Identical probes on different high-density oligonucleotide microarrays can produce different measurements of gene expression
© Zhang et al; licensee BioMed Central Ltd. 2006
Received: 30 March 2006
Accepted: 15 June 2006
Published: 15 June 2006
There are many potential sources of variability in a microarray experiment. Variation can arise from many aspects of the collection and processing of samples for gene expression analysis. Oligonucleotide-based arrays are thought to minimize one source of variability as identical oligonucleotides are expected to recognize the same transcripts during hybridization.
We demonstrate that although the probes on the U133A GeneChip arrays are identical in sequence to probes designed for the U133 Plus 2.0 arrays the values obtained from an experimental hybridization can be quite different. Nearly half of the probesets in common between the two array types can produce slightly different values from the same sample. Nearly 70% of the individual probes in these probesets produced array specific differences.
The context of the probe may also contribute some bias to the final measured value of gene expression. At a minimum, this should add an extra level of caution when considering the direct comparison of experiments performed in two microarray formats. More importantly, this suggests that it may not be possible to know which value is the most accurate representation of a biological sample when comparing two formats.
Microarrays are rapidly becoming a standard tool for molecular biologists. The technology can reveal a multitude of information from even the simplest biological experiment. The technique is primarily used by scientists to generate ideas in experiments colloquially referred to as "fishing expeditions". However, the result of even simple experiments is hundreds to thousands of gene expression changes. Because of this the cost of validation is prohibitive. The microarray technology itself is expensive and this often prevents replication. Furthermore, the preliminary experiments are usually composed of the minimum number of samples absolutely necessary to address the question. This minimalistic approach is also used for other common microarray experiments. Rather than replicate an experiment or increase the size of a study, it is increasingly common for researchers to turn to other published microarray experiments to attempt to verify or expand their findings. Early attempts at cross-platform verification of microarray experiments utilized data generated in other laboratories from independent experiments [1–6]. These attempts produced limited success and undoubtedly many more studies went unpublished because of the difficulty in reproducing the results of one assay in a second platform.
However, recent studies have shed a more positive light on cross-platform comparisons. Several studies have now shown that the probes used to detect transcripts are the root cause of differences between platforms [7–10]. The probes on some arrays do not detect the transcripts attributed to them [11, 12]. This is especially a concern for cDNA-based arrays [13, 14]. Many commercial platforms are utilizing oligonucleotide-based probes. However, commercial array producers do not always accurately identify the targets of their probes [15, 16]. There are also limitations to our understanding of the transcriptome and therefore two different probes that, theoretically, detect the same transcript may produce different measurements in microarray experiments because of cross-hybridization from unknown splice variants, gene families, and transcribed pseudogenes. Nonetheless, the correlation between different arrays is quite good if one restricts the analysis to sequence matched probes [9, 10, 17–19].
Oligonucleotide based arrays appear to have higher resolution and lower variability than cDNA based arrays [2, 20, 21]. They are also easier to compare across-platforms because the sequences can be easily cross-matched. Comparison is even easier when the exact same sequence is used on two different arrays. The most recent Affymetrix GeneChips contain probes that were previously available on earlier arrays. Therefore the sequences are identical and would be expected to produce similar measurements from the exact same samples. We tested this assumption and found that identical probes on the two arrays can produce different measurements.
Controlled hybridization experiment
We decided to investigate this phenomenon further to be certain that the differences were not due to unknown bias introduced during processing of the samples. To minimize the possible sources of bias, we processed 5 aliquots of the Reference RNA in parallel using the same reagents. Following the production of labeled RNA we prepared enough hybridization mixture from each preparation to hybridize both a U133A array and a U133 Plus array at the same time. We further processed the samples in parallel to minimize staining and scanning issues. Following this carefully controlled experiment we examined the U133A probeset values relative to the same probesets on the U133 Plus arrays. This time the t-test indicated that more than 5000 probesets were yielding different values on the two array types (p-value < 0.05). Nearly 10% of the evaluated probesets were strongly separated into distinct groups (2191 probesets, p-value < 0.01). There were 1568 probesets in which every recorded measurement from the A chip was larger than every recorded measurement from the Plus chip. There were 869 probesets in which every recorded measurement from the Plus chip was larger than every recorded measurement from the A chip. By chance alone we would only expect 43 such occurrences.
