Tech Notes

Q. What is the Lowess normalization in GeneSpring?

A. . Lowess Intensity dependent normalization is used for two-color data. It is a type of within-slide normalization scheme that adjusts for intensity-dependent variation due to dye properties. Dye bias is caused by inconsistencies in the relative fluorescence intensity between Cy5 and Cy3. These inconsistencies often result in non-linear relationship between the two dyes and a curve will occur in a scatter plot:

By applying Lowess normalization to your data in GeneSpring, a locally-weighted regression is fitted to the data to adjust the control value for each measurement. As a result, the plot of signal versus control value will be linear, as shown in the plot below:

NOTE: Lowess normalization is not recommended for arrays or regions with less than 1,000 genes.

More information on the Lowess normalization algorithm can be found in the GeneSpring user manual. Lowess normalization is referenced in the publication that can be found at: http://stat-www.berkeley.edu/users/terry/zarray/Html/normspie.html.

NOTE: Lowess normalization is a global normalization method. It is not intended for arrays in which most genes are expected to vary during the experiment.