Gene variant driven sub-stratification of breast cancer patients.

目前的乳腺癌诊断依赖于雌激素/孕激素受体（HR）和HER 2受体值的二元（阳性/阴性）值。这些共同形成激素受体状态（HRS），与临床乳腺癌治疗决策密切相关。人们越来越认识到，乳腺癌是一种非常异质性的疾病，实际上代表了许多不同的亚型。由于基于二元分类系统的治疗决策，患者结局差异很大，难以预测，并且个性化化疗的手段有限。该领域越来越多地在寻找超越激素受体状态的方法，以更细粒度的方式进一步对患者进行分层，一种有希望的方法是使用基因变异。目前这个项目的目的是开发一个计算模型，将遗传变异与亚分层表型激素表达和由此产生的肿瘤微环境联系起来。与我们的工业合作伙伴Onkolyze一起，我们将开发一个系统，使科学家和临床医生能够将乳腺癌分类为“遗传亚型”，并有可能为这些亚型设计具有更高疗效的个性化治疗策略。学生将开发一个AI模型，可以将基因变异与激素受体表达百分比联系起来，目前在现有的临床病理学工作流程中进行了表征。该模型将预测所得的表型和这些表型的可能的药物-细胞相互作用。我们的模型效用将在来自MCRC生物库和合作者的患者来源的细胞系和患者来源的微环境中进行验证，用于成像模态分析。这将能够在细胞水平上基于患者组织中发现的遗传变异预测治疗结果，而不会显著偏离现有的临床工作流程。学生将接受生物信息学应用于个性化肿瘤学的广泛的跨学科技能培训;具体而言，他们将接受来自AI/机器学习，神经网络，生物信息学和统计学的最先进技术的培训。此外，通过验证步骤，他们将学习体外分析，技术和细胞生物学方法，包括细胞培养和免疫荧光显微镜。学生将在相关的背景理论（例如细胞周期，DNA复制，化疗方法，临床工作流程和癌症）获得较强的理论理解。学生还将通过我们的工业合作伙伴现有的正在进行的合作和UoM的合作，了解建模和合作的多学科方法。在博士学位结束时，学生将有一个适合高质量论文，多个高影响力的出版物，强大的跨学科技能和行业联系的工作机构，使他们在制药/生物技术或学术界的职位极具竞争力。

Current breast cancer diagnosis relies on binary (positive/negative) values for estrogen/progesterone hormone receptors (HR), and HER2 receptor values. Together, these form the hormone receptor status (HRS), which is closely linked to clinical breast cancer treatment decisions. It is increasingly recognised that breast cancer is an extremely heterogenous disease which in reality represents many varied subtypes. As a result of treatment decisions based around a binary classification system, patient outcomes are highly varied, difficult to predict, and there are limited means to personalise chemotherapy. Increasingly, the field is looking for methods to move beyond hormone receptor status to further stratify patients in a more granular fashion, one promising way to do this is using gene variants. The aim of this current project is to develop a computational model to link genetic variants to the sub-stratified phenotypic hormone expression and resulting tumour microenvironment. With our industrial partner Onkolyze, we will develop a system that will allow scientists and clinicians to classify breast cancer into "genetic subtypes" and potentially design personalised treatment strategies with higher efficacy for these subtypes. The student will develop a AI model that can link gene variants to hormone receptor percentage expression, which is currently characterised during existing clinical pathology workflow. This model will predict the resultant phenotypes and possible drug-cell interactions of these phenotypes. Our models utility will be validated in patient-derived cell lines and patient-derived microenvironments from the MCRC biobank and collaborators, for imaging modality analysis. This will enable the prediction of treatment outcomes based on genetic variants found in patient tissue at a cellular level without deviating significantly from existing clinical workflow. The student will receive a broad interdisciplinary skills training in bioinformatics applied to personalised oncology; specifically, they will be trained in state-of-the-art techniques from AI/machine learning, neural networks, bioinformatics, and statistics. In addition, via the validation steps they will learn in-vitro analytical, technical and cell biology methods including cell culture and immunofluorescence microscopy. The student will gain a strong theoretical understanding in the relevant background theory (e.g. cell cycle, DNA replication, chemotherapy approaches, clinical workflow, and cancer). The student will also develop an understanding of a multidisciplinary approach to modelling and collaboration via our industrial partner's existing on-going collaborations and those at UoM. At the end of the PhD the student will have a body of work suitable for a high-quality thesis, multiple high impact publications, a strong interdisciplinary skill set and industry links, making them extremely competitive for posts in pharma/biotech or academia.