Although most microarray laboratories cannot mix array types for batch methods such as RMA or MBEI, we were able to generate artificial A CEL files from the appropriate probe values on the U133 Plus CEL files. This allowed us to perform RMA on the 10 arrays. Following quantile normalization and the model-based signal estimation of RMA more than 9000 probesets produced different values indicating that the array specific differences are not a consequence of a specific method of normalization or signal estimation. In addition, the A array files fabricated from the Plus array data allowed us to perform several types of normalization at the probe or probeset level similar to the detailed analysis of Choe et al. (). None of the standard methods used for normalizing microarray data or calculating the probeset values reduced the number of differences between the two array formats to a level expected by chance (data not shown).
Across array type versus within array type comparisons
When arrays of the same type were compared the probesets that appeared to be differentially measured were probably all due to random noise. Approximately 500 probesets appeared to have group specific values in any 5 array to 5 array comparison (p-value < 0.05) within the same array type. A different 5 array by 5 array comparison yielded a different group of 500 probesets. Less than 2% of the probesets were in common between any two gene lists produced by such 5 by 5 comparisons. This illustrates that the apparent differences were nothing more than random variation that happened to fall into distinct distributions when the groups were formed. In contrast, the probesets found to be different when comparisons were made across the array types were more consistent. Approximately 5000 differences were observed at a p-value < 0.05 when 5 U133 Plus arrays were compared to a group of 5 U133A arrays. There was 60 to 85% overlap between any two lists generated by comparisons across array types. This suggests that the same probesets were being identified no matter which set of samples were chosen for comparison. This is strong evidence that the differences are real and not chance events.
Latin square hybridization control
Comparison between the variability introduced by hybridization and staining effects alone, the cDNA synthesis and transcription process, and cross-platform comparisons. Each value represents the percentage of probesets in each quantile of the array where measurements diverged more than 2-fold from the mean.
Evaluation of individual probes and probesets
Cross-platform behavior of selected probesets that detect the same transcript in the reference RNA.
Plus chip range
A chip range
Adducin 3 (gamma)
dimethylarginine dimethylaminohydrolase 2
Lysosomal-associated multi-spanning membrane protein-5
nuclear transcription factor Y, gamma
polyglutamine binding protein 1
Putative prostate cancer tumor suppressor
ribosomal protein, large, P0
staufen, RNA binding protein (Drosophila)
U2 small nuclear RNA auxiliary factor 1-like 2
Sequence of the probes used to detect KIAA0676 protein and their location on the arrays.
Probe sequences for KIAA0676 protein
Plus chip location
A chip location
Probe sequences for KIAA0676 protein
Plus chip location
A chip location
The transcript for KIAA0676 is detected by 4 different probesets. Three of these probesets behave very similarly on the two array types. The fourth probeset behaves in the opposite manner (table 2). The individual probes in each probeset, and their locations on the two arrays, are shown in table 3. The probesets that show the same behavior are nearly identical while the divergent probeset is distinctly different from the other three. On average, 9 of the 11 individual probes are identical for the 3 similarly acting probesets. The identically designed probes are also located side-by-side on the respective arrays. Therefore, these three probesets are virtually identical in both sequence and location on the arrays and identical in the bias seen between the array types. This fact was also true of several other parallel probesets we examined (data not shown).
We looked at the measurements recorded for the individual probes that comprise these probesets. We found that 10 of the 11 probes produced higher values on the U133A arrays than on the U133 Plus arrays for probesets 206431_x_at, 215994_x_at, and 212054_x_at. Conversely, the probeset 212052_s_at, which behaved in the opposite manner, had more individual probes yielding a higher value on the U133 Plus arrays. Therefore bias at the probe level appears to generate the bias observed for the full probeset. We looked at all the individual perfect match probes for the probesets that produced different measurements on the two array types. When the probeset produced a larger value in the U133A arrays an average of 7.6 individual probes (out of 11) were higher in the A arrays than in the U133 Plus arrays. Conversely, when the Plus chip produced a higher value, an average of 7.7 probes from these probesets were higher in the U133 Plus environment.
