TABLE II.

Studies Examining the Correlation Among Microarray Technologies from 2004 to 2006

Publication	Platforms	Probe ID	Validation	AUthors' conclusion
Yauk et al. [2004]	Affymetrix, Agilent cDNA, Agilent oligo, Codelink (Amersham), Mergen, NIA cDNA	Unigene ID	None	Agreement Good platforms correlate well. Expression profiles clustered by biology rather than technology. cDNA platforms less sensitive than oligonucleotide
Woo et al. [2004]	Affymetrix, in-house spotted cDNA and oligonucleotide	MGI identifiers	None	Agreement Good concordance in expression level and statistical significance between Affymetrix and oligonucleotide arrays; cDNA arrays showed poor concordance with other platforms
Jurata et al. [2004]	Affymetrix, Agilent cDNA	Unigene ID	Real time RT-PCR	Moderate agreement Gene changes overlapping between the two platforms were co-directional; RT-PCR validation rates were similar
Park et al[2004]	Affymetrix, Agilent cDNA, custom oligonucleotides from three different sources printed by Agilent	Unigene ID	Real time RT-PCR	Agreement Expression level correlation low, but log ratios high; correlation is stronger for highly expressed genes
Mecham et al. [2004a]	Affymetrix, Agilent cDNA	Sequence matched	None	Agreement Cross-platform analysis greatly improved by sequence matching
Mah et al. [2004]; Two different labs	Affymetrix, cDNA	Unigene (sequence matched)	Real time RT-PCR	Disagreement Poor correlation when matched by expression level
Jarvinen et al. [2004]	Affymetrix, Agilent cDNA, custom-cDNA	Unigene ID	None	Moderate agreement Good correlations for commercial, whereas the correlations between the custom-made and either commercial platforms were lower. Discrepant findings due to clone errors, old annotations, or unknown causes
Shippy et al. [2004]	Affymetrix, Codelink (Amersham)	Unigene ID	Real time RT-Pcr	Agreement After noise adjustment (use precent)genes
Ulrich et al. [2004];ILSI/HESI collaboration	Affymetrix, Clontech, Incyte, NIEHS, Molecular Dynamics, PHASE-1	Comparison of pathways	Real time RT-PCR	Agreement Correlation of the biological pathways involved in response to toxicant exposure
Stec et al. [2005]	Affymetrix, Millennium Pharmaceuticals cDNA	UniGene ID, then sequence matched using BLAST	None	Moderate agreement Increased significantly after sequence matching; discrepant correlations between Affymetrix and cDNA measurements could be explained by probe sequence differences
Irizarry et al. [2005]; Lab-lab comparison (10 labs)	Affymetrix (5 labs), cDNA (3 labs), 2-color oligonucleotide (Qiagen 70mer; analyzed in 2 labs)	Unigene, locuslink, RefSeq	Real time RT-PCR	Agreement Among best performing laboratories; increased data quality with more stringent pre-processing
Dobbin et al. [2005]; Cross-laboratory comparison	Affymetrix (inter-and intra-laboratory correlation)	None required	None	Agreement Intra-laboratory correlation was only slightly stronger the inter-laboratory. Samples clustered by biology rather than laboratory
Larkin et al. [2005]	Affymetrix, TIGR cDNA	Sequence mapped TIGR	Real time RT-PCR	Agreement Biological treatment had a greater effect on gene expression than platform for 90% of the genes
Bammler et al. [2005]; Lab-lab comparison 7 labs, 12 platforms	Affymetrix, Agilent Amersham (Codelink), Compugen, Operon, 2 custom oligo, 5 custom cDNA,	Transcripts matched using NIA mouse index	None	Moderate agreement Standardized protocols and data analysis required
Pylatuik and Fobert [2005]	Affymetrix, Genomic Amplicon arrays, Operon oligo	Locus ID (Arabidopsis)	Northern blot	Moderate agreement Signal intensity-dependant
Shi et al. [2005a]	Tan et al., 2003; dataset	Genbank acc.No	N/A	Agreement Concluded that the quality of the original dataset was poor and inappropriate methods were applied for analysis. Alternate analysis had 10× >concordance
Bames et al. [2005]	Affymetrix, Illumina Beadarrays	Sequence matched using BLAST [Kent, 2002]	None	AgreementFor genes with high expression and concordance improved for probes that were verified to target same transcript
