From single-cell transcriptomic to single-cell fluxomic: characterising metabolic dysregulations for breast cancer subtype classification

尽管最近乳腺癌治疗取得了进展，但癌细胞的高度可变性及其相关的耐药性仍然是改善临床结果的巨大障碍。众所周知，癌细胞必须重新编程其细胞代谢（细胞内发生的维持生命的化学过程）以支持快速增殖并促进获得性耐药。然而，调节这种生物变化的潜在机制既没有得到充分的了解，也没有得到充分的治疗。直到最近，随着单细胞分析（一种允许分析单个癌细胞的新技术）的出现，已经有可能分析细胞水平的变化，这些变化有助于确定四种主要的乳腺癌亚型（即Luminal a， Luminal B， TNB和HER2阳性）并开发不同的治疗途径。然而，患者的生存率仍然很低，尤其是对于最具侵袭性的乳腺癌亚型，因为细胞的变化不能轻易地与促进耐药性的代谢状态的改变联系起来。此外，缺乏特定的工具来分析大量的单细胞代谢谱，使得在代谢水平上进行单细胞分析仍然不切实际。该提案旨在启动提赛德大学（英国）和康奈尔大学（美国）之间的国际合作，以表征来自四种主要乳腺癌亚型的32种不同乳腺癌细胞类型（即细胞系）的代谢特征，以识别代谢失调并做出明智的治疗决策。先进的计算技术（即人工智能）将被应用于识别每种癌症亚型中导致癌症进展的代谢反应和变化。拟议合作的最终目标是阐明乳腺癌亚型在代谢水平上的主要差异，为靶向药物的开发提供信息，并支持临床决策。首先，数学技术将被应用于开发32种不同乳腺癌细胞系的代谢模型（通过一组数学方程）。这32个模型将用数学方法描述不同癌细胞内部发生的代谢反应。这将通过整合PI （Occhipinti博士）在代谢建模方面的专业知识和国际合作伙伴（Betel博士）的单细胞分析知识来实现。其次，将应用先进的计算技术来确定影响32种癌细胞类型中每种细胞增殖的关键特征。这些特征将包括一组针对每种细胞类型的生物元素（即与癌症代谢相关的信息），可用于预测细胞特异性耐药性或为临床决策提供信息。最后，选定的关键特征（例如，对每种癌细胞类型的生长贡献最大的代谢反应）将通过计算和实验室实验进行验证，并通过项目期间安排的定期会议和讨论与乳腺癌临床医生和专家分享。具体而言，学术小组将协调广泛的活动，包括与英国国家医疗服务体系的乳腺癌专家定期举行会议，旨在通过知识和技能交流，对开发的计算模型提供反馈，同时促进医疗和计算领域不同部门之间的联系。拟议的项目汇集了来自医疗保健领域两个卓越中心的学者（即康奈尔大学的威尔康奈尔医学中心和蒂赛德大学的国家地平线中心），他们在细胞分析，代谢建模和人工智能方面有着良好的记录，以更好地了解癌症发展的代谢机制并改善癌症预后。

Despite the recent developments in breast cancer treatments, the high variability of cancer cells and their related drug resistance still pose a huge obstacle to improving clinical outcomes. It is now well-known that cancer cells must reprogram their cellular metabolism (chemical processes that occur within the cell to maintain life) to support rapid proliferation and promote acquired drug resistance. However, the underlying mechanisms regulating such biological changes are neither fully understood nor sufficiently treated. Only recently, with the advent of single cell analysis (a novel technique that allows the analysis of individual cancer cells), it has been possible to analyse changes at the cellular level that have helped in identifying four main breast cancer subtypes (i.e., Luminal A, Luminal B, TNB, and HER2 positive) and developing different treatment routes. However, patient survival remains low - especially for the most aggressive breast cancer subtypes - since cellular changes cannot be easily connected to an alteration in the metabolic state that promotes drug resistance. Moreover, the lack of specific tools to analyse a vast quantity of single cell metabolic profiles makes single-cell analysis at the metabolic level still impractical. This proposal aims at initiating an international collaboration between Teesside University (UK) and Cornell University (US) to characterise the metabolic profile of 32 different breast cancer cell types (i.e., cell lines) from the four main breast cancer subtypes to identify metabolic dysregulations and allow informed treatment decisions. Advanced computation techniques (i.e., artificial intelligence) will be applied to identify the metabolic reactions and changes responsible for cancer progression in each cancer subtype. The final objective of the proposed collaboration will be to elucidate the main differences among breast cancer subtypes at the metabolic level to inform the development of targeted drugs and support clinical decisions. First, mathematical techniques will be applied to develop metabolic models (through a set of mathematical equations) of 32 different breast cancer cell lines. These 32 models will mathematically describe the metabolic reactions taking place inside the different cancer cells. This will be achieved by integrating the expertise in metabolic modelling of the PI (Dr Occhipinti) with the knowledge of single cell analysis of the International Partner (Dr Betel).Second, advanced computational techniques will be applied to identify the key features affecting the proliferation of each of the 32 cancer cell types. Such features will include a set of biological elements (i.e., information related to cancer metabolism) specific to each cell type, which can be used to predict cell-specific drug resistance or inform clinical decisions. Finally, the selected key features (e.g. the metabolic reactions that are contributing the most to the growth of each cancer cell type) will be validated through computational and lab experiments and shared with breast cancer clinicians and experts through regular meetings and discussions that will be arranged during the project. Specifically, the academic team will coordinate a wide range of activities, including regular meetings with breast cancer experts from the NHS, designed to provide feedback on the developed computational model through knowledge and skills exchange while promoting connectivity across different sectors both in the medical and computational areas. The proposed project brings together academics from two centres of excellence in the healthcare sector (i.e., Weill Cornell Medicine at Cornell University and the National Horizon Centre at Teesside University), who have a strong track record in working with cell analysis, metabolic modelling, and artificial intelligence to better understand the metabolic mechanisms of cancer development and improve cancer outcomes.