We suspected that this bias might be due to a little bleed over from the surrounding probes on the array. We examined a number of individual probes. In many cases brighter probes surrounded the probes producing the higher values; however this was not universally true. More consistently, it appeared that the probes yielding higher values were just a little brighter on the arrays with the higher measurements. This suggested a possible manufacturing difference as the underlying cause rather than bleed over from neighboring probesets.
One of the most important considerations in performing a microarray experiment is that the data obtained at the end is an accurate representation of the RNA present in the biomaterial used to start the process. Bias can lead to erroneous conclusions if the bias happens to track with the experimental condition evaluated or obscures the differences because the bias is larger than the biological effect. Bias can come from many sources [25–27]. When properly identified, bias can be corrected and a proper analysis can proceed. Our data demonstrates that the array itself can contribute bias to gene expression measurements. It is quite understandable how different probe sequences could lead to different measurements in gene expression arrays. However, our data shows that even identical probe sequences can yield slightly different measurements of gene expression.
The impact of probe level variation depends on the nature of the question addressed in a microarray experiment. For a simple experiment performed in either platform the observed performance of a gene should be substantially reproduced by the alternative platform. However, if clinical samples are being accumulated for a disease state, the differences between platforms might prevent the pooling of samples processed in both platforms. This depends on the differences one finds in the experimental system. Small differences might be obscured while large differences would still be observable. We observed that many of the probesets with target genes present in the Universal Reference RNA showed differences greater than 2-fold. This bias could compromise experiments where the number of samples evaluated might otherwise allow one to detect differences less than 2-fold.
Our data shows that microarray data can be very consistent at the same time as it shows a bias. Overall, any two arrays produced with the Universal Reference RNA yielded fairly consistent microarray results. The final measurements form a tight cluster of values for most probesets and the overall correlation was 0.96 for any two arrays. However, a significant number of probesets produced two distinctly different or partially overlapping distributions when the array formats were viewed separately (figure 2 and 3). Therefore the bias due to the array is clearly evident even though the overall correlation was very high. Sometimes a measurement was higher on the U133A arrays and sometimes it was higher in the U133 Plus arrays. This could even be true for the same transcript, as one probeset yielded higher values and another probeset, detecting the same transcript, yielded lower values than the alternative platform. This shows that the bias was not due to processing, hybridization, or normalization artifacts. This observation has larger implications. It illustrates that it may not be possible to have all probes on an array performing exactly the same way so that all transcripts could be measured on equivalent scales.
The trend in microarray research is towards generating larger sets of data for analyzing complex biological problems and diseases. Institutions are pooling resources to attain the large datasets required for analysis. Several groups have already evaluated the comparability of data produced at distant sites [28–30]. There are also attempts to directly compare gene expression data from quite dissimilar microarray platforms as well as serial analysis of gene expression (SAGE) and RT-PCR [31–34]. This push towards the direct comparison of microarray data from distant sites and dissimilar platforms is likely to continue. It is important to consider the differences in measurement observed on different array types.
Since their introduction, microarrays have been forecast for clinical use. Many people believe that patterns of gene expression will be used to identify disease states. Diagnosis, prognosis, and treatment options are all believed to lie in patterns of gene expression. While some people begin the search for these patterns, the microarray community has also been working towards improving the technology to produce more reliable results. Some sources of bias can be minimized or corrected. Others will simply be accepted as part of the process. Classifiers developed to interpret microarray data must allow for any variation that occurs in the measurement of individual probes. It remains to be determined whether the variation caused by the type of array will interfere with the performance or development of classifiers. For the present time it seems wise to use a single array type for the evaluation of microarray data. If the classification of samples relies on small gene expression differences, the classification may be microarray format specific. However, large differences will probably still be observed despite the small differences reported here between the U133A arrays and the U133 Plus arrays.
