Application of Methods from Information Theory in Protein Interaction Analysis

我们在这一优先方案内的头两个资助期所做的工作表明，信息论的概念可以用来处理蛋白质-蛋白质识别的生物学问题。为此，我们开发了一种基于互信息(MI)概念的形式主义，允许量化蛋白质界面不同结构特征的信息含量。然而，蛋白质不仅与其他蛋白质相互作用，而且还与核酸(DNA、RNA)结合。因此，我们希望在下一个资金阶段调整和测试我们基于MI的方法来研究蛋白质-DNA识别的特异性。我们将示范研究调控蛋白AmtR和CCPA作为模型系统，它们在细菌氮和碳控制中发挥关键作用，但其DNA结合特异性尚不清楚。特别是，我们想要回答以下问题：哪些序列特征在蛋白质-DNA识别中具有信息量？我们能定义增强特定蛋白质-DNA识别的互信息的结构特征吗？能量强的相互作用是否也表现出最高的结合互信息？特定结合的大部分信息是由蛋白质-DNA直接接触(直接读出)编码的，还是隐含在DNA的构象性质(间接读出)编码的？AmtR和CCPA的DNA结合特异性是否与相似的结构特征有关？我们能否使用MI作为设计具有高结合亲和力的最佳DNA序列的合适度量？AmtR和CCPA的识别原理是否也适用于其他蛋白质-DNA复合体？实现这些目标需要生成大量的DNA序列数据集，将基于MI的方法适应于DNA序列，识别适合于信息论分析的结构特征，以及设计新的DNA序列。

Our work of the first two funding periods within this priority programme has demonstrated that the concepts from information theory can be adapted to treat the biological problem of protein-protein recognition. For that purpose, we have developed a formalism based on the concept of mutual information (MI) that allows quantifying the information content of different structural features of protein interfaces. Proteins, however, do not only interact with other proteins but do also bind nucleic acids (DNA, RNA). Therefore, we want to adapt and test our MI-based approach to investigate the specificity of protein-DNA recognition in the next funding period. We will exemplarily study the regulatory proteins AmtR and CcpA as model systems, which play a crucial role in bacterial nitrogen and carbon control, but for which DNA binding specificity is yet poorly understood.In particular, we want to answer the following questions:Which sequence features are informative in protein-DNA recognition? Can we define structural features that enhance the mutual information for specific protein-DNA recognition? Do energetically strong interactions also exhibit the highest mutual information for binding?Is the majority of the information for specific binding coded by direct protein-DNA contacts (direct readout) or implicitly coded in the conformational properties of the DNA (indirect readout)? Are similar structural features responsible for the DNA binding specificity of AmtR and CcpA? Can we use the MI as a suitable measure for the design of an optimal DNA sequence that is recognized with high binding affinity? Are the principles of recognition deduced for AmtR and CcpA also transferrable to other protein-DNA complexes? Achieving these goals requires the generation of a large dataset of DNA sequences, the adaptation of our MI-based approach to DNA sequences, the identification of suitable structural features for information-theoretic analysis, and the design of novel DNA sequences.