Dealing with reduced data acquisition times in Fluorescence Correlation Spectroscopy (FCS) for High-Throughput Screening (HTS) applications

Abstract

Fluorescence Correlation Spectroscopy (FCS) may be used to assay the binding of drug-like ligands to signaling proteins and other bio-particles. For High Throughput Screening (HTS), a competitive format is typically used in which the binding of an unlabeled compound results in displacement of a fluorescently labeled ligand. Unweighted curve-fitting of the normalized autocorrelation function (ACF) to a two-diffusion-component model can resolve the fractions of free and bound ligand if the diffusion rates differ sufficiently and if the experimentally estimated ACF has adequate precision. However, for HTS (and also for intracellular FCS studies) it is desirable to minimize the experimental data collection time. In this case, the precision of the ACF is limited and it becomes important to account for the statistical features of the ACF estimate when designing an assay. The errors at different points in the estimated ACF are correlated and hence least-squares fitting methods are not statistically rigorous. We compare different methods for estimating and curvefitting the ACF from the raw data of short duration FCS measurements. The methods are applied to data from experiments to assay binding of Alexa-488-labeled Bak peptide with Bcl-x L , which is an intracellular protein that acts to protect against programmed cell death. We present results from a detailed Monte Carlo simulation of the experiment, which is useful for validating short-duration assay capabilities. We also discuss the measurement of changes in steady state fluorescence anisotropy due to restricted rotational diffusion upon binding, which provides a complementary assay.

Data Reduction Methods for Application of Fluorescence Correlation Spectroscopy to Pharmaceutical Drug Discovery

“…In order to facilitate the study of realistic experimental conditions, such as triplet crossing, photobleaching, excitation saturation, detector dead time, and optical misalignment, to evaluate the effects of systematic and random errors in the ACF with short duration acquisition times, and to facilitate comparison of different data analysis strategies, a Monte Carlo simulation of single-molecule detection (SMD), FCS and fluorescence cross correlation spectroscopy (FCCS) has recently been developed [25]. The simulation is quite different from those that add random-number generated noise to a theoretical model ACF curve [31], but is an ab initio approach like those of Refs.…”

“…In this paper, we focus on data reduction and analysis aspects of FCS and the application of FCS to the determination of interactions between pairs of bio-molecules [25]. Section 2 begins with a brief overview on the use of fluorescence techniques for pharmaceutical compound screening and it introduces fluorescence methods for competitive ligand-binding measurements, including those complementary to FCS, such as methods based on fluorescence polarization and fluorescence brightness fluctuations.…”

Saturation effects in fluorescence correlation spectroscopy

“…For applications such as high throughput screening and intra-cellular studies with limited observation times, there is a need to make measurements with reduced data acquisition times [2]. In order to obtain adequate photon statistics and signal-to-noise for short measurement durations, it is desirable to increase the photon count rate by increasing the laser power.…”

Accounting for Triplet and Saturation Effects in FCS Measurements

“…Systematic errors can arise if other sources of fluctuations such as laser power fluctuations are present [18], if there is a corruption of the photon data due to effects such as detector dead time [34] or afterpulses [35,36], or if approximations that the correlator makes in determining the experimental ACF from the photon data stream become poor [37,38,39]. Systematic errors can also arise if the shape of the probe region from which fluorescence is collected (which is defined by the focused laser and the optical collection system) differs from that modeled by the fitting function [40].…”

“…Monte Carlo simulations [3,34,39,44,45,46,47,48,i] and numerical calculations [49,50,51,52,53,54,55,56] (which are also referred to as 'simulations' by some authors) have provided invaluable tools for quantitatively studying distortion and errors in the experimental ACF. Ref.…”