Vectorized data acquisition and fast triple-correlation integrals for Fluorescence Triple Correlation Spectroscopy

Abstract

Listen

Translate article

Fluorescence Correlation Spectroscopy (FCS) is widely used to quantify reaction rates and concentrations of molecules in vitro and in vivo. We recently reported Fluorescence Triple Correlation Spectroscopy (F3CS), which correlates three signals together instead of two. F3CS can analyze the stoichiometries of complex mixtures and detect irreversible processes by identifying time-reversal asymmetries. Here we report the computational developments that were required for the realization of F3CS and present the results as the Triple Correlation Toolbox suite of programs. Triple Correlation Toolbox is a complete data analysis pipeline capable of acquiring, correlating and fitting large data sets. Each segment of the pipeline handles error estimates for accurate error-weighted global fitting. Data acquisition was accelerated with a combination of off-the-shelf counter-timer chips and vectorized operations on 128-bit registers. This allows desktop computers with inexpensive data acquisition cards to acquire hours of multiple-channel data with sub-microsecond time resolution. Off-line correlation integrals were implemented as a two delay time multiple-tau scheme that scales efficiently with multiple processors and provides an unprecedented view of linked dynamics. Global fitting routines are provided to fit FCS and F3CS data to models containing up to ten species. Triple Correlation Toolbox is a complete package that enables F3CS to be performed on existing microscopes. Program summaryProgram title: Triple Correlation Toolbox (suite)Catalogue identifier: AEOP_v1_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEOP_v1_0.htmlProgram obtainable from: CPC Program Library, Queen’s University, Belfast, N. IrelandLicensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 50189No. of bytes in distributed program, including test data, etc.: 6135283Distribution format: tar.gzProgramming language: C/Assembly.Computer: Any with GCC and library support.Operating system: Linux and OS X (data acq. for Linux only due to library availability), not tested on Windows.RAM: ≥512 MB.Classification: 16.4.External routines: NIDAQmx (National Instruments), Gnu Scientific Library, GTK+, PLplot (optional)Nature of problem:Fluorescence Triple Correlation Spectroscopy required three things: data acquisition at faster speeds than were possible without expensive custom hardware, triple-correlation routines that could process 1/2 TB data sets rapidly, and fitting routines capable of handling several to a hundred fit parameters and 14,000 + data points, each with error estimates.Solution method:A novel data acquisition concept mixed signal processing with off-the-shelf hardware and data-parallel processing using 128-bit registers found in desktop CPUs. Correlation algorithms used fractal data structures and multithreading to reduce data analysis times. Global fitting was implemented with robust minimization routines and provides feedback that allows the user to critically inspect initial guesses and fits.Restrictions:Data acquisition only requires a National Instruments data acquisition card (it was tested on Linux using card PCIe-6251) and a simple home-built circuit.Unusual features:Hand-coded ×86-64 assembly for data acquisition loops (platform-independent C code also provided).Additional comments:A complete collection of tools to perform Fluorescence Triple Correlation Spectroscopy—from data acquisition to two-tau correlation of large data sets, to model fitting.Running time:1–5 h of data analysis per hour of data collected. Varies depending on data-acquisition length, time resolution, data density and number of cores used for correlation integrals.