Implementation of an FTIR spectral library of 486 filamentous fungi strains for rapid identification of molds

Filamentous fungi may cause food and feed spoilage and produce harmful metabolites to human and animal health such as mycotoxins. Identification of fungi using conventional phenotypic methods is time-consuming and molecular methods are still quite expensive and require specific laboratory skills. In the last two decades, it has been shown that Fourier transform infrared (FTIR) spectroscopy was an efficient tool for microorganism identification. The aims of this study were to use a simple protocol for the identification of filamentous fungi using FTIR spectroscopy coupled with a partial least squares discriminant analysis (PLS-DA), to implement a procedure to validate the obtained results, and to assess the transferability of the method and database. FTIR spectra of 486 strains (43 genera and 140 species) were recorded. An IR spectral database built with 288 strains was used to identify 105 different strains. It was found that 99.17% and 92.3% of spectra derived from these strains were correctly assigned at the genus and species levels, respectively. The establishment of a score and a threshold permitted to validate 80.79% of the results obtained. A standardization function (SF) was also implemented and tested on FTIR data from another instrument on a different site and permitted to increase the percentage of well predicted spectra for this set from 72.15% to 89.13%. This study confirms the good performance of high throughput FTIR spectroscopy for fungal identification using a spectral library of molds of industrial relevance. (C) 2014 Elsevier Ltd. All rights reserved.