244
Opinion
TRENDS in Biochemical Sciences
Vol.28 No.5 May 2003
Standardization of protocols in cDNA microarray analysis Vladimı´r Benesˇ and Martina Muckenthaler European Molecular Biology Laboratory, Meyerhofstrasse 1 D-69117 Heidelberg, Germany
Systematic variations can occur at various steps of a cDNA microarray experiment and affect the measurement of gene expression levels. Accepted standards integrated into every cDNA microarray analysis can assess these variabilities and aid the interpretation of cDNA microarray experiments from different sources. A universally applicable approach to evaluate parameters such as input and output ratios, signal linearity, hybridization specificity and consistency across an array, as well as normalization strategies, is the utilization of exogenous control genes as spike-in and negative controls. We suggest that the use of such control sets, together with a sufficient number of experimental repeats, in-depth statistical analysis and thorough data validation should be made mandatory for the publication of cDNA microarray data. Genome-wide expression profiling by cDNA microarrays is a powerful technique for monitoring changes in gene expression on a global scale in different organisms, tissues and temporal or spatial arrangements [1– 3]. Specialized microarrays focus on a selection of genes that are involved in specific biochemical pathways or diseases. Microarray co-hybridization assays are complex, multistep procedures involving array fabrication, fluorescentprobe labelling, hybridization and data analysis (Fig. 1). Laboratories around the world have developed a variety of protocols for each of these steps [4 –6]. It is widely acknowledged that this range of available methodology results in data variability between laboratories. However, in our opinion the different technologies available for microarray analysis are not problematic per se as long as the research community is able to straightforwardly assess the validity and quality of the data published in journals and microarray databases. Standardization of microarray technology with respect to the implementation of quality controls, such as exogenous negative and spikein controls would be a desirable prerequisite for the submission of microarray data to repositories and peerreviewed journals. Such controls should enable the assessment of input and output ratios, signal linearity, hybridization specificity and consistency across an array, as well as the assessment of normalization strategies. Furthermore, as it is common practice for most biological experiments, gene-expression data obtained by microarray technology should be controlled by a sufficiently Corresponding author: Vladimı´r Benesˇ (
[email protected]).
large number of biological repeat experiments as well as by independent validation of the results. Strict implementation of these measures would facilitate the cross comparison of data between laboratories. Several journals have already followed the Microarray Gene Expression Data (MGED; http://www.mged.org) call of the Society for standardization of microarray data publications [7] in that, upon submission, datasets must be made publicly available in one of the main public repositories [e.g. ArrayExpress (http://www.ebi.ac.uk/microarray/ArrayExpress) and NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/ geo)]. Data submission must be accompanied by a detailed description of the experimental procedures (with all its variables), as outlined by the MGED group (http://www. mged.org/Workgroups/MIAME/miame_checklist.html). We believe that the value of cDNA microarray data will be significantly increased for the research community if standardized quality controls, such as exogenous negative and spike-in controls, are included into otherwise tightly controlled experimental set-up. Variability in cDNA microarray data A microarray platform that is commonly used in genomewide gene expression profiling is the spotted cDNA microarray. In this system, PCR products amplified from cDNA libraries or genomic DNA are arrayed on a coated glass slide or nitrocellulose membrane using a precision robot. A common experimental design of cDNA microarray experiments is the comparison of a query or ‘experimental sample’ with a reference RNA sample. The reference sample should represent all the genes that are expressed in the tissues or cells to be analyzed. When a common reference sample is used in a series of experiments, it enables comparison of all experimental samples with a common denominator. Therefore, the data from multiple arrays are normalized into an integrated body of data. cDNAs are synthesized from the reference and experimental sample with fluorescent dyes (often Cy3 and Cy5) and are subsequently co-hybridized to the array. Following the removal of unbound material, the array is scanned with a laser scanner and the resulting image is analyzed to quantitate relative fluorescent intensities between the reference and the experimental sample for each gene that is represented on the microarray. Once these raw data are obtained, the local background (around each spot) is usually subtracted and the data are normalized. To perform a successful and high quality microarray experiment it is important to consider and control
