Title: Visualization,
screening and classification of cell cycle-regulated genes in yeast
by multi-dimensional
scaling, nonlinear dimension reduction and wavelet transform
We propose
a new and integrated approach for visualization,
screening,
and prediction of gene functions using microarray data,
based on multidimensional
scaling (MDS), nonlinear dimension
reduction
and wavelet transform. This approached is applied to
analyze the
cell cycle of yeast in Spellman et al. (1998).
The results
show that this new approach indeed provides a
visualization
tool, which displays the functional relationships
between genes,
like the periodical pattern of a cell cycle. This
representation
can be used to verify the functions of genes known
by experimental
methods in literature. Based on this, screening
and prediction
of the functions of all genes can be implemented.
The performances
of different methods in screening and prediction
are evaluated
by the Jackknife approach and compared in this
study. Hence,
this integrated approach suggests a new perspective
to classify
and discover the functions of genes by their
expression
profiles in microarray. The findings for yeast cell
cycle are
also reported and discussed.