|
Vibrational spectroscopy, with its sensitivity to biochemical changes and its potential for rapid noninvasive use, is a powerful tool for myriad clinical applications. A tremendous amount of research has been and continues to be reported, supporting existing applications and opening up exciting new avenues. As a result, the amount of data generated has exploded, demanding newer and faster analysis tools. It is no longer tenable to rely on programming experts for each and every problem since it restricts quick testing of ideas and exploring workflows. Knowledge of basic programming can help spectroscopy practitioners save enormous amounts of time spent on data analysis and channel that time toward experimentation. Learning general programming, however, can be time consuming and labor intensive. Therefore, this Spotlight aims to specifically teach only the commands necessary to analyze spectroscopic data (Raman/Fourier transform infrared (FTIR)) using MATLAB®. It explains how to build an analysis routine to apply a step-by-step combination of MATLAB commands and perform preprocessing and multivariate analysis directly from spectra-containing folders with a single click. As an example, an automated script that can import data from several folders, perform first derivatization, select a specific spectral range, perform area normalization and principal component analysis (PCA), plot PCA scores, save principal components, perform linear discriminant analysis (LDA) on PCA results, provide confusion matrix, cross-validate the LDA by the leave-one-out method, and perform predictions using the LDA model, all with a single click, is discussed in detail. A script for a support vector machine is also dealt with briefly. Using these scripts, the reader can build their own script dedicated to the routines used in their laboratory by making minor changes. As an example, modification of the code to automate mean and standard deviation calculations is included. The Spotlight is specifically meant for specialists from backgrounds other than mathematics and programming who wish to automate repetitive analysis and thus avoid technical jargon.
|