http://tibs.trends.com 0968-0004/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0968-0004(03)00068-9
Opinion
TRENDS in Biochemical Sciences
Vol.28 No.5 May 2003
245
Box 1. cDNA microarray experiments – points to consider (a) Here, we list the points to consider during a cDNA microarray experiment starting from gene, to spot, to insight:
3
Genome-wide expression profiling vs specialized microarrays Selection and sequence verification of cDNA samples
2 Establishment of the technological microarray platform † Synthesis and purification of gene fragments † Surface chemistry † Spotting conditions † Array design
Background cut-off 1
0
0
1
2
Preparation of the experimental and the reference sample † High quality RNA extraction from cultured cells, tissues, patient biopsies, laser capture microdissection taking into consideration that the experimental and the reference samples must be treated identically
3
(b)
Choice of methodology for the synthesis of fluorescent-labelled cDNA † Yield of purified total RNA † Accuracy, sensitivity, background noise † Labour intensity and working time † Financial aspects
3
2
Implementation of controls (non-specific background, normalization and ratio)
Background cut-off
1
Number and type of replicates (technical, biological) Data acquisition and evaluation Data normalization (global, intensity-dependent)
0 0
1
2
3 Ti BS
Fig. 1. Spike-in and negative controls contribute to the quality assessment of a cDNA microarray experiment. Shown is a scatter plot representation of a microarray experiment performed on a specialized cDNA microarray platform [42]. The expression levels of genes in the experimental sample (labelled with Cy5) are represented on the y-axis, and of genes in the control sample (labelled with Cy3) on the x-axis. Both axes are shown in logarithmic scales. Blue spots represent empty spots; green spots represent negative controls, selected from bacterial and Arabidopsis sequences. Red spots show spike-in controls that have been added to equal amounts to the experimental and the reference sample before probe synthesis. Grey spots represent the experimental genes on the microarray. Spots with intensities above the background cut-off value are considered detectably expressed in this experiment. (a) Accurate total RNA quantitation: non-regulated genes and spike-in controls are represented as equally expressed when the experiment is normalized by using either all genes on the microarray or the spike-in controls. (b) Inaccurate RNA quantitation skews the representation of the spike-in controls in relation to equally expressed genes on the microarray. Nevertheless, the spikein controls can still be used to determine the ratio cut-off value when the experiment is normalized by the spike-in controls only, and therefore allow for a quality estimation of the microarray experiment. Comparable scenario as shown in panel B can arise when the majority of genes represented on the array change their expression levels (e.g. when general transcription is influenced by a treatment of cells). In this case, spike-in controls are essential for data evaluation.
numerous variations that can occur at each step (Box 1). Generally, experimental and systematic variations can be distinguished: experimental variability can be controlled by careful experimental design [8] and through a sufficient number of experimental repeats; systematic variations have to be addressed by controls on the array. A possible source for systematic variations can be the irregular deposition of PCR amplified cDNAs on the glass surface by different printing pins (including ‘carry-over’ of the samples between adjacent sample wells caused by inferior washing of the pins) or biases associated with different fluorescent dyes. It has been recognized that fluorescent http://tibs.trends.com
Interpretation of the microarray data (comparison, clustering, selforganizing maps) Independent validation of the data (quantitative reverse transcriptase real-time PCR, Northern blot, in situ hybridization)