Microarray data is often useful beyond the intentions of the original experiment. The microarray community is continuously developing standards for microarray processing and data management that allows scientists to utilize microarray data held in repositories. Much like sequence data, this stored gene expression information can be used in a multitude of creative ways. The comparison of microarray data between two formats, or even between two laboratories using the same format, requires knowledge about the sources of error that can arise in a microarray experiment. We have demonstrated that the same sequence can provide slightly different measurements of gene expression in different array formats. This implies that the comparison of microarray data between formats may require an additional, array specific, correction factor for each probe. The larger implication is that it may not be possible to establish equivalent correlations between the measured value in an array and the absolute value in a biological sample for every gene on an array. If sequence, as well as sequence context, introduces subtle adjustments to the final measured value of a transcript, then it may not be possible to know which measurement is the most accurate measurement of transcript abundance. Therefore attempts to perform cross-platform verification either have to use gene specific correction factors or be satisfied with similarity rather than exact replication.
The source of RNA was the Universal Human Reference RNA from Stratagene (Catalog number 740000, Stratagene, La Jolla, CA) for all samples. This RNA was processed using the established Affymetrix protocols for the generation of biotin-labeled cRNA, hybidization, staining, and scanning. A more complete description of this process is available in the papers by Warrington et al.  or Dobbins et al. . Hybridizations were performed as indicated in, either HG-U133A (U133A) arrays, or Human Genome U133 Plus 2.0 (U133 Plus) arrays from Affymetrix, Inc. (Santa Clara, CA). The U133A arrays were scanned on an Affymetrix GeneChip® scanner 3000 at a 2.5 μm resolution and the U133 Plus chips were scanned at 1.56 μm resolution. All data was initially generated from the scanned image files using the MAS 5.0 software embedded in the GeneChip Operating Software from Affymetrix. The data was initially normalized globally to an average intensity of 500. The signal values from each array were then exported to a text file. The probesets in the U133 Plus arrays but not found on the U133A arrays were removed from these files and then all the arrays used in this study were renormalized in a single group using the iterative process described by Li and Wong as well as by Wang et al. [22, 23]. This process insured that the final normalization was based on the most stable gene expression measurements across the arrays and that the probesets on the U133 Plus arrays, but not represented on the U133A arrays, did not influence the final normalized values.
All data used in this manuscript are publicly available for download. The initial samples placed on U133A arrays are available at the Gene Expression Database Portal (, experiment IDs 615–618). The samples hybridized to U133 Plus Arrays are from a series evaluating the effect of starting RNA quantity. The array data is available at the Gene Expression Omnibus () under accession number GSE3062. The samples processed and then hybridized to both a U133A array and a U133 Plus array are available under the accession number GSE3061.
The Latin Square data set used for comparison was generated by scientists at Affymetrix and is available from the company's website .
For the probe level analysis the values in each CEL file were loaded into a database. The probe values corresponding to those probesets found on the U133A arrays were extracted from each U133 Plus array and used to generate an artificial U133A CEL file. All the CEL files were then normalized to the same average value and individual probe values were compared across arrays. This technique also allowed comparisons using batch methods such as RMA or MBEI and perform other normalization techniques on the probe level data. The artificial U133A CEL files and corresponding U133A CEL files were loaded into a local implementation of the RMA program available through Bioconductor to calculate the expression values.
This work was funded by a National Functional Genomics Center Grant from the Department of Defense, DAMD 17-02-2-0051. The authors would like to thank Jeremy Harbig for providing computational support for extracting the individual probe values from the CEL files and producing artificial CEL files.
