Limits...
Pre-processing and differential expression analysis of Agilent microRNA arrays using the AgiMicroRna Bioconductor library.

López-Romero P - BMC Genomics (2011)

Bottom Line: AgiMicroRna uses the linear model features implemented in the limma package to assess the differential expression between different experimental conditions and provides links to the miRBase for those microRNAs that have been declared as significant in the statistical analysis.AgiMicroRna is a rational collection of Bioconductor functions that have been wrapped into specific functions in order to ease and systematize the pre-processing and statistical analysis of Agilent microRNA data.The development of this package contributes to the Bioconductor project filling the gap in microRNA array data analysis.

View Article: PubMed Central - HTML - PubMed

Affiliation: Epidemiology, Atherothrombosis and Imaging Department, Centro Nacional de Investigaciones Cardiovasculares Carlos III, Melchor Fernández Almagro 3, E-28029 Madrid, Spain. plopez@cnic.es

ABSTRACT

Background: The main research tool for identifying microRNAs involved in specific cellular processes is gene expression profiling using microarray technology. Agilent is one of the major producers of microRNA arrays, and microarray data are commonly analyzed by using R and the functions and packages collected in the Bioconductor project. However, an analytical package that integrates the specific characteristics of microRNA Agilent arrays has been lacking.

Results: This report presents the new bioinformatic tool AgiMicroRNA for the pre-processing and differential expression analysis of Agilent microRNA array data. The software is implemented in the open-source statistical scripting language R and is integrated in the Bioconductor project (http://www.bioconductor.org) under the GPL license. For the pre-processing of the microRNA signal, AgiMicroRNA incorporates the robust multiarray average algorithm, a method that produces a summary measure of the microRNA expression using a linear model that takes into account the probe affinity effect. To obtain a normalized microRNA signal useful for the statistical analysis, AgiMicroRna offers the possibility of employing either the processed signal estimated by the robust multiarray average algorithm or the processed signal produced by the Agilent image analysis software. The AgiMicroRNA package also incorporates different graphical utilities to assess the quality of the data. AgiMicroRna uses the linear model features implemented in the limma package to assess the differential expression between different experimental conditions and provides links to the miRBase for those microRNAs that have been declared as significant in the statistical analysis.

Conclusions: AgiMicroRna is a rational collection of Bioconductor functions that have been wrapped into specific functions in order to ease and systematize the pre-processing and statistical analysis of Agilent microRNA data. The development of this package contributes to the Bioconductor project filling the gap in microRNA array data analysis.

Show MeSH

Related in: MedlinePlus

Pre-processing steps. AgiMicroRna includes two distinct pre-processing protocols for transforming the raw probe level data into the processed data that contain the summarized and normalized microRNA gene signals. The first protocol comprises the following steps: 1) acquisition of the TGS processed by AFE, and 2) normalization between arrays by scale or quantile methods. The second option uses the RMA algorithm via the following steps: 1) the raw mean signal can be background corrected (optional, see recommendations in the text); 2) the signal is normalized between arrays by quantile normalization, and 3) the probe level data is summarized into a single microRNA measure. Selected microRNAs can be filtered out according to the Flags assigned to each probe by the Agilent Extraction Software.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3037903&req=5

Figure 1: Pre-processing steps. AgiMicroRna includes two distinct pre-processing protocols for transforming the raw probe level data into the processed data that contain the summarized and normalized microRNA gene signals. The first protocol comprises the following steps: 1) acquisition of the TGS processed by AFE, and 2) normalization between arrays by scale or quantile methods. The second option uses the RMA algorithm via the following steps: 1) the raw mean signal can be background corrected (optional, see recommendations in the text); 2) the signal is normalized between arrays by quantile normalization, and 3) the probe level data is summarized into a single microRNA measure. Selected microRNAs can be filtered out according to the Flags assigned to each probe by the Agilent Extraction Software.

Mentions: Agilent microRNA microarrays interrogate each microRNA gene with different probe sets. To make statistical inferences, a summary expression measure for each microRNA, possibly normalized between arrays, is needed. AgiMicroRna includes two alternative strategies for pre-processing the raw probe level data to yield a summarized and normalized microRNA gene signal (Figure 1). The first approach is based simply on normalization of the AFE-processed TGS between arrays. The AFE-processed TGS is a background-subtracted signal and hence might contain negative values. Therefore to obtain positive values before log transformation, AgiMicroRna either adds a small positive constant to all TGS signals or sets all negative TGS values to 0.5. This TGS signal can be used to make statistical inferences after a normalization step, either using the quantile or scale methods integrated in AgiMicroRna or any normalization method implemented in another Bioconductor package. The other approach incorporated in AgiMicroRna yields a summary microRNA gene measure using the RMA algorithm [10,12]. In the RMA algorithm implemented in AgiMicroRna, the signal can be first background corrected by fitting a normal + exponential convolution model to the vector of observed intensities [18]. When using the RMA algorithm, it might be a better option to omit background correction [12]. Whether or not the signal has been background corrected, the arrays are then normalized by quantile, and finally an estimate of the microRNA gene signal is obtained by fitting a linear model to the log2 transformed probe intensities. This model produces an estimate of the microRNA gene signal corrected for the probe effect.


