Although the definition of the calibration factor, introduced in point (ii) above, is simple, its experimental determination is more challenging. While quantum efficiencies can be determined relatively easily, detection efficiencies are much more difficult to measure in practice. Therefore, the experimental determination of this calibration parameter is based on a different principle. Although several methods have been devised for this purpose, all of them are based on the concept, introduced by Trón et al.(9), of comparing the intensities of an equal number of excited acceptors and excited donors. The original approach was developed for measuring FRET between fluorescent antibodies. In such a system, a sample is labeled with only donor‐conjugated antibodies, and another sample is labeled with only acceptor‐conjugated antibodies against the same epitope. These two samples contain an equal number of donor‐tagged and acceptor‐tagged antibodies if a large enough number of cells are averaged. So that the ratio of the intensities of the acceptor‐only sample, measured in the FRET channel (M_A), and the donor‐only sample, measured in the donor channel (M_D), correspond to an equal number of donors and acceptors, the intensities are to be corrected with the degrees of labeling (DOL) of the donor‐tagged (L_D) and acceptor‐tagged antibodies (L_A). If the requirement of an equal number of excited donors and acceptors is to be met, the intensity ratio is to be further corrected with the molar absorption coefficients of the donor (ε_D) and the acceptor (ε_A) at the donor excitation wavelength:

((1))

Since the DOLs (L_D, L_A) are determined for the antibody stock solutions, the correctness of the formula hinges upon the assumption that the DOL of the stock is identical to the mean DOL of the cell‐bound antibody fraction. It has recently been explicitly shown that the average DOL of the cell‐bound fraction is often lower than the mean DOL of the stock due to diminished affinity of fluorescently labeled antibody species. Therefore, correction of the intensities with the DOL of the stock leads to a misestimation (12). It has also been pointed out that a slightly modified formula is required for the calculation of parameter α if fluorophore saturation takes place, that is, when most fluorophores are in the excited state (13). This situation is common when using confocal microscopies equipped with high numerical aperture objectives.

Since it is fairly easy to generate a 1:1 donor–acceptor ratio using fluorescent protein‐based FRET constructs, a plethora of approaches have been devised for such systems. In one of these methods, the loss in sensitized acceptor emission, due to partial acceptor photobleaching, is compared to consequent donor dequenching in order to obtain parameter G (14). Several approaches based on a series of donor–acceptor fluorescent protein constructs exhibiting different FRET values have been developed. Although the formalisms are somewhat different, the donor intensity and the sensitized acceptor intensities are compared in all of them. Initially, a regression approach utilizing an arbitrary number (at least two) of donor–acceptor constructs have been published (15) followed by another paper using two such constructs (16). Practically, the same principle is used in single‐molecule FRET measurements, in which the series of different donor–acceptor fluorescent protein constructs is replaced by different conformations of a single construct. The calibration factor, designated by G or α in other FRET approaches, is termed γ in single‐molecule FRET methods, and it is determined by regression (10). In two consecutive publications, an iterative and a closed‐form approach has been developed for determining G (or α, or γ) for a single donor–acceptor fluorescent protein construct (17, 18).

Recently, different measures of central tendency were evaluated for the determination of parameter G using a previously published method based on two donor–acceptor fluorescent protein constructs (16). Although the mean, the median, and the mode provided statistically insignificantly different estimates for G, the precision (reproducibility) of the mode was the best (19). Menaesse et al. (in this issue, page XXX) went one step further in simplifying the determination of G by using a single donor–acceptor fluorescent protein construct with known FRET efficiency. Using the calibrated FRET efficiency of the construct G was determined by regressing the sensitized emission on the quenched donor intensity. Besides describing the calibration method, the authors also compare this new approach to a previous one based on two FRET constructs with unknown energy transfer efficiencies (16, 19). The authors concluded that a single construct with known FRET efficiency provides a more precise estimation with fewer images compared to the previously used method. It was pointed out that the confidence interval of the estimation is broadened if a FRET construct with low FRET efficiency is used. An admitted drawback of the proposed approach is the requirement for a FRET construct with known energy transfer efficiency. Inaccuracy or uncertainty of the FRET efficiency of the calibration construct (e.g. because of the presence of unpaired donors due to incomplete maturation of the acceptor fluorescent protein) could obviously undermine the reliability of the proposed method. Since FRET calibration constructs with known energy transfer efficiency and rapidly maturing fluorescent proteins suitable for FRET are increasingly available, this approach is a viable and simple alternative for the determination of parameter G. Since the real value of FRET measurements compared to other, less quantitative approaches is the predictability and modelability of the readout parameter, the energy transfer efficiency, any improvement increasing the reproducibility and simplicity of the approach will definitely contribute to its successful implementation in biological research.

