In:
The Journal of Immunology, The American Association of Immunologists, Vol. 184, No. 1_Supplement ( 2010-04-01), p. 144.11-144.11
Abstract:
Gene expression microarrays are widely used to study genome-wide gene expression profiles. Many microarray analysis methods are available, making it challenging to decide which method to use. While the effectiveness of some of these methods has been assessed using artificial spike-in data sets, analytical approaches that work well with spike-in data may not work as well with data from real biological samples. To evaluate these methods we applied Gene Ontology (GO) term co-clustering as a comparative tool to evaluate 300 different data processing pipelines composed of various background correction, normalization and summarization methods using real biological data, based on the premise that an improvement in any step of microarray data analysis should be reflected in improved co-clustering of related genes. Our results suggest that background correction has little affect on GO term co-clustering characteristics, normalization has a big impact, some summarization methods constantly outperform others, some interdependencies exist among these methods, e.g., a good normalization method may work well with one summarization method but not another. An automated method for running and assessing these 300 pipelines has been made available through the Immunology Database and Analysis Portal (ImmPort; www.immport.org) - a public database resource funded by NIAID to support the management and analysis of clinical and mechanistic data generated by their funded investigators.
Type of Medium:
Online Resource
ISSN:
0022-1767
,
1550-6606
DOI:
10.4049/jimmunol.184.Supp.144.11
Language:
English
Publisher:
The American Association of Immunologists
Publication Date:
2010
detail.hit.zdb_id:
1475085-5
Permalink