Computational prediction of gene function

在接下来的五年里，数以万计的新生物将完成它们的基因组测序。这些新数据带来了巨大的挑战和前所未有的机遇。 新的挑战是回答这个问题：我们如何理解这些基因组？虽然我们可以识别它们里面的基因，但我们怎么知道这些基因有什么作用呢？有关基因功能的许多信息很难单独从基因组序列中提取出来，需要进一步的实验才能确定。然而，这种类型的实验比基因组测序困难得多，成本也高得多。传统上，科学家们专注于少数具有代表性的“模型”生物来进行这些更详细的实验。然而，模式生物中的基因功能与其他密切相关的生物中的基因功能之间并不总是有明确的联系。这项提议的一个重点是试图定义在什么条件下我们可以自信地预测基因的功能在两个生物体之间是保守的。我们将通过训练计算算法来预测基因功能在两个模式生物之间何时保守来实现这一目标。在这种情况下，我们知道正确的答案，因此可以评估我们的方法做得有多好，然后再将它们应用于我们不知道正确答案的情况。 这个机会令人兴奋的部分是，我们可以使用这些新数据来帮助我们分析生物样本中包含的细胞。这种技术有着广泛的用途，从分析癌症肿瘤到检测血液中的感染性物质。我们将寻求这个机会，首先构建一个来自不同生物体的细胞类型特定特征的库，然后使用该库，以及一种新的计算技术(我们将开发)，我们将分析新的生物样本。 这两个项目代表了两个或更多学生的优秀论文项目；我们预测，如果成功，它们将对科学界乃至整个社会产生实质性和持久的影响。

Over the next five years, tens of thousands of new organisms will have their genomes sequenced. These new data present a substantial challenge and an unprecedented opportunity. The new challenge is answering the question: How do we understand these genomes? Although we can identify the genes within them, how can we tell what the genes do? Much about the function of a gene is difficult to extract from genome sequence alone and requires further experimentation to determine. However, this type of experimentation is much more difficult and costly than genome sequencing. Traditionally, scientists have focused on a small number of representative "model" organisms to perform these more detailed experiments. However, there is not always a clear link between gene function in model organisms and that in other, closely related ones. One focus of this proposal is to attempt to define the conditions under which we can confidently predict that a gene's function is conserved between two organisms. We will accomplish this goal by training computational algorithms to predict when gene function is conserved between two model organisms. In this situation, we know the correct answer and can therefore evaluate how well our methods are doing before applying them to situations in which we do not know the correct answer. The exciting part of this opportunity is that we can use these new data to help us to analyze biological samples in terms of what cells are contained within. Such a technique has a wide variety of uses, ranging from analyzing cancerous tumours to detecting infectious agents in blood. We will pursue this opportunity by first constructing a library of cell type specific signatures from different organisms and then using that library, along with a new computational technique (that we will develop), we will analyze new biological samples. These two projects represent excellent thesis projects for two or more students; and we predict that if successful, they will have a substantial and lasting impact on the scientific community and, potentially, society in general.