Quantitative translational neuroscience: Bridging preclinical and human neuroscience research

神经精神和神经发育障碍的治疗往往无法从成功的临床前试验过渡到真实的世界疗效。这在很大程度上是由于缺乏针对这些疾病的良好动物模型，这些疾病通常只针对人类的认知和神经过程。这些问题在小鼠模型的情况下尤其普遍。老鼠是一个受欢迎的模型物种，因为我们了解它们的许多遗传学并能够操纵它。然而，老鼠的大脑更小，并且与我们的大脑不同。老鼠和人类的大脑究竟有多大的不同，以及这如何影响从一个物种到另一个物种的结果翻译仍然是未知的。解决这个问题需要新的工具，使我们能够直接比较老鼠和人类大脑的组织。传统上，这是很难做到的，因为需要大量的数据，而且小鼠和人类大脑的大小和形状如此不同，很难找到一个参考框架。我们开创了一种解决这一问题的办法，我们称之为“共同空间办法”。实际上，我们使用来自两个物种的高通量全脑数据，然后根据由两个大脑中可以识别的共同特征组成的抽象特征空间来描述不同的大脑。例如，我们可以根据在任何特定区域表达的基因来描述大脑。如果我们选择老鼠和人类共有的基因，这将使我们能够在同一个“基因空间”中描述两个大脑。这是一种简单但非常强大的方法，可以直接比较不同物种的大脑组织。在这里，我们将使用这种方法来了解：（1）小鼠和人类大脑的每个部分有多相似。来自小鼠和人类的新的、公开可用的、高质量的基因表达和MRI数据使我们能够使用两个物种的相同类型的数据来比较大脑。我们的共同空间方法使我们能够确定小鼠大脑的每个部分与人类大脑的每个部分有多相似。(2)人类大脑的哪些方面是我们无法从老鼠身上理解的。如果我们能在同一个空间描述两个大脑，我们也可以评估人脑的哪些部分与老鼠的不同，在任何大脑组织水平上。这实际上向我们展示了老鼠-人类翻译的局限性。(3)是否有任何系统的规则来预测某种神经或精神疾病是否可以用老鼠模型来理解。通过大型国际财团，现在可以获得一系列疾病的大脑变化数据集。我们可以将它们与我们的翻译极限数据联合收割机结合起来，看看是否有某些预测因素表明某种疾病是否可以在小鼠中成功建模。然后，我们将通过比较某些条件下的小鼠模型与相同条件下的人类患者数据来测试这些预测因子。(4)开发一种方法，更好地将特定的疾病小鼠模型与特定的患者联系起来。许多精神疾病的小鼠模型不是一种，而是多种--每一种都对疾病的一个非常小的方面具有很高的结构效度。为了优化成功的翻译，将特定的小鼠品系与特定的患者相匹配将是有益的。我们的公共空间方法将使我们能够做到这一点，节省转化研究所需的工作量和动物数量。这项研究将主要基于牛津大学，但将受益于世界各地的合作者。大型财团已经开始绘制一系列疾病的大脑变化，并开始收集大量的人类和非人类成像数据。然而，迄今为止，这样的联盟从未弥合临床前动物研究和人类临床研究之间的差距。该项目将打破这些孤岛，为转化神经科学建立一个新的定量框架。

Therapies for neuropsychiatric and neurodevelopmental disorders often fail to make the transition from successful preclinical trials to real world efficacy. This is for a large part due to the lack of good animal models for these diseases, which often target uniquely human cognitive and neural processes. These problems are especially prevalent in the case of the mouse model. Mice are a popular model species, since we understand much of their genetics and are able to manipulate it. However, mouse brains are smaller and differently organized from ours. How different the brains of the mouse and human are exactly and how this influences translations of results from one species to the other remains largely unknown.Solving this problem requires new tools that allow us to directly compare the organization of the mouse and human brain. Traditionally, this is difficult to do due to the vast amounts of data required and the fact that the mouse and human brain are of such different size and shape that it's difficult to find a reference frame. We have pioneered an approach to solve this issue, which we term the 'common space approach'. In effect, we use high-throughput, whole-brain data from both species and then describe the different brains in terms of an abstract feature space consisting of common features that can be identified in both brains. For instance, we might describe the brains in terms of which genes are expressed in any given area. If we select genes that are shared by mice and humans, this will allow us to describe both brains in the same 'gene space'. This is a simple but extremely powerful way to directly compare brain organization across species.Here, we will use this approach to understand:(1) how similar each part of the mouse and human brain are. New, openly available, high quality gene expression and MRI data from both the mouse and the human allow us to compare the brains using the same type of data for the two species. Our common space approach allows us to determine how similar each part of the mouse brain is to each part of the human brain.(2) which aspects of the human brain we cannot understand based on the mouse. If we can describe the two brains in the same space, we can also assess which parts of the human brain are very distinct from the mouse, at any level of brain organization. This, in effect, shows us the limit of the mouse-human translation. (3) whether there are any systematic rules that predict whether a certain neurological or psychiatric disease can be understood using a mouse model. Though large international consortia, datasets of brain changes in a range of diseases are now available. We can combine them with our limits-of-translation data to see if there are certain predictors that indicate whether a disease can be successfully modelling in the mouse. We will then test these predictors by comparing mice models of certain conditions with human patient data from the same conditions.(4) develop a way to better relate specific mouse models of disease with specific patients. Many psychiatric disorders have not one mouse model, but many--each with a high construct validity for a very small aspect of the disease. To optimize successful translation, it would be beneficial to match specific mouse strains to specific patients. Our common space approach will allow us to do this, saving on the amount of work and the number of animals needed for translational research.This research will mostly be based at the University of Oxford, but will benefit from collaborators across the world. Large consortia have started mapping out brain changes in a range of diseases and have started to collect large amounts of human and non-human imaging data. However, to date such consortia never bridged the gap between preclinical animal research and human clinical research. This project will break those silos, building a new quantitative framework for translational neuroscience.