Pre-processing and differential expression analysis of Agilent microRNA arrays using the AgiMicroRna Bioconductor library.

López-Romero P - BMC Genomics (2011)

Pre-processing steps. AgiMicroRna includes two distinct pre-processing protocols for transforming the raw probe level data into the processed data that contain the summarized and normalized microRNA gene signals. The first protocol comprises the following steps: 1) acquisition of the TGS processed by AFE, and 2) normalization between arrays by scale or quantile methods. The second option uses the RMA algorithm via the following steps: 1) the raw mean signal can be background corrected (optional, see recommendations in the text); 2) the signal is normalized between arrays by quantile normalization, and 3) the probe level data is summarized into a single microRNA measure. Selected microRNAs can be filtered out according to the Flags assigned to each probe by the Agilent Extraction Software.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3037903&req=5

Figure 1: Pre-processing steps. AgiMicroRna includes two distinct pre-processing protocols for transforming the raw probe level data into the processed data that contain the summarized and normalized microRNA gene signals. The first protocol comprises the following steps: 1) acquisition of the TGS processed by AFE, and 2) normalization between arrays by scale or quantile methods. The second option uses the RMA algorithm via the following steps: 1) the raw mean signal can be background corrected (optional, see recommendations in the text); 2) the signal is normalized between arrays by quantile normalization, and 3) the probe level data is summarized into a single microRNA measure. Selected microRNAs can be filtered out according to the Flags assigned to each probe by the Agilent Extraction Software.
Mentions: Agilent microRNA microarrays interrogate each microRNA gene with different probe sets. To make statistical inferences, a summary expression measure for each microRNA, possibly normalized between arrays, is needed. AgiMicroRna includes two alternative strategies for pre-processing the raw probe level data to yield a summarized and normalized microRNA gene signal (Figure 1). The first approach is based simply on normalization of the AFE-processed TGS between arrays. The AFE-processed TGS is a background-subtracted signal and hence might contain negative values. Therefore to obtain positive values before log transformation, AgiMicroRna either adds a small positive constant to all TGS signals or sets all negative TGS values to 0.5. This TGS signal can be used to make statistical inferences after a normalization step, either using the quantile or scale methods integrated in AgiMicroRna or any normalization method implemented in another Bioconductor package. The other approach incorporated in AgiMicroRna yields a summary microRNA gene measure using the RMA algorithm [10,12]. In the RMA algorithm implemented in AgiMicroRna, the signal can be first background corrected by fitting a normal + exponential convolution model to the vector of observed intensities [18]. When using the RMA algorithm, it might be a better option to omit background correction [12]. Whether or not the signal has been background corrected, the arrays are then normalized by quantile, and finally an estimate of the microRNA gene signal is obtained by fitting a linear model to the log2 transformed probe intensities. This model produces an estimate of the microRNA gene signal corrected for the probe effect.

Bottom Line: AgiMicroRna uses the linear model features implemented in the limma package to assess the differential expression between different experimental conditions and provides links to the miRBase for those microRNAs that have been declared as significant in the statistical analysis.AgiMicroRna is a rational collection of Bioconductor functions that have been wrapped into specific functions in order to ease and systematize the pre-processing and statistical analysis of Agilent microRNA data.The development of this package contributes to the Bioconductor project filling the gap in microRNA array data analysis.

View Article: PubMed Central - HTML - PubMed

Affiliation: Epidemiology, Atherothrombosis and Imaging Department, Centro Nacional de Investigaciones Cardiovasculares Carlos III, Melchor Fernández Almagro 3, E-28029 Madrid, Spain. plopez@cnic.es

ABSTRACT

Background: The main research tool for identifying microRNAs involved in specific cellular processes is gene expression profiling using microarray technology. Agilent is one of the major producers of microRNA arrays, and microarray data are commonly analyzed by using R and the functions and packages collected in the Bioconductor project. However, an analytical package that integrates the specific characteristics of microRNA Agilent arrays has been lacking.

Results: This report presents the new bioinformatic tool AgiMicroRNA for the pre-processing and differential expression analysis of Agilent microRNA array data. The software is implemented in the open-source statistical scripting language R and is integrated in the Bioconductor project (http://www.bioconductor.org) under the GPL license. For the pre-processing of the microRNA signal, AgiMicroRNA incorporates the robust multiarray average algorithm, a method that produces a summary measure of the microRNA expression using a linear model that takes into account the probe affinity effect. To obtain a normalized microRNA signal useful for the statistical analysis, AgiMicroRna offers the possibility of employing either the processed signal estimated by the robust multiarray average algorithm or the processed signal produced by the Agilent image analysis software. The AgiMicroRNA package also incorporates different graphical utilities to assess the quality of the data. AgiMicroRna uses the linear model features implemented in the limma package to assess the differential expression between different experimental conditions and provides links to the miRBase for those microRNAs that have been declared as significant in the statistical analysis.

Conclusions: AgiMicroRna is a rational collection of Bioconductor functions that have been wrapped into specific functions in order to ease and systematize the pre-processing and statistical analysis of Agilent microRNA data. The development of this package contributes to the Bioconductor project filling the gap in microRNA array data analysis.

Show MeSH
Related in: MedlinePlus