Recognition and alignment of variables from UV-vis chromatograms and application to industrial enzyme digests classification

- Development of a Mathlab program to classify industrial enzymes from raw UV-vis chromatograms.
- Combination of alignment and identification to obtain higher amount of identified peaks.
- Use of realignment to improve the subsequent lineal discrimant analysis model.
- Successful classification of spiked real samples by the developed methodology.

In the last years, industrial applications of chemometrics have largely increased due to their capacity to extract important information from complex records as chromatograms or spectra data. The use of chemometric methods also can avoid the use of detectors of elevated cost. In this work, a procedure to recognize the relevant chemical information contained in complex UV-vis chromatograms, after a trypsin digestion, to identify the three enzyme main classes (proteases, amylases and cellulases) commonly employed in the cleaning industry, has been developed. In order to recognize the chromatogram peaks, six indices of peak identity or identifiers were defined. A program written in MATLAB was elaborated to accomplish multiple comparisons between chromatograms to construct 3rd order tensors, which contain the common peaks of two or more chromatograms. Using a training test and these tensors, the target and sample chromatograms were ordered according to the proximity to its respective class centroids. Further, the peaks with the best warranties of being correctly recognized as belonging to characteristic peptides, common to at least two chromatograms, were used to align the sample chromatograms. Afterwards, to construct an LDA model for enzyme classification, the relative peak area of the aligned and identified peaks were employed. The LDA model was validated showing a 100% of prediction capability by leave-one-out, by dividing the samples in training and evaluation sets and also by the successful prediction of some spiked real samples.