- Kuo WP, Jenssen TK, Butte AJ, Ohno-Machado L, Kohane IS: Analysis of matched mRNA measurements from two different microarray technologies. Bioinformatics. 2002, 18 (3): 405-412. 10.1093/bioinformatics/18.3.405.PubMedView ArticleGoogle Scholar
- Li J, Pankratz M, Johnson JA: Differential gene expression patterns revealed by oligonucleotide versus long cDNA arrays. Toxicol Sci. 2002, 69 (2): 383-390. 10.1093/toxsci/69.2.383.PubMedView ArticleGoogle Scholar
- Sorlie T, Tibshirani R, Parker J, Hastie T, Marron JS, Nobel A, Deng S, Johnsen H, Pesich R, Geisler S, Demeter J, Perou CM, Lonning PE, Brown PO, Borresen-Dale AL, Botstein D: Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A. 2003, 100 (14): 8418-8423. 10.1073/pnas.0932692100.PubMedPubMed CentralView ArticleGoogle Scholar
- Culhane AC, Perriere G, Higgins DG: Cross-platform comparison and visualisation of gene expression data using co-inertia analysis. BMC Bioinformatics. 2003, 4: 59-10.1186/1471-2105-4-59.PubMedPubMed CentralView ArticleGoogle Scholar
- Nimgaonkar A, Sanoudou D, Butte AJ, Haslett JN, Kunkel LM, Beggs AH, Kohane IS: Reproducibility of gene expression across generations of Affymetrix microarrays. BMC Bioinformatics. 2003, 4 (1): 27-10.1186/1471-2105-4-27.PubMedPubMed CentralView ArticleGoogle Scholar
- Bloom G, Yang IV, Boulware D, Kwong KY, Coppola D, Eschrich S, Quackenbush J, Yeatman TJ: Multi-platform, multi-site, microarray-based human tumor classification. Am J Pathol. 2004, 164 (1): 9-16.PubMedPubMed CentralView ArticleGoogle Scholar
- Tan PK, Downey TJ, Spitznagel ELJ, Xu P, Fu D, Dimitrov DS, Lempicki RA, Raaka BM, Cam MC: Evaluation of gene expression measurements from commercial microarray platforms. Nucleic Acids Res. 2003, 31 (19): 5676-5684. 10.1093/nar/gkg763.PubMedPubMed CentralView ArticleGoogle Scholar
- Mecham BH, Wetmore DZ, Szallasi Z, Sadovsky Y, Kohane I, Mariani TJ: Increased measurement accuracy for sequence-verified microarray probes. Physiol Genomics. 2004, 18 (3): 308-315. 10.1152/physiolgenomics.00066.2004.PubMedView ArticleGoogle Scholar
- Mecham BH, Klus GT, Strovel J, Augustus M, Byrne D, Bozso P, Wetmore DZ, Mariani TJ, Kohane IS, Szallasi Z: Sequence-matched probes produce increased cross-platform consistency and more reproducible biological results in microarray-based gene expression measurements. Nucleic Acids Res. 2004, 32 (9): e74-10.1093/nar/gnh071.PubMedPubMed CentralView ArticleGoogle Scholar
- Carter SL, Eklund AC, Mecham BH, Kohane IS, Szallasi Z: Redefinition of Affymetrix probe sets by sequence overlap with cDNA microarray probes reduces cross-platform inconsistencies in cancer-associated gene expression measurements. BMC Bioinformatics. 2005, 6 (1): 107-10.1186/1471-2105-6-107.PubMedPubMed CentralView ArticleGoogle Scholar
- Knight J: When the chips are down. Nature. 2001, 410 (6831): 860-861. 10.1038/35073680.PubMedView ArticleGoogle Scholar
- Kothapalli R, Yoder SJ, Mane S, Loughran TPJ: Microarray results: how accurate are they?. BMC Bioinformatics. 2002, 3 (1): 22-10.1186/1471-2105-3-22.PubMedPubMed CentralView ArticleGoogle Scholar
- Halgren RG, Fielden MR, Fong CJ, Zacharewski TR: Assessment of clone identity and sequence fidelity for 1189 IMAGE cDNA clones. Nucleic Acids Res. 2001, 29 (2): 582-588. 10.1093/nar/29.2.582.PubMedPubMed CentralView ArticleGoogle Scholar
