomarker development: ATRX mutation and CCDC69 expression may serve as markers of immune activity, guiding future efforts in precision oncology.

  • Hypothesis generation: The identified axis provides a new angle for experimental studies, particularly around mast-cell biology in sarcomas.

Limitations and Next Steps

The authors acknowledge important limitations. The sample size was relatively small, and experimental validation was not included. They call for larger cohorts, especially including patients treated with immunotherapy, to confirm the model’s predictive value. Mechanistic studies are also needed to directly test how ATRX influences CCDC69 and mast-cell function in LMS.

Take-Home Messages

  • Monocyte differentiation states can capture immune heterogeneity in LMS through the OncoImmune gene set.

  • A seven-gene prognostic model shows potential to personalize outcome prediction.

  • ATRX mutation connects genomics to immunity by lowering CCDC69 and impairing mast-cell function.

  • The proposed ATRX–CCDC69–mast-cell axis offers a new research direction in understanding LMS progression and immune responses.

“We believe our findings can pave the way for new directions in LMS research… [our] ML-based prognostic risk model… demonstrates practical utility in predicting long-term outcomes for individual patients.”

 

The translation of the preceding English text in Chinese:

 

理解平滑肌肉瘤的挑战

平滑肌肉瘤（leiomyosarcoma，LMS）是一种来源于平滑肌的恶性肿瘤，可发生于子宫、胃肠道或其他软组织。它是成人最常见的软组织肉瘤之一。对于晚期或不可切除的LMS患者，治疗选择有限。标准一线治疗的疗效往往有限：无进展生存期（PFS）仅约5个月，总生存期（OS）约为14–16个月。

尽管免疫治疗已改变部分癌症的结局，但仅有少数LMS患者从中获益。这一有限疗效凸显了开发能够反映肿瘤免疫微环境（tumor immune microenvironment，TIME）复杂性的新增生物标志物的必要性。在TIME中，单核细胞——可分化为树突状细胞或巨噬细胞的免疫细胞——通过多种方式塑造肿瘤反应。其行为受营养水平、氧浓度以及肿瘤来源因子等信号的影响。理解单核细胞分化如何与LMS中的免疫活性相联系，可能为治疗开辟新方向。

一种新的整合式方法

本研究将单核细胞向树突状细胞分化的单细胞RNA测序与来自LMS肿瘤及正常组织的bulk RNA测序数据相结合。通过对这些数据的交叉整合，研究者界定了一组311个“OncoImmune”基因，该基因集富集于细胞因子与免疫反应通路。以此为基础，团队探索了免疫活性状态如何与LMS的肿瘤行为相连接。

免疫状态分层

OncoImmune基因揭示了LMS中不同的免疫激活模式，显示单核细胞的分化状态能够反映肿瘤的免疫环境。这种联系提示，这类基因特征有望作为生物标志物，用于指导临床决策或识别新的治疗靶点。

七基因风险模型

基于OncoImmune的枢纽基因，研究团队构建了一个机器学习预后模型。该模型仅依赖7个基因即可将患者分为高风险与低风险组，并在长期结局预测中表现出中等准确性。作者将其描述为具备实际应用价值的“概念验证”型工具，可用于LMS的个体化风险评估。

将基因组学与免疫相连

研究的关键发现之一是ATRX突变、CCDC69表达与肥大细胞功能之间的关系。携带ATRX突变的肿瘤表现出较低的CCDC69水平，而这与更高的风险及肿瘤微环境中肥大细胞活性受损相关。长期以来以免疫作用著称的肥大细胞，在此被视为塑造LMS进展的重要参与者。

ATRX–CCDC69–肥大细胞轴

综合而言，研究支持这样一个工作模型：ATRX突变抑制CCDC69，从而导致肥大细胞功能受损并带来更差的患者结局。作者提出将该ATRX–CCDC69–肥大细胞轴作为LMS的潜在免疫学与预后标志物。

对研究与临床实践的启示

  • 风险分层： 七基因模型提供了一种紧凑工具，可将LMS患者划分为不同预后组，潜在用于优化试验设计或患者监测。

  • 标志物开发： ATRX突变与CCDC69表达可作为免疫活性的标志，指导未来精准肿瘤学工作。

  • 假设生成： 所识别的信号轴为实验研究提供新切入点，尤其是聚焦肉瘤中肥大细胞生物学的研究。

局限性与下一步工作

作者也承认存在重要局限性：样本量相对较小，且缺乏实验性验证。后续需要更大规模的队列，特别是纳入接受免疫治疗的患者，以验证模型的预测价值。同时，还需开展机制性研究，以直接检验ATRX如何影响CCDC69及肥大细胞功能在LMS中的作用。

要点总结

  • 通过OncoImmune基因集，单核细胞分化状态可以捕捉LMS中的免疫异质性。

  • 七基因预后模型显示出实现个体化结局预测的潜力。

  • ATRX突变通过降低CCDC69并损害肥大细胞功能，将基因组学与免疫联系起来。

  • 所提出的ATRX–CCDC69–肥大细胞轴为理解LMS进展与免疫反应提供了新的研究方向。

“我们相信，我们的研究结果可为LMS研究开辟新的方向……[我们的] 基于机器学习（ML）的预后风险模型……在预测个体患者的长期结局方面表现出实际应用价值。”

Reference:

Jingrong Deng, Changfa Shu, Dong Wang, Richard Nimbona, Xingping Zhao, Dabao Xu

OncoImmune machine-learning model predicts immune response and prognosis in leiomyosarcoma.

Biomol Biomed [Internet]. 2025 Jun. 4 [cited 2025 Sep. 19];25(10):2308–2323.

Available from: https://www.bjbms.org/ojs/index.php/bjbms/article/view/12342

Additional information:

We invite submissions for our upcoming thematic issues, including:

More news: Blog

Editor: Merima Hadžić

Be the first to comment

Leave a Reply