Saturday, December 17, 2005
Lattice of ontologies
A review of this paper
is suggested:
http://www.genome.org/cgi/content/full/11/7/1149
A philosophical
position is defined and taken.
Two scientific
methodologies are identified. The first
is the use of what are called “systematic methods” to reconstruct evolutionary history
of various gene expressions, ie phenotypes.
The second also uses
systematic methods to make an analysis of gene expression.
In cognitive science
“systematic methods” means a characteristic of research
studies that involves the careful and systematic use of research
procedures. In gene research the phrase
often means the use of cluster techniques such as Kohonen self-organizing maps
or some type of latent semantic indexing type similarity matrix – which also
results in clusters of data – indicating some stable phenomenon. (see {link} )
Systematic
methods for organizing gene expression data require a means of measuring quantitatively
if two expression profiles are similar to each other. In this regard it is
useful to consider the values that make up the expression profile for a single
gene as a series of coordinates, which define a vector, and to consider the
data for a microarray experiment as a matrix, where the genes define the rows,
and the arrays, or experiments define the columns. Once we consider these data
as vectors, we can use standard mathematical techniques to measure their
similarity. One distance metric that can thus be used is the Pearson
Correlation, which is essentially a measure as to how similar the directions in
which two expression vectors point are. The Pearson Correlation treats the
vectors as if they were the same (unit) length, and is thus insensitive to the
amplitude of changes that may be seen in the expression profiles. A second
distance measure that can be used is the Euclidean distance, which measures the
absolute distance between two points in space, which in this case are defined
by two expression vectors. The Euclidean distance thus takes into account both
the direction and the magnitude of the vectors. {link}
The results
of this analysis has to be placed into a formalism, and this formalism is often
a tree like structure with relationships, properties, attributes and
facets.