- Handley D, Serban N, Peters D, O'Doherty R, Field M, Wasserman L, Spirtes P, Scheines R, Glymour C: Evidence of systematic expressed sequence tag IMAGE clone cross-hybridization on cDNA microarrays. Genomics. 2004, 83 (6): 1169-1175. 10.1016/j.ygeno.2003.12.010.PubMedView ArticleGoogle Scholar
- Gautier L, Moller M, Friis-Hansen L, Knudsen S: Alternative mapping of probes to genes for Affymetrix chips. BMC Bioinformatics. 2004, 5: 111-10.1186/1471-2105-5-111.PubMedPubMed CentralView ArticleGoogle Scholar
- Harbig J, Sprinkle R, Enkemann SA: A sequence-based identification of the genes detected by probesets on the Affymetrix U133 plus 2.0 array. Nucleic Acids Res. 2005, 33 (3): e31-10.1093/nar/gni027.PubMedPubMed CentralView ArticleGoogle Scholar
- Stec J, Wang J, Coombes K, Ayers M, Hoersch S, Gold DL, Ross JS, Hess KR, Tirrell S, Linette G, Hortobagyi GN, Fraser Symmans W, Pusztai L: Comparison of the predictive accuracy of DNA array-based multigene classifiers across cDNA arrays and Affymetrix GeneChips. J Mol Diagn. 2005, 7 (3): 357-367.PubMedPubMed CentralView ArticleGoogle Scholar
- Eschrich S, Yang I, Bloom G, Kwong KY, Boulware D, Cantor A, Coppola D, Kruhoffer M, Aaltonen L, Orntoft TF, Quackenbush J, Yeatman TJ: Molecular staging for survival prediction of colorectal cancer patients. J Clin Oncol. 2005, 23 (15): 3526-3535. 10.1200/JCO.2005.00.695.PubMedView ArticleGoogle Scholar
- Hwang KB, Kong SW, Greenberg SA, Park PJ: Combining gene expression data from different generations of oligonucleotide arrays. BMC Bioinformatics. 2004, 5: 159-10.1186/1471-2105-5-159.PubMedPubMed CentralView ArticleGoogle Scholar
- Woo Y, Affourtit J, Daigle S, Viale A, Johnson K, Naggert J, Churchill G: A comparison of cDNA, oligonucleotide, and Affymetrix GeneChip gene expression microarray platforms. J Biomol Tech. 2004, 15 (4): 276-284.PubMedPubMed CentralGoogle Scholar
- Yauk CL, Berndt ML, Williams A, Douglas GR: Comprehensive comparison of six microarray technologies. Nucleic Acids Res. 2004, 32 (15): e124-10.1093/nar/gnh123.PubMedPubMed CentralView ArticleGoogle Scholar
- Li C, Wong WH: Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc Natl Acad Sci U S A. 2001, 98 (1): 31-36. 10.1073/pnas.011404098.PubMedPubMed CentralView ArticleGoogle Scholar
- Wang Y, Lu J, Lee R, Gu Z, Clarke R: Iterative normalization of cDNA microarray data. IEEE Trans Inf Technol Biomed. 2002, 6 (1): 29-37. 10.1109/4233.992159.PubMedView ArticleGoogle Scholar
- Choe SE, Boutros M, Michelson AM, Church GM, Halfon MS: Preferred analysis methods for Affymetrix GeneChips revealed by a wholly defined control dataset. Genome Biol. 2005, 6 (2): R16-10.1186/gb-2005-6-2-r16.PubMedPubMed CentralView ArticleGoogle Scholar
- Ramdas L, Coombes KR, Baggerly K, Abruzzo L, Highsmith WE, Krogmann T, Hamilton SR, Zhang W: Sources of nonlinearity in cDNA microarray expression measurements. Genome Biol. 2001, 2 (11): RESEARCH0047-10.1186/gb-2001-2-11-research0047.PubMedPubMed CentralView ArticleGoogle Scholar
- Bakay M, Chen YW, Borup R, Zhao P, Nagaraju K, Hoffman EP: Sources of variability and effect of experimental approach on expression profiling data interpretation. BMC Bioinformatics. 2002, 3 (1): 4-10.1186/1471-2105-3-4.PubMedPubMed CentralView ArticleGoogle Scholar
