Signal-Pair Correlation Analysis of Single-Molecule Trajectories
Signal-Pair Correlation Analysis of Single-Molecule Trajectories
复制标题
DOI：
10.1002/anie.201104033
复制
发表时间：
2011-01-01
影响因子：
16.6
通讯作者：
Woodside, Michael T.
中科院分区：
文献类型：
--
作者：
Hoffmann, Armin;Woodside, Michael T.
单分子（SM）方法极大地扩展了生物分子动态过程的研究。自从首次研究单离子通道[1]以来，诸如 Fçrster 共振能量转移 (FRET) 和力谱 [2] 等技术已经出现，可用于研究广泛的现象。通过直接观察单个分子随时间的结构变化，可以识别分子占据的状态，绘制出它们之间可能的转变，并测量转变率。这些信息已被用来构建各种大分子过程的独特详细图像，从离子通道功能 [1] 到分子运动运动、[3] 酶活性、[4] 基于 RNA 的调节、[5] 和生物分子折叠。 [6, 7] SM 方法的一个关键特征是能够观察和表征亚群，尤其是稀有或瞬态状态。然而，这些状态的动力学分析可能具有挑战性。通常，测量的信号轨迹被映射到一组离散状态，[8] 例如通过使用阈值或步长查找算法，[9] 或更复杂的最大似然方法，[10, 11] 特别是与隐马尔可夫模型 (HMM) 结合。[12, 13] 然后使用停留时间分布从状态轨迹获取动力学信息。 [14]这种方法的缺点是任何阻碍状态识别的因素（例如噪声或重叠信号、寿命差异大）都会引入难以控制的偏差。另一种策略是通过分析信号记录中的相关性来直接提取动力学，这具有许多优点：它对噪声和过滤伪影更稳健，相关拟合函数易于计算，并且拟合结果可用于直接测试动力学模型。 [15]信号强度相关性以前曾被用来分析光物理过程 [16, 17] 和二态折叠等现象，[18] 但这种分析对于具有相似时间尺度的多态系统或过程来说具有挑战性，因为系统中所有转变的速率都被折叠成单个相关函数，很难从中单独恢复。 [19] 在这里，我们提出了一种新型的相关分析，不是基于整个信号，而是基于与不同信号相关的离散信号范围。态，即使在多态系统中也可以确定转变速率和动力学方案。相关分析只需要与给定状态相关的部分信号，因此可以选择范围以最小化噪声数据中状态之间的重叠。最近，一种相关的方法被应用于通过 SM FRET 测量蛋白质扩散。[20]这种信号对相关方法甚至适用于由于噪声而重叠的状态、占用率低的状态以及非常相似或有数量级差异的速率的轨迹，所有这些问题都可能阻碍其他方法。该方法涉及两个步骤：首先，将信号（扩展、FRET 效率、电流……）分为离散范围，并计算所有范围对（“信号对”）之间的时间相关性。接下来，通过为每个状态分配一定的信号范围并将它们之间的所有互相关性与针对给定动力学方案导出的函数进行拟合来测试特定的动力学模型。通过对所有可能的方案重复拟合，可以根据经验验证正确的方案并确定相关的速率。通过使用信号对直方图可以简化信号范围的选择。这些包含有关当前状态以及它们之间的转换的有价值的信息，而无需识别......
Single-molecule (SM) methods have dramatically expanded the study of dynamic processes in biomolecules. Since the first studies of single ion channels,[1] techniques such as Fçrster Resonance Energy Transfer (FRET) and force spectroscopy [2] have emerged which can be used to study a broad range of phenomena. By directly observing structural changes in a single molecule over time, the states occupied by the molecule can be identified, the possible transitions between them mapped out, and the transition rates measured. Such information has been used to build uniquely detailed pictures of various macromolecular processes, from ion-channel function [1] to molecular-motor motion,[3] enzymatic activity,[4] RNA-based regulation,[5] and biomolecular folding.[6, 7] A key feature of SM approaches is the ability to observe and characterize sub-populations, especially rare or transient states. Kinetic analysis of these states can, however, be challenging. Often the measured signal trajectories are mapped onto a set of discrete states,[8] for example by using thresholding or step-finding algorithms,[9] or more-sophisticated maximum-likelihood methods,[10, 11] especially in combination with hidden Markov modeling (HMM).[12, 13] Dwelltime distributions are then used to obtain kinetic information from the state trajectories.[14] A drawback of this approach is that any factors hindering state identification (eg noisy or overlapping signals, large differences in lifetimes) can introduce biases for which it is difficult to control. An alternate strategy is to extract kinetics directly by analyzing correlations in the signal record, which has a number of advantages: it is more robust against noise and filtering artifacts, correlation fitting functions are easily calculated, and the fits can be used to test kinetic models directly.[15] Signal–intensity correlations have been used previously to analyze phenomena such as photo-physical processes [16, 17] and two-state folding,[18] but such analysis has proven challenging for multi-state systems or processes having similar timescales, because the rates for all the transitions in the system are folded into a single correlation function from which they are difficult to recover separately.[19]Herein we present a new type of correlation analysis, based not on the entire signal but rather on discrete ranges of the signal associated with different states, which allows transition rates and kinetic schemes to be determined even in multi-state systems. Only part of the signal associated with a given state is needed for correlation analysis, hence the ranges can be chosen to minimize overlap between states in noisy data. A related approach was recently applied to protein diffusion measured by SM FRET.[20] This signal-pair correlation method works even for trajectories with states that overlap because of noise, states with low occupancy, and rates that are very similar or differ by orders of magnitude, all issues that can hinder other approaches. The method involves two steps: First, the signal (extension, FRET efficiency, current…) is divided into discrete ranges and the time correlations between all pairs of ranges (“signal pairs”) are calculated. Next, specific kinetic models are tested by assigning a certain signal range to each state and fitting all cross-correlations between them with the functions derived for a given kinetic scheme. By repeating the fits for all possible schemes, the correct scheme can be validated empirically and the associated rates determined. The selection of signal ranges is simplified by using signal-pair histograms. These contain valuable information about the states present and the transitions between them, without the need to identify …