A Digital Platform for Processing Fluorescence Spectroscopy Data Using Simulation Modelling and Machine Learning Algorithms

Abstract

A digital computational platform is proposed for processing fluorescence spectroscopy data, which implements complex analysis of experimental information based on the simulation modelling and machine learning algorithms. Data analysis includes partitioning biophysical data into clusters according to the degree of likeness in some measure of similarity, finding the median cluster members (medoids), applying the data reduction method and visualizing the experimental data in a two-dimensional space. Analysis of the medoids is carried out by the analytical or simulation models of optical processes occurring in molecular systems. The visualization of data clusters in the original and transformed feature spaces is done with the aim of user interaction. As a demonstrative example, the platform FluorSimStudio is implemented for processing time-resolved fluorescence measurements (https://dsa-cm.shinyapps.io/FluorSimStudio). The digital platform is an open system and allows addition of complex analysis models, taking into account the development of new modelling and analysis algorithms.