Interrogating the mouse visual system by automated analysis of voluntary behaviour

视网膜变性和不可逆的光感受器疾病影响全球约2.2亿人，目前无法治愈。在这些疾病中，视网膜中的光敏神经元死亡，使视网膜无法对光做出反应，患者失明。由于最近的科学和技术进步，目前正在开发几种恢复这些人视力的方法。在将这些潜在疗法用于人类之前，衡量其有效性的唯一方法是询问它们是否可以恢复与人类相同致盲条件的动物的视力。目前，几乎所有的工作都是在实验室小鼠中进行的，因为患有视网膜变性的小鼠品系很容易获得。要知道治疗对这些动物是否有效，可以归结为回答一个简单的问题：老鼠的视力有多好？这个问题目前很难回答。因为你不能问老鼠他们看到了什么，研究人员必须使用间接的迹象表明老鼠在他们的环境中发现了视觉特征。在这个建议中，我将讨论如何实现这一目标的问题。我将从一个人道的方法开始，但目前还不能像我们希望的那样有效地回答这个问题。该测试依赖于这样的观察：当老鼠检测到视觉场景的变化时，它们通常会以行为的变化做出反应。目前，行为的变化被测量的是老鼠在他们的环境中移动的距离。我相信，如果我能更好地测量老鼠的行为，我可以使这个测试更加强大。我的项目的第一步是开发一种自动技术来识别不同类型的行为。为什么这一点很重要？想象一个男人站在一个房间里。过了一段时间，一束光从外面闪了出来，投射到窗户上。接下来会发生什么？男人可能会走到窗户边，也可能只是从他现在的位置转过头去看它。我们如何才能了解这个人是否已经看到了光明？唯一的办法就是观察他的动作。如果我们仅仅依靠他走向窗户，我们可能会失败，然而，如果我们也能够确定他何时将头转向窗户，我们将有更好的机会。为了实现这种可能性，我将跟踪身体点（如头部和上背部和下背部），我将使用一种算法来识别这些点何时发生突然运动。这种技术被称为变点分析，它将使我能够识别不同行为（如行走或头部运动）的开始或停止时间，并将它们与视觉环境中实验操作发生的时间联系起来。有时，视觉环境的单一操作不能引起自发反应，必须重复多次才能获得良好的视觉检测估计。然而，这可能是有问题的。为什么？为什么？回到房间里的那个人，想象一下闪光以固定的时间间隔发生多次。过了一段时间，男人会希望他们发生，失去兴趣。然而，如果闪光以不规则的间隔出现，彼此之间的距离足够远，那么人可能会继续对它们做出反应。我的项目的第二步是优化这些实验操作的时间，以最大限度地提高自发反应。我的项目的最后一步是使用这些技术来测量和比较三种最新和最有前途的基因治疗策略在视网膜变性后恢复视力的有效性。我的项目将提供一种更强大的方法来评估小鼠的视觉感知，并提供一个基准数据集来比较基因疗法的效果。它将能够更好地从动物研究结果推断出人类状况的改善，并减少用于测试的小鼠数量。

Retinal degeneration and irreversible photoreceptor diseases affect ~220 million worldwide and are currently not curable. In these diseases the light sensitive neurons in the retina die, leaving the retina unable to respond to light and the patient blind. Several approaches for restoring vision to these people, arising from recent scientific and technological advances, are currently under development. The only way to measure the effectiveness of these potential therapies before trailing them on humans is to ask whether they can restore vision to animals that suffer the same blinding conditions as humans. At present almost all of that work is undertaken in laboratory mice because strains of mice suffering retinal degeneration are readily available. Knowing whether treatments have been effective in these animals boils down to answering a simple question: how well can the mice see? This question is, at the moment, difficult to answer. As you cannot ask mice what they have seen, researchers must use indirect indications that mice have detected visual features in their environment. In this proposal I tackle the problem of how to achieve this. I will start with an approach that is humane but at present cannot answer the question as effectively as we'd like. The test relies on the observation that when mice detect a change in the visual scene they often respond with a change in behaviour. At present the change in behaviour that is measured is how far mice move around their environment. I believe that I can make this test much more powerful if I can measure mouse behaviour better. The first step of my project is to develop an automatic technique to identify distinct types of behaviour. Why this is important? Imagine a man standing in a room. After some time a light is flashed from the outside and projected through the window. What happens next? The man might walk to the window or instead he might just turn his head to look at it from his current position. How can we understand whether the man has seen the light? The only way to do that is by looking at his movements. If we only rely on him walking to the window we might fail in the task, however, if we are also able to determine when he turns his head towards the window we will stand better chances. In order to implement this possibility I will track body points (such as head and upper and low back) and I will use an algorithm to identify when sudden movements of these points occur. This technique is called changepoint analysis and will allow me to identify start or stop times of different behaviours (such as walking or head movements) and relate them to the times when experimental manipulations in the visual environment occur. Sometimes single manipulations of the visual environment fail to elicit a spontaneous response and they have to be repeated several times to obtain a good estimate of visual detection. However this can be problematic. Why? Back to the man in the room imagine that the flash occurs multiple times at regular time intervals. After a while the man will expect them to happen and lose interest. However if the flash is presented at irregular intervals sufficiently distant from each other the man might continue to react to them. The second step of my project is to optimise the timing of these experimental manipulations in order to maximise spontaneous reactions. The final step of my project is to use these techniques to measure and compare the effectiveness of three of the most recent and promising gene therapy strategies for restoring vision following retinal degeneration. My project will deliver a more powerful approach to assess visual perception in mice and a benchmark dataset to compare the effects of gene therapies. It will enable to better extrapolate from outcomes of animal studies to improvements in human condition and reduce the number of mice used for testing.