Optimizing Parallel Factor (PARAFAC) Assisted Excitation-Emission Matrix Fluorescence (EEMF) Spectroscopic Analysis of Multifluorophoric Mixtures

Parallel factor (PARAFAC) analysis is the most commonly used mathematical technique to analyse the excitation-emission matrix fluorescence (EEMF) data sets of mutifluorophoric mixtures. PARAFAC essentially performs the mathematical chromatography on the EEMF data sets and helps in extracting pure excitation, pure emission and contribution profiles of each of the fluorophores without requiring any pre-separation step. The application of PARAFAC requires the initialisation of the spectral variables that is usually achieved by performing the singular value decomposition (SVD) analysis on EEMF data sets. One of the problem with SVD based initialisation is that it orthogonalises the data sets and makes the PARAFAC modelling of the EEMF data sets computationally challenging task that needs to be taken care. To address this issue, the present introduces an alternate approach for initialising the spectral variables for performing the PARAFAC analysis. The proposed approach essentially involve initialisation of the spectral variables with random numbers in a constraint manner. The proposed approach is found to provide the desired computational economy, robustness and analytical effectiveness to the PARAFAC analysis of EEMF data sets.