Project VI: Normalization and standardization in
microarray experiments
One
problematic area in the analysis of microarray data is that of array
normalization: typically there is considerable systematic, but not expression
related variation within an array and variation across replicate arrays. Before a set of differentially-expressed
genes can be identified, the array must be normalized so that spurious results
are not reported. The sources of this
unwanted variation between arrays are often due to slight differences in the
construction of the arrays, or different spotting machinery, or imperfect
imaging results. This phenomenon arises
in cDNA and oligonucleotide arrays.
Many
methods of normalization/standardization of microarray data have been proposed,
and this project will investigate some of them.
Methods of within-array adjustment (correcting for nonlinear
relationships between total signal ad observed
match/mismatch difference) and between-array standardization (baseline methods,
quantile normalization) will be studied.
A
large test data set, concerning the transcriptome derived from human
chromosomes 21 and 22 will be investigated in some detail.