Ensemble framework to infer large-scale causal gene regulatory networks from transcriptomic data

现在已经确定，复杂的生物表型不是由单个基因决定的，而是由相互作用的基因和基因产物组成的网络决定的。因此，破译基因调控网络(GRN)的结构对于进一步了解人类细胞的基本过程至关重要。然而，细胞内领域中分子相互作用的映射仍然是从高通量生物学数据中产生生物学知识的管道中的瓶颈。 生物信息学和高性能计算的最新进展使人们有可能从转录数据集合中推断出非定向的大规模调控网络。然而，很少有网络推理方法可以推断预测的基因相互作用的方向性(因果关系)，尽管这是更好地解释GRN过程中的关键。推断大规模GRN的另一个挑战在于定量评估其有效性。流行的验证程序包括：(I)表明正在研究的算法在通过构建已知其真实基础网络的模拟数据集上执行得很好；(Ii)使用不完整的“金标准”数据集，例如已知的转录因子及其靶标，这些数据集很难概括从转录数据中可以推断出的相互作用；以及(Iii)使用低通量的实验室实验来验证几个预测的相互作用，这些相互作用只代表所推断的GRN中非常小且可能存在偏见的一部分。 我们的研究项目从几个角度解决了这些问题。首先，我们将在网络生物学领域最新贡献的基础上，开发新的机器学习方法，使之能够对大规模因果GRN进行推理。这些方法将是新颖的，它们实现了网络和因果推理的集成方法，并集成了从生物医学文献中提取的先验。其次，我们将扩展我们最近发布的验证框架，以定量评估我们的新方法，并将其与最先进的网络推理技术进行比较。第三，我们将开发一个计算平台，允许生物学家利用我们的计算工具，将他们感兴趣的基因或途径充分映射到大型GRN中。 为了实现这些目标，我将依靠我与著名的计算机科学家、生物信息学家和生物学家的密切合作，这不仅将确保我的方法应用于真实的和新的数据，而且还将持续评估这些方法的发展，以便影响参与生物和医学研究的更大的研究人员社区。

It is now established that complex biological phenotypes are not governed by single genes but instead by networks of interacting genes and gene products. As a consequence, deciphering the structure of the gene regulatory network (GRN) is crucial to further our understanding of fundamental processes in human cells. However, the mapping of molecular interactions in the intracellular realm remains the bottleneck in the pipeline to produce biological knowledge from high-throughput biological data. Recent advances in bioinformatics and high-performance computing made it possible to infer undirected large-scale regulatory networks from collections of transcriptomic data. However very few network inference methods can infer the directionality (causality) of predicted gene interactions, despite this being key in the process of better interpreting GRNs. Another challenge when inferring large-scale GRNs consists in quantitatively assessing their validity. Popular, however weak, validation procedures include (i) showing that the algorithm under study performs well on simulated datasets for which the true underlying network is known by construction; (ii) using incomplete ‘gold standard’ datasets, such as known transcription factors and their targets, which poorly recapitulate the interactions that can be inferred from transcriptomic data; and (iii) using low-throughput laboratory experiments to validate a few predicted interactions, which represent only a very small and potentially biased part of the inferred GRN. Our research program addresses these issues from several angles. First we will build on our recent contributions in the field of network biology to develop new machine learning methods enabling inference of large-scale causal GRNs. These methods will be novel in their implementation of an ensemble approach for network and causality inference, and their integration of priors extracted from the biomedical literature. Second we will extend our recently published validation framework to quantitatively assess and compare our new methods with state-of-the-art network inference techniques. Third we will develop a computational platform to allow biologists to leverage our computational tools to adequately map their genes or pathways of interest into large GRNs. To achieve these objectives, I will rely on my close collaborations with renowned computer scientists, bioinformaticians, and biologists, This will not only ensure that my methods are applied to real and new data, but also that the development of these methods are being continuously evaluated in order to impact the larger community of researchers involved in biological and medical studies.