笔记详情
标题
RNA Quality and Microarray Analysis
内容
RNA Quality and Microarray Analysis
RNA from primary lung cancers were obtained as part of collaborations with William Gerald at Memorial Sloan-Kettering Cancer Center (New York dataset) and Chi-Leung Lam and Maria Wong at the University of Hong Kong. All samples were collected with appropriate consent and internal review board approval. Cell line RNA was extracted from cell lines maintained in the Minna laboratory at UT Southwestern Medical Center at Dallas as described above.
The quality of total RNA for all samples was analyzed by formaldehyde gel and/or by capillary electrophoresis on the Experion System (Bio-Rad). Total RNA was labeled and amplified by our genomics core facility, according to manufacturer's instructions (Affymetrix [http://www.affymetrix.com]). cRNA was reanalyzed after labeling to ensure optimal amplification for most of the samples.
cRNA was hybridized to U133 Plus 2.0 (~47,000 transcripts) or U133A (~18,400 transcripts) (Affymetrix), and scanned by our microarray core facility (http://microarray.swmed.edu). Expression analysis of microarray data was performed using several algorithms: Robust Multichip Averaging (RMA) [31,32], Microarray Analysis Suite 5.0 (Affymetrix), MATRIX 1.29 (an array analysis program written by GL [unpublished data]; see below), NIH-DAVID [33], Cluster, and TreeView [34].
After scanning, arrays were checked for quality using GCOS (Gene Chip Operating Software) from Affymetrix and then normalized using either RMA or MATRIX 1.29. For log ratio calculations using MAS5 normalization (MATRIX 1.29), the only requirement was that the numerator be present (Affymetrix p-value < 0.065). Data were then logged and renormalized. For RMA normalization, all data were compiled using RMA Express, or RMA through R or BRBArrayTools.
MATRIX (MicroArray TRansformation In eXcel) is a Microsoft Visual Basic program that allows import of multiple CHP files (saved as text file format) from Affymetrix MicroArray Suite 5.0 into an Excel spreadsheet where median normalization, comparison of arrays using log ratios and t-tests, color display, and hierarchical clustering can be performed. Specifically, expression signals are first log2-transformed and color coded such that higher signals are displayed as darker (blue) colors. Absent (high detection p-value) signals are optionally coded separately on a gray scale. For comparison of samples or classes of samples, log2 ratios (i.e., difference of log2-transformed signals) are calculated. If samples are compared, the stronger signals must have a present call (detection p-value < 0.05). If classes of samples are compared (as log ratios of the means), the median of the detection p-values for the class with the highest mean expression value must be less than 0.05. Two-sample t-tests are further calculated to filter out univariate non-significant differential expression. Hierarchical clustering was performed using average linkage with a Pearson correlation metric. All analyses are performed using extensive gene annotation and all probes are BLAST-verified. MATRIX has not been released, as it is still under development. While this program was used extensively in these studies, all analyses were reproduced using publicly available software. Please contact Luc Girard (Luc.Girard@utsouthwestern.edu ) for further details. 点击翻译
RNA from primary lung cancers were obtained as part of collaborations with William Gerald at Memorial Sloan-Kettering Cancer Center (New York dataset) and Chi-Leung Lam and Maria Wong at the University of Hong Kong. All samples were collected with appropriate consent and internal review board approval. Cell line RNA was extracted from cell lines maintained in the Minna laboratory at UT Southwestern Medical Center at Dallas as described above.
The quality of total RNA for all samples was analyzed by formaldehyde gel and/or by capillary electrophoresis on the Experion System (Bio-Rad). Total RNA was labeled and amplified by our genomics core facility, according to manufacturer's instructions (Affymetrix [http://www.affymetrix.com]). cRNA was reanalyzed after labeling to ensure optimal amplification for most of the samples.
cRNA was hybridized to U133 Plus 2.0 (~47,000 transcripts) or U133A (~18,400 transcripts) (Affymetrix), and scanned by our microarray core facility (http://microarray.swmed.edu). Expression analysis of microarray data was performed using several algorithms: Robust Multichip Averaging (RMA) [31,32], Microarray Analysis Suite 5.0 (Affymetrix), MATRIX 1.29 (an array analysis program written by GL [unpublished data]; see below), NIH-DAVID [33], Cluster, and TreeView [34].
After scanning, arrays were checked for quality using GCOS (Gene Chip Operating Software) from Affymetrix and then normalized using either RMA or MATRIX 1.29. For log ratio calculations using MAS5 normalization (MATRIX 1.29), the only requirement was that the numerator be present (Affymetrix p-value < 0.065). Data were then logged and renormalized. For RMA normalization, all data were compiled using RMA Express, or RMA through R or BRBArrayTools.
MATRIX (MicroArray TRansformation In eXcel) is a Microsoft Visual Basic program that allows import of multiple CHP files (saved as text file format) from Affymetrix MicroArray Suite 5.0 into an Excel spreadsheet where median normalization, comparison of arrays using log ratios and t-tests, color display, and hierarchical clustering can be performed. Specifically, expression signals are first log2-transformed and color coded such that higher signals are displayed as darker (blue) colors. Absent (high detection p-value) signals are optionally coded separately on a gray scale. For comparison of samples or classes of samples, log2 ratios (i.e., difference of log2-transformed signals) are calculated. If samples are compared, the stronger signals must have a present call (detection p-value < 0.05). If classes of samples are compared (as log ratios of the means), the median of the detection p-values for the class with the highest mean expression value must be less than 0.05. Two-sample t-tests are further calculated to filter out univariate non-significant differential expression. Hierarchical clustering was performed using average linkage with a Pearson correlation metric. All analyses are performed using extensive gene annotation and all probes are BLAST-verified. MATRIX has not been released, as it is still under development. While this program was used extensively in these studies, all analyses were reproduced using publicly available software. Please contact Luc Girard (Luc.Girard@utsouthwestern.edu ) for further details. 点击翻译
来源
PLoS Med. 2006 December; 3(12): e486.
类别
