Premium
Assessing quality and normalization of microarrays: case studies using neurological genomic data
Author(s) -
Hershey A. D.,
Burdine D.,
Liu C.,
Nick T. G.,
Gilbert D. L.,
Glauser T. A.
Publication year - 2008
Publication title -
acta neurologica scandinavica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.967
H-Index - 95
eISSN - 1600-0404
pISSN - 0001-6314
DOI - 10.1111/j.1600-0404.2007.00979.x
Subject(s) - normalization (sociology) , bioconductor , outlier , dna microarray , preprocessor , microarray analysis techniques , microarray , database normalization , computer science , gene chip analysis , data mining , data pre processing , identification (biology) , computational biology , artificial intelligence , biology , pattern recognition (psychology) , gene , genetics , gene expression , botany , sociology , anthropology
Background – Genomic analysis using microarray tools has the potential benefit of enhancing our understanding of neurological diseases. The analysis of these data is complex due to the large amount of data generated. Many tools have been developed to assist with this, but standard methods of analysis of these tools have not been established. Objective – This study analyzed the sensitivity and specificity of different analytical methods for gene identification and presents a standardized approach. Methods – Affymetrix HG‐U133 plus 2.0 microarray datasets from two neurological diseases – chronic migraine and new‐onset epilepsy – were used as source data and methods of analysis for normalization of data and identification of gene changes were compared. Housekeeping genes were used to identify non‐specific changes and gender related genes were used to identify specific changes. Results – Initial normalization of data revealed that 5–10% of the microarray were potential outliers due to technical errors. Two separate methods of analysis (dChip and Bioconductor) identified the same microarray chips as outliers. For specificity and sensitivity testing, performing a per‐gene normalization was found to be inferior to standard preprocessing procedures using robust multichip average analysis. Conclusions – Technical variation in microarray preprocessing may account for chip‐to‐chip and batch‐to‐batch variations and outliers need to be removed prior to analysis. Specificity and sensitivity of the final results are best achieved following this identification and removal with standard genomic analysis techniques. Future tools may benefit from the use of standard tools of measurement.