- Churchill GA: Fundamentals of experimental design for cDNA microarrays. Nat Genet. 2002, 32 Suppl: 490-495. 10.1038/ng1031.PubMedView ArticleGoogle Scholar
- Dobbin KK, Beer DG, Meyerson M, Yeatman TJ, Gerald WL, Jacobson JW, Conley B, Buetow KH, Heiskanen M, Simon RM, Minna JD, Girard L, Misek DE, Taylor JM, Hanash S, Naoki K, Hayes DN, Ladd-Acosta C, Enkemann SA, Viale A, Giordano TJ: Interlaboratory comparability study of cancer gene expression analysis using oligonucleotide microarrays. Clin Cancer Res. 2005, 11 (2 Pt 1): 565-572.PubMedGoogle Scholar
- Irizarry RA, Warren D, Spencer F, Kim IF, Biswal S, Frank BC, Gabrielson E, Garcia JG, Geoghegan J, Germino G, Griffin C, Hilmer SC, Hoffman E, Jedlicka AE, Kawasaki E, Martinez-Murillo F, Morsberger L, Lee H, Petersen D, Quackenbush J, Scott A, Wilson M, Yang Y, Ye SQ, Yu W: Multiple-laboratory comparison of microarray platforms. Nat Methods. 2005, 2 (5): 345-350. 10.1038/nmeth756.PubMedView ArticleGoogle Scholar
- Bammler T, Beyer RP, Bhattacharya S, Boorman GA, Boyles A, Bradford BU, Bumgarner RE, Bushel PR, Chaturvedi K, Choi D, Cunningham ML, Deng S, Dressman HK, Fannin RD, Farin FM, Freedman JH, Fry RC, Harper A, Humble MC, Hurban P, Kavanagh TJ, Kaufmann WK, Kerr KF, Jing L, Lapidus JA, Lasarev MR, Li J, Li YJ, Lobenhofer EK, Lu X, Malek RL, Milton S, Nagalla SR, O'Malley J P, Palmer VS, Pattee P, Paules RS, Perou CM, Phillips K, Qin LX, Qiu Y, Quigley SD, Rodland M, Rusyn I, Samson LD, Schwartz DA, Shi Y, Shin JL, Sieber SO, Slifer S, Speer MC, Spencer PS, Sproles DI, Swenberg JA, Suk WA, Sullivan RC, Tian R, Tennant RW, Todd SA, Tucker CJ, Van Houten B, Weis BK, Xuan S, Zarbl H: Standardizing global gene expression analysis between laboratories and across platforms. Nat Methods. 2005, 2 (5): 351-356. 10.1038/nmeth754.PubMedView ArticleGoogle Scholar
- Edgar R, Domrachev M, Lash AE: Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002, 30 (1): 207-210. 10.1093/nar/30.1.207.PubMedPubMed CentralView ArticleGoogle Scholar
- Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, Barrette T, Pandey A, Chinnaiyan AM: ONCOMINE: a cancer microarray database and integrated data-mining platform. Neoplasia. 2004, 6 (1): 1-6.PubMedPubMed CentralView ArticleGoogle Scholar
- Shen D, He J, Chang HR: In silico identification of breast cancer genes by combined multiple high throughput analyses. Int J Mol Med. 2005, 15 (2): 205-212.PubMedGoogle Scholar
- Barnes M, Freudenberg J, Thompson S, Aronow B, Pavlidis P: Experimental comparison and cross-validation of the Affymetrix and Illumina gene expression analysis platforms. Nucleic Acids Res. 2005, 33 (18): 5914-5923. 10.1093/nar/gki890.PubMedPubMed CentralView ArticleGoogle Scholar
- Warrington JA, Nair A, Mahadevappa M, Tsyganskaya M: Comparison of human adult and fetal expression and identification of 535 housekeeping/maintenance genes. Physiol Genomics. 2000, 2 (3): 143-147.PubMedGoogle Scholar
- The Gene Expression Database Portal. [http://gedp.nci.nih.gov/]
- Gene Expression Omnibus. [http://www.ncbi.nlm.nih.gov/projects/geo/]
- Affymetrix technical support data. [http://www.affymetrix.com/support/technical/sample_data/datasets.affx]
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.