Question: Comparing Microarray Expression Values
gravatar for Liz
7.8 years ago by
Liz30 wrote:

Is it possible to compare microarray expression values that have been normalized by different methods?

For example, I would like to compare published data sets, one has been normalized "by log2-transformation, subtraction of median, and division by standard deviation" and the other used lowess normalization.

Also, I would like to compare my own data with these data sets - I have performed a one colour array - which normalization method would be the most appropriate? I have no experience with bioinformatics tools or coding and need the most simple method possible.

data microarray • 1.8k views
ADD COMMENTlink written 7.8 years ago by Liz30
gravatar for Will
7.8 years ago by
United States
Will4.5k wrote:

Sadly you're probably SOL if you don't have the raw data before normalization. Without knowing the column medians (and standard deviations) from the first set and the Lowless parameters you wont be able to get comparable levels.

You can try to do a meta analysis. This basically consists of finding significant genes from each dataset individually. Then finding enriched functions/pathways using a tool like David. Then comparing the overlapping functions across the datasets.

Sadly if a study only publishes normalized microarray data its nearly impossible to reverse-engineer it into other contexts.

ADD COMMENTlink written 7.8 years ago by Will4.5k

@Liz That's right, It's essential to have raw data generated by scanner. So you can use raw foreground and background intensities to normalize data in similar way (according to commonly used method for a given platform). Then You can use one of the rank methods (RankProd package in R) or MetaArray (computes Probability of expression = POE which gives transforms data to the same scale) and for example t-test from multtest package (all software from bioconductor).

ADD REPLYlink written 7.7 years ago by boczniak767640
gravatar for brentp
7.8 years ago by
Salt Lake City, UT
brentp23k wrote:

You could try doing some rank order stuff, where you just give each probe a number based on it's order. There seems to be some precedent: "Combining Results of Microarray Experiments: A Rank Aggregation Approach" (pdf)

ADD COMMENTlink written 7.8 years ago by brentp23k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1523 users visited in the last hour