MRI-based mapping of regional genomic diversity in Glioblastoma

描述(由申请人提供)：我们建议开发一种基于图像的胶质母细胞瘤诊断系统(GBM)，该系统可以识别肿瘤几乎总是复发的区域内潜在的治疗耐药的遗传基础。这将有助于为GBM患者提供个性化护理。目前的治疗选择对所有患者都是程式化的和统一的，没有考虑到导致治疗抵抗和悲观预后的广泛的遗传多样性。具体地说，每个患者的GBM是独特的异质性，由多个不同的亚克隆群体组成，这些亚克隆群体对治疗的敏感性不同。这种多样性导致肿瘤对靶向治疗的反应不一致，并允许耐药克隆作为复发疾病重新繁殖。此外，常规MRI常规指导手术切除强化的肿瘤核心，但留下邻近非强化实质内的肿瘤群，或肿瘤周围的脑(BAT)。蝙蝠是辅助治疗的主要靶点，因为它藏匿着几乎普遍复发的残留肿瘤种群。管理蝙蝠体内的基因组多样性应该为治疗选择提供信息，但这个区域几乎从来没有活检过，因为它在传统的MRI上评估很差。目前，还没有系统的方法来解决肿瘤内的异质性来表征BAT内的区域基因组多样性。为了解决这一关键需求，这一探索性方案将开发和测试一种新的作图系统，该系统将多参数MRI与图像引导的组织分析和机器学习(ML)算法相结合，以描绘GBM的区域基因组变异。该系统使用常规MRI来识别主要的肿瘤亚成分：强化核心、BAT和中央坏死。在这些子成分中，先进的MRI(灌注、扩散、纹理)将进一步表征代表潜在基因组状态的表型表达的区域性肿瘤属性(即血管生成、渗透性、侵袭性和增殖)。这些MRI特征将指导来自不同肿瘤亚区的立体定向活检，以产生匹配的MRI和基因组数据对。ML算法将结合这些数据来估计每个肿瘤的区域基因组多样性，包括未经手术采样的BAT区域。我们统一了一支由多学科调查人员组成的团队，这些调查人员来自与我们有实质性和长期合作的机构。我们的团队提供实现研究目标所需的多个研究领域的专业知识，包括：1)图像处理和分析；2)图像引导的立体定向手术和联合配准；3)机器学习(ML)方法的发展；以及4)BAT内肿瘤的全面基因组询问和细胞生物学。如果成功，这里提出的工作将对GBM患者的诊断和治疗产生重大影响。这可能会改善临床结果，使范式从“一种疗法适用于所有人”转变为一种基于突变的方法，即选择针对个别肿瘤群体的组合疗法。

DESCRIPTION (provided by applicant): We propose to develop an image-based diagnostic system for Glioblastoma (GBM) that identifies the potential genetic underpinnings of treatment resistance within the zone where tumor almost always recurs. This should facilitate the delivery of individualized care for patients with GBM. Current therapy selection is formulaic and uniform for all patients and does not account for broad genetic diversity that contributes to treatment resistance and dismal prognosis. Specifically, each patient's GBM is uniquely heterogeneous and comprised of multiple distinct subclonal populations with differing susceptibilities to therapy This diversity causes tumors to respond non-uniformly to targeted therapy and allows resistant clones to repopulate as recurrent disease. Additionally, conventional MRI routinely guides surgical resection of enhancing tumor core, but leaves behind tumor populations within adjacent non-enhancing parenchyma, or brain around tumor (BAT). The BAT represents the primary target of adjuvant therapy because it harbors the residual tumor populations that nearly universally recur. Curating the genomic diversity within BAT should inform treatment selection, but this region is almost never biopsied because it is poorly evaluated on conventional MRI. Currently, there is no systematic method that addresses intratumoral heterogeneity to characterize the regional genomic diversity within BAT. To address this critical need, this exploratory proposal will develop and test a novel mapping system that integrates multi-parametric MRI with image-guided tissue analysis and machine learning (ML) algorithms to delineate regional genomic variations in GBM. This system uses conventional MRI to identify major tumoral subcomponents: enhancing core, BAT, and central necrosis. Within these subcomponents, advanced MRI (perfusion, diffusion, texture) will further characterize regional tumor properties (i.e., angiogenesis, permeability, invasion, and proliferation) that represent phenotypic expression of underlying genomic status. These MRI traits will guide stereotactic biopsies from distinct tumoral subregions to generate matched pairs of MRI and genomic data. An ML algorithm will incorporate these data to estimate regional genomic diversity throughout each tumor, including BAT areas that have not been surgically sampled. We have unified a multi-disciplinary team of investigators from institutions that have substantial and long-standing collaborations. Our group offers expertise in multiple fields of study that are necessary to accomplish the research aims, including: 1) image processing and analytics; 2) image-guided stereotactic surgery and coregistration; 3) development of machine learning (ML) methodology; and 4) comprehensive genomic interrogation and cell biology of tumor within BAT. If successful, the work proposed here should significantly impact how GBM patients are diagnosed and treated. This potentially improves clinical outcomes by enabling a paradigm shift from "one treatment fits all" to a mutations-based approach that selects combinatorial therapies targeting individual tumor populations.