Pre-processing in biochemometrics: correction for path-length and temperature effects of water in FTIR bio-spectroscopy by EMSC

The term 'biochemometrics' is proposed. Methodology is presented for simplifying high-resolution phenotyping measurements, in terms of multivariate modelling methods in order to stabilise FTIR bio-spectroscopy data. Irrelevant gas contributions from water vapour and CO2 in the instrument light path are modelled and removed, in order to avoid having to wait for N-2 purging. Variations in the infrared spectroscopy (IR) spectrum of water with temperature are described in terms of two model component spectra. These additive water variations are quantified and eliminated by Extended Multiplicative Signal Correction (EMSC), along with various physical signal variations of additive and multiplicative nature due to, for example, sample or instrument temperature. Sample temperature is predicted from the EMSC model parameters. The models developed for Attenuated Total Reflection (ATR) measurements of pure water at different temperatures are tested successfully in independent water samples as well as in the in vivo monitoring of Candida albicans growing and decaying on the ATR crystal of the same instrument. Copyright (c) 2007 John Wiley & Sons, Ltd.