虽然上面 (ii) 点中介绍的校准因子的定义很简单，但它的实验确定更具挑战性。虽然可以相对容易地确定量子效率，但在实践中测量检测效率要困难得多。因此，该校准参数的实验确定基于不同的原理。尽管为此目的设计了几种方法，但所有方法都基于 Trón 等人（9)，比较相同数量的受激受体和受激供体的强度。最初的方法是为测量荧光抗体之间的 FRET 而开发的。在这样的系统中，一个样品仅用供体偶联的抗体标记，而另一个样品仅用针对相同表位的受体偶联抗体进行标记。如果对足够多的细胞进行平均，这两个样本包含相同数量的供体标记和受体标记抗体。因此，在 FRET 通道中测量的仅受体样品的强度 ( M _A ) 和在供体通道中测量的仅供体样品的强度之比( M _D)，对应于相同数量的供体和受体，强度将通过供体标记 ( L _D ) 和受体标记抗体 ( L _A )的标记度 (DOL) 进行校正。如果要满足激发供体和受体数量相等的要求，则强度比应进一步用供体激发波长处的供体（ε _D）和受体（ε _A）的摩尔吸收系数进行校正：

((1))

由于 DOL ( L _D , L _A ) 是针对抗体原液确定的，公式的正确性取决于库存的 DOL 与细胞结合抗体部分的平均 DOL 相同的假设。最近已经明确表明，由于荧光标记抗体种类的亲和力降低，细胞结合部分的平均 DOL 通常低于库存的平均 DOL。因此，用股票的 DOL 校正强度会导致错误估计 ( 12 )。也有人指出，如果发生荧光团饱和，即当大多数荧光团处于激发态时，计算参数 α 需要稍微修改公式（13 )。这种情况在使用配备高数值孔径物镜的共聚焦显微镜时很常见。

由于使用基于荧光蛋白的 FRET 构建体很容易产生 1:1 的供体 - 受体比例，因此已经为此类系统设计了大量方法。在这些方法之一中，将由于部分受体光漂白导致的敏化受体发射损失与随后的供体去猝灭进行比较，以获得参数G ( 14 )。已经开发了几种基于一系列表现出不同 FRET 值的供体 - 受体荧光蛋白构建体的方法。尽管形式主义有些不同，但都比较了供体强度和敏化受体强度。最初，已经发表了一种利用任意数量（至少两个）供体-受体结构的回归方法（15) 之后是另一篇使用两个这样的结构的论文 ( 16 )。实际上，在单分子 FRET 测量中使用了相同的原理，其中一系列不同的供体 - 受体荧光蛋白构建体被单个构建体的不同构象所取代。校准因子在其他 FRET 方法中由G或 α指定，在单分子 FRET 方法中称为 γ，它由回归确定 ( 10 )。在连续的两份出版物中，已经开发了一种迭代和封闭形式的方法来确定单个供体-受体荧光蛋白构建体的 G（或 α 或 γ） ( 17, 18 )。

最近，使用先前公布的基于两种供体-受体荧光蛋白构建体的方法，对确定参数G 的不同集中趋势测量进行了评估( 16 )。尽管平均值、中位数和众数为G提供了统计上不显着的估计值，但众数的精确度（再现性）是最好的 ( 19 )。梅内斯等人。（本期第 XXX 页）通过使用已知 FRET 效率的单一供体-受体荧光蛋白构建体，进一步简化了G的测定。使用构造G的校准 FRET 效率通过将敏化发射对淬灭的供体强度进行回归来确定。除了描述校准方法外，作者还将这种新方法与之前基于两种能量转移效率未知的 FRET 结构的方法进行了比较 ( 16, 19）。作者得出的结论是，与以前使用的方法相比，具有已知 FRET 效率的单个构造提供了更精确的估计，图像更少。有人指出，如果使用具有低 FRET 效率的 FRET 构造，则估计的置信区间会扩大。所提出的方法的一个公认的缺点是需要具有已知能量转移效率的 FRET 构造。校准构建体的 FRET 效率的不准确或不确定性（例如，由于受体荧光蛋白的不完全成熟导致未配对供体的存在）显然会破坏所提出方法的可靠性。由于具有已知能量转移效率和适用于 FRET 的快速成熟荧光蛋白的 FRET 校准结构越来越多，格。由于与其他定量较少的方法相比，FRET 测量的真正价值在于读出参数的可预测性和可建模性、能量转移效率，任何提高该方法可重复性和简单性的改进肯定会有助于其在生物研究中的成功实施。