The use of spectra derived from six independent but identical experiments
also allowed an assessment of the variability between separate
experiments and whether individual experiments performed at different
times could be distinguished by FT-IR spectroscopy. By plotting
the data shown in Fig. 2c on the basis of experiment, it could be
shown that clustering of non-stressed (standard TAP) and nutrient
stressed (Low-P and Low-N TAP) samples was good even though spectral
data was derived from different experiments (Fig. S2). Whilst there
was some variation accounted for by experimental timing, such as some
variation from Experiment 6 (Fig. S2b, c), these differences do not affect
the ability of PCA to distinguish between samples grown in different nutrient
replete and nutrient limited media. It should be noted that all six
experiments were performed under environmentally controlled conditions
with parameters such as light intensity and temperature as identical
as possible. This suggests that if a standard metabolic fingerprint is
obtained, such as a species-specific fingerprint, with a high level of experimental
control species classification may be possible, depending
on whether species-dependent characteristics can be determined from
FT-IR spectral information.