dyes such as Cy3 and Cy5 exhibit different quantum yields and are differentially sensitive to photobleaching [9,10]. Depending upon the type of the activated surface, these dyes also show varying background levels (E. Furlong, pers. commun.). Although this phenomenon has not been thoroughly studied, it has been indicated that the direct incorporation of Cy3 and Cy5 modified-nucleotide analogues into the cDNA might introduce sequence-specific artefacts [11,12]. This is likely to be caused by the variable and differing rates by which these bulky nucleotide analogues are incorporated into the synthesized DNA molecules by reverse transcriptase, and might be further exacerbated by irregularities of the reverse transcriptase processivity, which is dependent upon the base composition of a template. Such experimentally introduced variability in signal intensities can have severe effects on gene expression values and can result in misidentification of differentially expressed genes. An efficient way to recognize this type of experimental artefact, provided that the same labelling protocol is followed throughout a given set of experiments, is to exchange the dyes between the experiment and the reference sample (a so-called ‘dye-swap’ or ‘flip-dye’ experiment) [8,13,14]. In addition to different dye usage, the choice of method for fluorescent cDNA labelling can significantly influence the expression pattern obtained
246
Opinion
TRENDS in Biochemical Sciences
from a microarray experiment, and care should be taken to select a protocol that is appropriate for the experimental system [15]. Moreover, instrumental variations in the spotter specifications (e.g. spot size and precision of DNA deposition) and/or the scanner specifications (e.g. detection limit and dynamic range) can affect the data and, thus, reduce their comparability between different laboratories [9,16]. A balancing act – normalization A process called normalization [17] can balance many of these systematic variations. In recent years, several normalization techniques have been developed [18,19]; the underlying principle behind the different methods requires the identification of genes that are not affected by experimental conditions and, thus, should show a ratio of the arithmetic mean equal to one between the reference and the experimental sample. Many researchers have used a global normalization strategy in their experiments, which works on the assumption that most genes in a microarray experiment do not change their expression, and that only a minor fraction of genes is differentially expressed. This strategy has been further refined by application of intensity-dependent algorithms for local normalization [18,20]. Although these ways to normalize cDNA microarray data might be adequate under certain circumstances, we believe that exogenous spike-in controls are valuable tools for improvement of the normalization process. Complex techniques require complex controls In its simplest form, housekeeping genes have served as controls for genes whose expression is expected to be constant across samples. However, recent studies have demonstrated that genes that do not change in their expression levels in response to a variety of experimental conditions, do not exist. Even the expression of actin or glyceraldehyde 3-phosphate dehydrogenase (GAPDH) – genes that are extensively used to normalize gene expression in the conventional northern blot experiments – can vary considerably under certain experimental conditions (M. Muckenthaler, unpublished observation). In addition, many housekeeping genes might not be representative of all intensity values on the array. Another approach to control variability in microarray experiments is to spot genomic DNA in multiple dilutions on the array and use the signals obtained after hybridization for normalization purposes. A universally applicable approach to normalize gene expression data and to assess important parameters in cDNA microarray experiments (i.e. input and output ratios, signal linearity, hybridization specificity and consistency across an array) is the application of exogenous control genes. External controls are chosen to avoid cross-hybridization with mRNAs from the organism studied, but should be similar in general characteristics, such as GC content and length, and should contain a poly(A) tail [21]. These control genes are often selected from bacterial or plant biochemical pathways that are not represented in metazoan organisms and, thus, can be used on vertebrate or insect arrays. The exogenous controls are included in every sub-grid of the cDNA microarray; this design aids the recognition of variations within the slide http://tibs.trends.com
Vol.28 No.5 May 2003
