The impact of ecological competition and cooperation on cancer adaptive therapy

该项目旨在使用数学和计算模型，结合临床前和临床数据，加深对空间相互作用(“生态学”)在调节癌症治疗耐药性方面的重要性的理解。具体目标是：1)了解各种生态位如何调节癌症耐药性，2)建立整合癌症进化和生态学的多尺度模型，3)开发利用空间的多药物适应性疗法。在大多数已经扩散到转移性疾病的癌症中，治疗失败几乎是板上钉钉的。试图消除所有现在可能栖息在患者体内多个部位的癌细胞会导致癌细胞通过进化获得治疗耐药性。事实上，治疗阻力可能是治愈这种疾病或将其维持在不损害患者生活质量的水平上的最大障碍。最近，数学肿瘤学和癌症生物学紧密结合在一起，为癌症提供了利用进化而不是忽视它的新的临床治疗方法。这些策略成功的关键是预测适应，从而在治疗策略无效之前调整它们。最初的工作将集中在开发一种基于试剂的模型，该模型将耐药性视为一种可塑性状态，受微环境调节，并使用2-3种药物的组合进行治疗。了解每种药物出现耐药和恢复敏感的时间尺度，以及潜在的协同作用，对于决定如何有效地对药物进行排序以管理耐药性至关重要。临床前数据将可用于支持转移性黑色素瘤以及复发卵巢癌的这一项目。最终，我们的目标是开发一个强大的平台，在存在多种药物和多重耐药策略的情况下，向临床应用提供最佳形式的适应治疗。以前的工作主要集中在肿瘤对单一药物的反应上，而数学方法目前还不够详细，无法为临床医生提供实际指导。除此之外，还将开展深度强化学习在适应性治疗中的新应用方面的工作，特别是开发和表征独立学习适应性策略的模型。这位学生和莫菲特癌症中心的合作者进行的初步研究表明，在简单(非空间)肿瘤模型上训练的深度学习算法能够超越人类开发的策略，并可能通过转移学习应用于更复杂的(空间)模型。这项工作是与佛罗里达州莫菲特癌症中心的工业主管Alexander Anderson和Robert Gatenby合作进行的，他们将就研究的数学、建模和临床方面提供建议。该项目属于EPSRC数学生物学研究领域。它还有助于EPSRC改变医疗保健的战略优先事项，有可能提高耐药癌症患者的生活质量。

This project aims to use mathematical and computational modelling, combined with preclinical and clinical data, to develop a deeper understanding of the importance of spatial interactions ("ecology") in regulating therapeutic resistance in cancer. Specific goals are to: 1) understand how various ecological niches regulate cancer resistance, 2) build multiscale models that integrate cancer evolution and ecology, 3) develop multi-drug adaptive therapies that exploit space. Therapy failure is virtually assured in most cancers that have disseminated into metastatic disease. Trying to eliminate all cancer cells that may now inhabit multiple locations in the patient results in cancer cells acquiring therapeutic resistance via evolution. In fact, therapy resistance is probably the largest impediment to curing the disease, or maintaining it at a level that does not compromise the patient's quality of life. Recently, mathematical oncology and cancer biology have been tightly integrated to deliver novel clinical treatments for cancer that exploit evolution, rather than ignoring it. The key to success in such strategies is to anticipate adaptation and thus adjust therapeutic strategies before they become ineffective. Initial work will focus on the development of an agent-based model that considers resistance as a plastic state, modulated by the microenvironment, and treated with a combination of 2-3 drugs. Understanding the time scale of the emergence of resistance and the return to sensitivity for each drug, as well as potential synergies, will be critical in deciding how to sequence the drugs effectively to manage resistance. Preclinical data will be available to support this project in metastatic melanoma, as well as recurrent ovarian cancer. Ultimately, the goal is to develop a robust platform for informing clinical application on the best forms of adaptive therapy when there are multiple drugs, and multiple resistance strategies. Previous work has focused on the tumour response to a single drug, and mathematical approaches are currently insufficiently detailed to offer practical guidance to clinicians. In addition to this, work will also be conducted on a novel application of deep reinforcement learning to adaptive therapy, with a particular focus on developing and characterising models that independently learn adaptive strategies. Preliminary studies by the student and collaborators at the Moffitt Cancer Centre have suggested that deep learning algorithms trained on simple (non-spatial) tumour models are able to outperform human-developed strategies, and may be applied to more complex (spatial) models through transfer learning. This work is conducted in collaboration with industrial supervisors Alexander Anderson and Robert Gatenby based at the Moffitt Cancer Centre in Florida, who will advise on mathematical, modelling and clinical aspects of the research. This project falls within the EPSRC mathematical biology research area. It also contributes towards EPSRC's strategic priority to transform healthcare, with the potential to improve quality of life for patients with treatment resistant cancers.