Gwinn et al [2005]	Affymetrix, Amersham (Codelink), cDNA	Amersham (locuslink ID) Affymetrix (GenBank)–links made via IDs given by company; genes of interest used probe sequence	Real-time RT-PCR	Disagreement Each platform yielded unique gene expression profiles
Petersen et al. [2005];	Affymetrix, in-house cDNA, in-house Operon oligonucleotide	Unigene	Real time RT-PCR	Agreement High concordance for significant expression ratios and 1.5 to 2-fold changes (93–99%)
Carter et al. [2005]	Affymetrix, Stanford cDNA	Sequence matched	None	Agreement Re-examination of NCI-60 cell line data. Overlapping probes correlate well
Wang et al. [2005]; 5 data sets were either collected from a public source or generated in house	Affymetrix, Agilent oligo, cDNA	Unigene ID	None	Agreement Data were more consistent between two commercial platforms and less consistent between custom arrays and commercial arrays; expression at the gene level exhibited an acceptable level of agreement. Lab and sample effect was greater than platform effect
Schlingemann et al. [2005]	Affymetrix, in-house long oligo	Unigene ID	Real time RT-PCR	Agreement Similar profiles and strong correlations were found for the 2 platforms
Warnat et al. [2005]	6 different cDNA and oligo array studies previously published from several laboratories	Unigene ID	None	Agreement Integrated raw microarray data from different studies for supervised classifications. More platforms better for predictive analysis
Ali-Seyed et al. [2006]	Affymetrix, Applied Biosystems	Promote analysis	Real time RT-PCR	Agreement AB more sensitive and more correlated with RT-PCR
Severgnini et al. [2006]	Affymetrix, Amersham (codelink)	Locuslink ID	Real time RT-PCR	Disagreement Only 9 genes found to be differentially expressed in common out of 42 (Affymetrix) and 105 (Codelink) in total
de Reynies et al. [2006]	Affymetrix, GE Healthcare (Amersham), Agilent	Sequence mapped	Real time RT-PCR	Moderate agreement 1 color more precise than 2 color; Affymetrix and Agilent were more concordant based on detection of differential genes
Wang et al. [2006]	Agilent, Applied biosystems	Sequence matched (BLAST)	Real time RT-PCR	Agreement 1375 genes confirmed with RT-PCR
Kuo et al. [2006]; Lab-lab comparison as well	Affymetrix, Amersham, Mergen, ABI, custom cDNA, MGH, MWG, Agilent, Compugen, Operon	Probes sequence matched within 1 exon (Unigene, LocusLink, RefSeq, Refseq exon)	Real time RT-PCR	Strong agreement Commercial better than in-house, 1-color better than 2-color
Shi et al. [2006]; Lab to lab comparison (3 sites); used commercial RNA sources; MAQC	Affymetrix, Agilent (1 and 2 color), Applied Biosystems, Eppendorf, GE Healthcare, Illumina, in-house spotted Operon oligonucleotide	Probes sequence mapped to RefSeq and to AceView using 30 probe for genes with multiple oligonucleotides	Real time RT-PCR; TaqMan1 (Roche Molecular Systems); StaRTPCR and QuantiGene carried out by Canales et al. [2006]	Strong agreement Intra-platform consistency across test sites and high inter-platform concordance with respect to differentially expressed genes; high correlation between QRT-PCR values and microarray results
Guo et al. [2006]; Included 2 test sites for Affymetrix; MAQC	Affymetrix, Agilent, Applied Biosystems, GE Healthcare	Probes sequence mapped to RefSeq	None	Agreement High inter-site and cross-platform concordance in the detection of differential gene expression using fold change rankings; fold change ranking outperforms other analysis methods
Canales et al. [2006]; MAQC	Affymetrix, Agilent, Applied Biosystems, Eppendorf, GE Healthcare, Illumina	Probes sequence mapped to RefSeq	Real time RT-PCR; TaqMan1 (997 genes), StaRT-PCR (205 genes), and QuantiGene (244 genes)	Agreement High correlation between gene expression values and microarray results. Main variable was probe sequence and target location