surface (e.g. unevenness and coating), the performance of the printing tips and the assessment of local hybridization artefacts. The samples (which are to be hybridized to the array) are then spiked with the exogenous control RNAs that have been obtained by in vitro transcription and that contain engineered poly(A) tails. External controls can be used as negative controls if no corresponding mRNA is present in the RNA samples to be analyzed; negative controls help to determine the noise of a microarray experiment. As a rule-of-thumb, genes are considered detectable and expressed if the signal intensity exceeds one-to-two standard deviations of the average signal intensities of all negative controls. Such a cut-off value removes data from spots with low intensity – which have high variance – and improves slide-to-slide comparisons. External controls are also used as positive or spike-in controls; in this case, exogenous RNAs are added to the reference and the experimental samples in predetermined concentrations before the synthesis of fluorescent-labelled cDNAs. Spike-in controls need to be titrated to cover the entire range of signal intensities obtained in a microarray experiment to be representative for all detectable genes. Spike-in RNAs added in equal amounts to the reference and experimental sample can serve as normalization controls. This process requires consistent and tightly controlled isolation protocols and accurate quantitation of the initial amount of total RNA used for each sample, including controls, because inaccurate total RNA measurement might skew the relative representation of the actual RNA sample in respect to the spike-in RNAs (Fig. 1). The use of exogenous spike-in controls is of particular importance for the normalization of microarray data from specialized microarrays that contain only a reduced number of genes or when experimental conditions are analyzed of which a large proportion of genes represented on the array is expected to change its expression. In addition, exogenous spike-in controls can be used to evaluate an expression ratio observed in a microarray experiment. This might not be relevant for studies focusing mainly on large expression changes (e.g. diseases with large chromosomal rearrangements). However, in many publications, the ratio cut-off value for significant differential expression is chosen arbitrarily at twofold (without any experimental validation), a certain reproducibility of microarray data at or above this level was demonstrated previously [22]. Nevertheless, for many biological systems it would be useful to reliably detect a 1.5-fold change in mRNA content because it can be highly significant for the physiology or pathology of organisms. High-quality microarray experiments combined with a larger number of experimental repeats will support the measurement of these small differences in gene expression [23]. Inadequate experimental design or inconsistent application of fluorescent cDNA-labelling techniques will, however, not even support a ratio cut-off value of two [15]. Using spike-in RNAs, we have previously suggested that a cut-off value for differential expression can be defined in the following way: spike-in RNAs added in equal amounts to the reference and the experimental sample are normalized such that the ratio
Opinion
TRENDS in Biochemical Sciences
equals to one [15]; then, the deviation of each individual spike-in control from the normalized value is calculated and used to determine the standard deviation; the cut-off ratio for regulation in each microarray experiment is calculated as two standard deviations above and below one [15]. Intensity-dependent normalization strategies can differentially assign cut-off ratios to low- and highintensity spots on the array. This is important because lowintensity signals on the array, which often correspond to genes with low expression levels, show high variance and inferred results are much more prone to error. For calculating the cut-off ratio, it is important to normalize using the signals obtained from spike-in RNAs only (excluding the signals from the other spots on the array) because errors in total RNA quantitation might result in a skewed representation of the spike-in RNAs. If total RNA quantitation is inaccurate, the sum of spike-in RNAs is usually shifted parallel to the experimental spots in a scatter-plot representation of the data (Fig. 1). Spike-in RNAs added in pre-defined ratios to the experimental and reference sample will further assure that the experimental ratios obtained are significant and help to assess the dynamic range of a microarray experiment. Cut-off values for differential expression (spike ratio cut-off) using spike-in controls and cut-off values for the noise of a microarray experiment (background cut-off) using negative controls enable exclusion of low-quality slides from data evaluation and can provide parameters to assess slide quality in public microarray databases. In principle, if the selected set of external control genes is large enough it can be used for spike-in at different stages of the microarray procedure (e.g. before T7 RNA polymerase amplification of low-starting material, before