Metabolomics is the the name given to a range of methodologies
that all have a similar goal, namely, comprehensive understanding
of biological systems. Details vary by application and preferred
technology methodology, but the same general concepts apply
regardless of the specific approach.
First, an experimental design is executed, wherein treatment
effects are defined (e.g. diseased vs non-diseased). Next, samples
are taken and subjected to an analytical chemistry technique to
measure as many chemical species as possible. Applying statistics
to the resulting data, signals are identified that distinguish the
treatment groups.
Finally, in the most informative application of metabolomics,
these changes can be mapped to individual chemical species, which
themselves are mapped to biochemical pathways. This level of
biochemical understanding can reveal drug or disease mechanisms of
actions, as well as uncover alternate or unexpected biological
effects.
However, in the more common approach used today, individual
chemical species are not identified on a large scale, but the many
changing signals are used to construct mathematical models based on
a training set, which, presumably, can be used to classify test
samples. Methods can be supervised or unsupervised. Although this
classification approach is more prevalent today, primarily because
it is much easier to apply than the approach where individual
chemicals are identified, it is starting to lose favor, since after
more than 15 years, it has never been demonstrated to be useful for
gaining useful, previously unknown biological insights.