A parameter estimation method for fluorescence lifetime data

BMC Research Notes

Ping Ma³  Taekjip Ha⁴  Hajin Kim¹  Daniel Sewell² 

[1] Center for Soft and Living Matter, Institute for Basic Science, Ulsan, Republic of Korea;Department of Statistics, University of Illinois Urbana-Champaign, 725 S. Wright St., Champaign 61820, IL, USA;Department of Statistics, University of Georgia, 101 Cedar Street, Athens 30602, GA, USA;Department of Physics, University of Illinois Urbana-Champaign, 1110 W. Green St., Urbana, 61801, IL, USA

关键词: Mixture distribution;    Overfitting data;    Numerical optimization;    Single-molecule fluorescence lifetime;   

Others  :  1231986
DOI  :  10.1186/s13104-015-1176-y

 received in 2013-11-26, accepted in 2015-05-19,  发布年份 2015

【 摘 要 】

Background

When modeling single-molecule fluorescence lifetime experimental data, the analysis often involves fitting a biexponential distribution to binned data. When dealing with small sample sizes, there is the potential for convergence failure in numerical optimization, for convergence to local optima resulting in physically unreasonable parameter estimates, and also for overfitting the data.

Results

To avoid the problems that arise in small sample sizes, we have developed a gamma conversion method to estimate the lifetime components. The key idea is to use a gamma distribution for initial numerical optimization and then convert the gamma parameters to biexponential ones via moment matching. A simulation study is undertaken with 30 unique configurations of parameter values. We also performed the same analysis on data obtained from a fluorescence lifetime experiment using the fluorophore Cy3. In both the simulation study and the real data analysis, fitting the biexponential directly led to a large number of data sets whose estimates were physically unreasonable, while using the gamma conversion yielded estimates consistently close to the true values.

Conclusions

Our analysis shows that using numerical optimization methods to fit the biexponential distribution directly can lead to failure to converge, convergence to physically unreasonable parameter estimates, and overfitting the data. The proposed gamma conversion method avoids these numerical difficulties, yielding better results.

【 授权许可】

   
2015 Sewell et al.

【 图 表 】

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

【 参考文献 】

• [1]Sorokina M, Koh H, Patel SS, Ha T: Fluorescent lifetime trajectories of a single uorophore reveal reaction intermediates during transcription initiation. J Am Chem Soc 2009, 131(28):9630-9631.
• [2]Kollner M, Wolfrum J (1992) How many photons are necessary for fluorescence-lifetime measurements? Chem Phys Lett 200(1,2):199–204
• [3]Xie XS: Single-molecule approach to dispersed kinetics and dynamic disorder: probing conformational fluctuation and enzymatic dynamics. J Chem Phys 2002, 117(24):11024-11032.
• [4]Duncan R, Bergmann A, Cousin M, Apps D, Shipston M: Multi-dimensional time-correlated single photon counting (TCSPC) fluorescence lifetime imaging microscopy (FLIM) to detect fret in cells. J Microsc 2004, 215(1):1-12.
• [5]Novikov E, Hofkens J, Cotlet M, Maus M, De Schryver FC, Boens N: A new analysis method of single molecule fluorescence using series of photon arrival times: theory and experiment. Spectrochim Acta Part A Mol Biomol Spectrosc 2001, 57(11):2109.
• [6]Sasaki K, Masuhara H: Analysis of transient emission curves by a convolved autoregressive model. Appl Opt 1991, 30(8):977-980.
• [7]Enderlein J, Erdmann R: Fast fitting of multi-exponential decay curves. Opt Commun 1997, 134(1):371-378.
• [8]Turton D, Reid G, Beddard G: Accurate analysis of fluorescence decays from single molecules in photon counting experiments. Anal Chem 2003, 75(16):4182-4187.
• [9]Laurence TA, Chromy BA: Efficient maximum likelihood estimator fitting of histograms. Nat Methods 2010, 7(5):338-339.
• [10]Maus M, Cotlet M, Hofkens J, Gensch T, De Schryver FC, Schaffer J, et al.: An experimental comparison of the maximum likelihood estimation and nonlinear least-squares fluorescence lifetime analysis of single molecules. Anal chem 2001, 73(9):2078-2086.
• [11]Edel JB, Eid JS, Meller A: Accurate single molecule fret efficiency determination for surface immobilized dna using maximum likelihood calculated lifetimes. J Phys Chem B 2007, 111(11):2986-2990.
• [12]Dempster AP, Laird NM, Rubin DB (1977) Maximum likelihood from incomplete data via the EM algorithm. J Royal Stat Soc Ser B 39:1–37. (with discussion)
• [13]Nelder JA, Mead R: A simplex method for function minimization. Comput J 1965, 7(4):308-313.
• [14]McKinnon KIM: Convergence of the Nelder–Mead simplex method to a nonstationary point. SIAM J Optim 1998, 9(1):148-158.
• [15]Enderlein J, Goodwin PM, Van Orden A, Patrick Ambrose W, Erdmann R, Keller RA: A maximum likelihood estimator to distinguish single molecules by their fluorescence decays. Chem Phys Lett 1997, 270(5):464-470.
• [16]Rothwell P, Berger S, Kensch O, Felekyan S, Antonik M, Wöhrl B, et al.: Multiparameter single-molecule fluorescence spectroscopy reveals heterogeneity of hiv-1 reverse transcriptase: primer/template complexes. Proc Natl Acad Sci 2003, 100(4):1655-1660.
• [17]Enderlein J, Sauer M: Optimal algorithm for single-molecule identification with time-correlated single-photon counting. J Phys Chem A 2001, 105(1):48-53.
• [18]Digman MA, Caiolfa VR, Zamai M, Gratton E: The phasor approach to fluorescence lifetime imaging analysis. Biophys J 2008, 94(2):14-16.
• [19]Kim G-H, Legresley SE, Snyder N, Aubry PD, Antonik M: Single-molecule analysis and lifetime estimates of heterogeneous low-count-rate time-correlated fluorescence data. Appl Spectrosc 2011, 65(9):981-990.
• [20]Le Cam L, Lo Yang G: Asymptotics in statistics: some basic concepts. Springer, New York; 2000.
• [21]Sanborn ME, Connolly BK, Gurunathan K, Levitus M: Fluorescence properties and photophysics of the sulfoindocyanine Cy3 linked covalently to DNA. J Phys Chem B 2007, 111(37):11064-11074.