fluorescent cDNA synthesis and before hybridization), thus, allowing for the quality of many individual steps of the microarray procedure to be monitored. Exogenous control genes should become standard reagents in any microarray laboratory because this method is already applied by industrial providers of microarray technology (e.g. Affymetrix Inc. GeneChipw microarrays). Moreover, exogenous control genes can be obtained from several commercial sources (e.g. Amersham’s ScoreCard and Stratagene’s SpotReport). However, regardless of whether spike-in RNAs prepared in the laboratory or whether commercial products are being used, it is necessary to always keep in mind the issue of batch-tobatch variations, which can make interpretation of the results problematic if not directly misleading. Despite the wealth of information that can be obtained regarding the quality of a microarray experiment when exogenous controls are integrated into microarray analysis, to date, it has not been required to include them into the cDNA microarray experiments as a prerequisite for submission of microarray data for publication. Repeat experiments are needed for robust statistics Exogenous controls are not a substitute for thorough statistical analysis of microarray data to evaluate the expression levels of individual genes. Many diverse methods are available for statistical analysis [24,25], however, the quality of each method is dependent on the http://tibs.trends.com
Vol.28 No.5 May 2003
247
number of ‘spot replicates’ and ‘biological replicates’ integrated into a microarray experiment. Unfortunately, in the literature there have been cases of microarray experiments done without replication [26]. If clone collections containing less than a few thousand genes are used for spotting, each individual clone can be replicated multiple times at different locations of the array. With such a microarray layout, random and systematic measurement errors in the microarray process that might affect the data can be easily assessed. Owing to multiple spot replicates, more reliable information can be gained from a single experiment. Genome-wide arrays are limited by space and, thus, multiple clone replicates can rarely be spotted. As a consequence, the number of hybridization experiments needs to be increased. Regardless of the scale, replicate experiments in general should include multiple ‘biological repeats’ (e.g. RNA extractions from independent tissue samples or cell lines). ‘Biological repeats’ show a higher degree of variability in comparison to mere replicates of one biological sample but are essential for statistical analysis and for the generalization of conclusions [8,9]. Independent validation is indispensable Once statistically significant expression ratios are established it is important to confirm the obtained data using one of the available alternative methods of gene expression measurement. Clearly, differential expression of only a limited, crucial subset of genes can be verified. Currently, the techniques most frequently used are quantitative reverse transcriptase real-time PCR (qRT– PCR), northern blotting, ribonuclease protection assay or in situ hybridization, which is particularly suitable for highly regulated genes [27]. It has been demonstrated that the relative changes in mRNA expression between experimental and reference samples are consistently underestimated on cDNA microarrays. Attempts have been made to correlate expression ratios obtained from microarray experiments and qRT–PCR by applying a mathematical algorithm [28]. Why is independent data confirmation so important? In addition to the various artefacts that can be generated by microarray technology, the cDNA clone collections in use for spotting are often not sequence-verified before PCR amplification. Recent reports have demonstrated that even well maintained clone sets contain between one and five percent of wrongly assigned or contaminated clones [29,30]. Despite all the variability issues associated with cDNA microarray assays, reproducible high-quality data can be obtained from microarray experiments when optimized laboratory protocols are in use [16,31]. In this manuscript we have outlined how exogenous controls in combination with multiple biological repeats and thorough statistical analysis, if implemented for two-dye experiments, would enable assessment of the validity and quality of the data obtained. Which controls should be implemented before publication of microarray data? Provided that the number of biological repeats is sufficient (i.e. number of experiments, and size of the reference and
248
Opinion
TRENDS in Biochemical Sciences
