Premium
Reliable identification of closely related Issatchenkia and Pichia species using artificial neural network analysis of Fourier‐transform infrared spectra
Author(s) -
Büchl Nicole R.,
Wenning Mareike,
Seiler Herbert,
MietkeHofmann Henriette,
Scherer Siegfried
Publication year - 2008
Publication title -
yeast
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.923
H-Index - 102
eISSN - 1097-0061
pISSN - 0749-503X
DOI - 10.1002/yea.1633
Subject(s) - biology , artificial neural network , identification (biology) , fourier transform , biological system , fourier analysis , pattern recognition (psychology) , computational biology , artificial intelligence , computer science , mathematics , ecology , mathematical analysis
A reliable identification system for closely related species of the genera Issatchenkia and Pichia was established, using artificial neural network‐based Fourier‐transform infrared (FTIR) spectroscopy; 16 common Pichia species and all five known Issatchenkia species were included. A total of 238 strains isolated from a large variety of habitats were used as reference strains to generate an artificial neural network (ANN) identification system. This system consists of 10 single subnets connected to an ANN with four consecutive levels. An internal validation of the system, using unknown spectra of each reference strain, yielded an identification rate of 99.2%. To evaluate the performance of the ANN in routine diagnostics, 1608 spectra of 179 strains unknown to the ANN were used as a test dataset in an external validation. An overall identification rate of 98.6%, including a success rate of 100% for two common species, P. anomala and P. membranifaciens , demonstrates considerable potential of this FTIR‐based artificial neural network for the identification of closely related yeast species. Copyright © 2008 John Wiley & Sons, Ltd.