Discriminating complex biological materials using pyrolysis-mass spectrometry-chemometrics

Abstract

Chemometrics using supervised and unsupervised multivariate pattern recognition techniques is shown to be effective for differentiating complex pyrolysis-mass spectrometry (Py-MS) data from genetically different, but morphologically similar, bio-materials. Three studies using various multivariate pattern recognition approaches are presented to demonstrate the Py-MS-chemometric method. In one study Py-MS coupled with unsupervised pattern recognition (Py-MS-PR) was used to discriminate among fish egg samples representing Colorado squawfish ( Ptychocheilus lucius, Girard, 1856), Pahranagat roundtail chub ( Gila robusta jordani, Baird and Girard, 1853) and two populations of cut-throat trout ( Oncorhynchus clarki, Richardson, 1836). The Py-MS procedure provided characteristic chemotaxonomic information that allowed separation of the fishes at or below the currently recognized species level. In another study, a chemotaxonomic analysis of leafy spurge ( Euphorbia esula L.) populations was undertaken using Py-MS-PR. Accessions of E. esula corresponding to eight different populations were effectively characterized/classified into distinct biotypes based on the chemical variation in their plant latex as reflected in their Py-mass spectra. The last study details the methodology and results of Py-MS-PR analysis of leaf material from shadscale ( Atriplex confertifolia, [Torr & Frem] Wats.) samples representing four ploidy levels from eight geographic locations, collected over two seasons (August 88 and April 1989). The chemical information contained in the Py-MS data, when processed by unsupervised pattern recognition, clearly discriminated each geographic location and gave some correlation to ploidy. When supervised pattern recognition (discriminant function analysis) was applied to the Py-MS data, the samples could be correctly classified as to ploidy regardless of season.