experimental groups) to obtain statistically meaningful datasets, the experiments utilizing two-dye microarray technology should include experimental repeats (from our own experience, a comparison of six reference mice with six experimental mice provides a good starting point to obtain reliable measurement of gene expression) with a switch of dyes between the experimental and the reference samples. To assess input and output ratios, signal linearity, hybridization specificity and consistency across individual slides as well as to facilitate slide-to-slide comparisons, the implementation of exogenous controls is recommended. In addition, spike-in controls are useful tools to assess whether the ratio cut-off that has been chosen is supported by the experiment. This is of particular importance in experimental systems that exhibit only reduced differential gene regulation, for example, when studying neurological questions [32]. Ideally, a common spike-in control and reference set for each species used in every microarray experiment would limit variability problems [33] and facilitate normalization, evaluation and comparison of expression profiles obtained from independent measurements. A series of recently published articles [23,34 – 37] addressing the issue of a common reference set and its construction indicates that an effort to resolve the problem of comparability of microarray results is to be seriously undertaken. Concluding remarks Microarray technology will undoubtedly become one of the most intensively utilized methodologies in coming years. The results obtained will be starting points for many dazzling insights into biological questions. It is still early days for this methodology and, thus, standards for submitting primary data to databases and peer-reviewed journals for publication should be implemented now by databases curators and editors. An open letter of the Microarray Gene Expression Data (MGED) society to the scientific journals [38] has found its resonance in editorial boards of several journals [39– 41], which have already undertaken first standardization attempts for manuscripts containing microarray data, and their numbers continuously rising. For publication in the journals of the Nature Publishing Group, microarray data must be submitted to one of the existing microarray databases, either to ArrayExpress in EBI or to NCBI Gene Expression Omnibus. In addition, all experimental procedures (with all their variables) used to obtain the data must be reported according to the guidelines proposed by the MGED society [7]. A further step forward in the standardization of microarray technology would be the mandatory inclusion of positive and negative controls in every microarray experiment in addition to the implementation of thorough statistical analysis and validation of the data. Acknowledgements We thank our colleagues Eileen Furlong, Bruno Galy, Belen Minana, David Ibberson and Jos de Graaf for critical reading of the article and helpful discussion. We also appreciate the constructive comments of the referees. http://tibs.trends.com
Vol.28 No.5 May 2003
References 1 Mills, J.C. et al. (2001) DNA microarrays and beyond: completing the journey from tissue to cell. Nat. Cell Biol. 3, E175 – E178 2 Liotta, L. and Petricoin, E. (2000) Molecular profiling of human cancer. Nat. Rev. Genet. 1, 48 – 56 3 Marcotte, E.R. et al. (2001) DNA microarrays in neuropsychopharmacology. Trends Pharmacol. Sci. 22, 426 – 436 4 Hedge, P. et al. (2000) A concise guide to cDNA microarray analysis. Biotechniques 29, 548– 562 5 Klebes, A. et al. (2002) Expression profiling of Drosophila imaginal discs. Genome Biol. 3, R0038 6 Gasch, A.P. (2002) Yeast genomic expression studies using DNA microarrays. Methods Enzymol. 350, 393 – 414 7 Brazma, A. et al. (2001) Minimum information about a microarray experiment (MIAME)-toward standards for microarray data. Nat. Genet. 29, 365 – 371 8 Yang, Y.H. and Speed, T. (2002) Design issues for cDNA microarray experiments. Nat. Rev. Genet. 3, 579 – 588 9 Worley, J. et al. (2000) A system approach to fabricating and analysing DNA microarrays. In Microarray Chip Technology (Schena, M., ed.), pp. 65– 86, Eaton Publishing 10 Wildsmith, S.E. et al. (2001) Maximization of signal derived from cDNA microarrays. Biotechniques 30, 202– 208 11 Kerr, M.K. and Churchill, G.A. (2001) Experimental design for gene expression microarrays. Biostatistics 2, 183 – 201 12 Baugh, L.R. et al. (2001) Quantitative analysis of mRNA amplification by in vitro transcription. Nucleic Acids Res. 29, E29 13 Tseng, G.C. et al. (2001) Issues in cDNA microarray analysis: quality filtering, channel normalization, models of variations and assessment of gene effects. Nucleic Acids Res. 29, 2549 – 2557 14 Kerr, M.K. and Churchill, G.A. (2001) Statistical design and the analysis of gene expression microarray data. Genet. Res. 77, 123 – 128 15 Richter, A. et al. (2002) Comparison of fluorescent tag DNA labeling methods used for expression analysis by DNA microarrays. Biotechniques. 33, 620– 628, 630 16 Holloway, A.J. et al. (2002) Options available – from start to finish – for obtaining data from DNA microarrays II. Nat. Genet. 32, 481 – 489 17 Quackenbush, J. (2002) Microarray normalization and transformation. Nat. Genet. 32, 496 – 501 18 Yang, Y.H. et al. (2002) Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res. 30, E15 19 Kroll, T.C. and Wo¨lfl, S. (2002) Ranking: a closer look on globalisation methods for normalisation of gene expression arrays. Nucleic Acids Res. 30, E50 20 Cleveland, W.S. (1979) Robust locally weighted regression and smoothing scatterplots. J. Am. Stat. Assoc. 74, 8924– 8936 21 Eickhoff, B. et al. (1999) Normalization of array hybridisation experiments in differential gene expression analysis. Nucleic Acids Res. 27, E33 22 Wodicka, L. et al. (1997) Genome-wide expression monitoring in Saccharomyces cerevisiae. Nat. Biotechnol. 15, 1359 – 1367 23 Yang, I.V. et al. (2002) Within the fold: assessing differential expression measures and reproducibility in microarray assays. Genome Biology 3, 0062.1 – 0062.12 24 Nadon, R. and Shoemaker, J. (2002) Statistical issues with microarrays: processing and analysis. Trends Genet. 18, 265 – 271 25 Pan, W. et al. (2002) How many replicates of arrays are required to detect gene expression changes in microarray experiments? A mixture model approach. Genome Biol. 3, 0022.1 – 0022.10 26 Lee, T.M-L. et al. (2000) Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridisations. Proc. Natl. Acad. Sci. U. S. A. 97, 9834– 9839 27 Chuaqui, R.F. et al. (2002) Post-analysis follow-up and validation of microarray experiments. Nat. Genet. 32, 509 – 514 28 Yuen, T. et al. (2002) Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays. Nucleic Acids Res. 30, E48 29 Halgren, R.G. et al. (2001) Assessment of clone identity and sequence fidelity for 1189 IMAGE cDNA clones. Nucleic Acids Res. 29, 582 – 588 30 Taylor, E. et al. (2001) Sequence verification as quality-control step for production of cDNA microarrays. Biotechniques 31, 62 – 65 31 Bowtell, D. and Sambrook, J. (2003) In DNA Microarray: A Molecular
Opinion
32 33 34
35 36
TRENDS in Biochemical Sciences
Cloning Manual (Bowtell, D. and Sambrook, J., eds) Cold Spring Harbor Laboratory Press Lockhart, D.J. and Barlow, C. (2001) Expressing what’s on your mind: DNA arrays and the brain. Nat. Rev. Neurosci. 2, 63 – 68 Spruill, S.E. et al. (2002) Assessing sources of variability in microarray gene expression data. Biotechniques 33, 916 – 923 Kim, H. et al. (2002) Use of RNA and genomic DNA references for inferred comparisons in DNA microarray analyses. Biotechniques 33, 924 – 930 Sterrenburg, E. et al. (2002) A common reference for cDNA microarray hybridisations. Nucleic Acids Res. 30, E116 Tallat, A.M. et al. (2002) Genomic DNA standards for gene expression profiling in Mycobacterium tuberculosis. Nucleic Acids Res. 30, E104
Vol.28 No.5 May 2003
37 Weil, M.R. et al. (2002) Toward a universal standard: comparing two methods for standardizing spotted microarray data. Biotechniques 32, 1310– 1314 38 Nerup, J. (2002) An open letter to the scientific journals. Bioinformatics 18, 1409 39 No author, (2002) Microarray standards at last. Nature 419, 323 40 No author, (2002) Coming to terms with microarrays. Nat. Genet. 32, 333– 334 41 Firestein, G.S. and Pisetsky, D.S. (2002) DNA microarrays: boundless technology or bound by technology? Guidelines for studies using microarray technology. Arthritis Rheum. 46, 859 – 861 42 Muckenthaler, M. et al. (2002) Relationships and distinctions in iron regulatory networks responding to interrelated signals. Blood DOI 10.1182/blood-2002-07-2140
Have you seen our ‘Protein Synthesis’ series, which began in the October 2002 issue? Articles published to date: The protein synthesis world (Editorial) Schimmel, P. (2002) Trends Biochem. Sci. 27, 505 Nuclear export of ribosomal subunits Johnson, A.W., Lund, E. and Dahlberg, J. (2002) Trends Biochem. Sci. 27, 580585 Making sense of mimic in translation termination Nakamura, Y. and Ito, K. (2003) Trends Biochem. Sci. 28, 99-105 Translation initiation and viral tricks Schneider, R.J. and Mohr, I. (2003) Trends Biochem. Sci. 28, 130-136 Regulation of mRNA translation by 5Aˆ and 3Aˆ UTR-binding factors Wilkie G.S., Dickson K.S. and Gray N.K. Trends Biochem. Sci. DOI:10.1016/S0968-0004(03)00051-3 How introns influence and enhance eukaryotic gene expression LeHir H., Nott, A. and Moore M.J. Trends Biochem. Sci. DOI:10.1016/S0968-0004(03)00052-5 Other articles planned for the series: Insights into the decoding mechanism from recent ribosome structures Ogle J.M., Carter A.P. and Ramakrishnan V. Trends Biochem. Sci. (This issue) Protein Synthesis/UPR Kaufman, R
http://tibs.trends.com
249