CummeRbund.QcReport (v1) BETA

Summary

CummeRbund is a visualization package designed to help you navigate through the many inter-related files produced from a Cuffdiff RNA-Seq differential expression analysis and visualize the relevant results.  CummeRbund helps promote rapid analysis of RNA-Seq data by aggregating, indexing, and allowing you to easily visualize and create publication-ready figures of your RNA-Seq data. 

CummeRbund is a collaborative effort between the Computational Biology group led by Manolis Kellis at MIT's Computer Science and Artificial Intelligence Laboratory, and the Rinn Lab at the Harvard University department of Stem Cells and Regenerative Medicine.  This document is adapted from the CummeRbund manual for release 2.0.0.

Usage

Unlike most modules in GenePattern, the CummeRbund reporting modules require the entire output of a Cufflinks.cuffdiff job as they work with not just one or two files but rather with all of the Cuffdiff output files.  Simply drag the top-level Cuffdiff job folder into the 'cuffdiff.input' parameter from the 'Jobs' tab ('Recent Jobs' in versions of GenePattern before 3.8.0) or from the Job Results page.  The CummeRbund modules can also be directly accessed from the context menu of jobs in either of these locations.  Remember, you are submitting the entire job folder and not just a single file.

Alternatively, once a given job has been run through any one of the CummeRbund reporting modules, a reusable database file named cuffData.db will be produced that can be submitted in place of the Cuffdiff job for other CummeRbund reports.  You can use this file for job submission via all of the usual GenePattern mechanisms or you can submit the entire CummeRbund job folder for a subsequent CummeRbund job in the same way as described above for Cuffdiff jobs.  You are highly encouraged to reuse these database files wherever possible as your jobs will run much quicker and use less storage space than by starting from scratch with a Cuffdiff job.

CummeRbund.QcReport will produce a variety of result files in the form of both plots and text tables; these are described further in the Output Files section below.  You can use the feature.level parameter to control whether these should be generated at the level of genes, isoforms, transcript start sites (TSS), or coding sequences (CDS).  Note, however, that some result files are always produced at the genes level regardless of setting, while there are other result files where these settings do not apply and for which that output is generated irrespective of the chosen feature level; see the Output Files section for details.

The report.as.aggregate parameter controls whether results will be reported with replicates split out separately or together in aggregated samples.  Similar to feature.level, however, note that some result files are always generated for aggregate samples regardless of this setting.

If you used the merged.gtf output from Cuffmerge as the GTF input to Cuffdiff, then the text reports may contain Cuffdiff tracking IDs instead of gene symbols.  These will take the  form of XLOC_00001 for genes, TCONS_00000001 for isoforms, TSS1 for tss references, and P1 for cds references.  You can use the attempt.to.merge.names parameter to have the module try to merge gene symbols into these files alongside these IDs.

For more information on using RNA-seq modules in GenePattern, see the RNA-seq Analysis page.

References

Trapnell C, Hendrickson D,Sauvageau S, Goff L, Rinn JL, Pachter L. Differential analysis of gene regulation at transcript resolution with RNA-seq. Nature Biotechnology. 2013;31:46-53.

Links

CummeRbund website and manual.

Parameters

* - required

Input Files

1. <cuffdiff.input> (required)
   A Cuffdiff job, a previous CummeRbund job, or a cuffData.db file from a previous CummeRbund job.  Unlike most modules in GenePattern, the CummeRbund reporting modules require the entire output of a Cufflinks.cuffdiff job as they work with not just one or two files but rather with all of the Cuffdiff output files.  Simply drag the top-level Cuffdiff job folder into the 'cuffdiff.input' parameter from the 'Jobs' tab ('Recent Jobs' in versions of GenePattern before 3.8.0) or from the Job Results page.  The CummeRbund modules can also be directly accessed from the context menu of jobs in either of these locations.  Remember, unless you use a cuffData.db file, you are submitting the entire job folder and not just a single file.

Output Files

1. cuffData.db
   The RSQLite database created from the original Cuffdiff job.  This file can be used in other CummeRbund jobs to avoid the need for extra computation and storage, in which case the new job will instead hold a link back to the file from the original job.
2. QC.Boxplot
   A box plot of summary statistics for FPKM values (Fragments Per Kilobase of transcript per Million mapped read) depicting groups of numerical data through their quartiles, calculated across all samples (or replicates) in the Cuffdiff dataset.  See this explanation from the Cufflinks website and the this entry from the Cufflinks FAQ for more information about FPKM values.
3. QC.Dendrogram
   A dendrogram based on Jensen-Shannon distances between conditions, across all samples (or replicates) in the Cuffdiff dataset.  Dendrograms can provide insight into the relationships between conditions for various genesets (e.g. significant genes used to draw relationships between conditions).
4. QC.Density
   A density plot of log10 FPKM values across all samples (or replicates) in the Cuffdiff dataset.  This can be used to assess the distributions of FPKM scores across samples by visualizing an estimate of the underlying probability density function.  This Wikipedia entry gives some background on density plots.
5. QC.DimensionalityReduction.mds
   A plot of a projection of expression estimates across samples (or replicates) onto a smaller number of dimensions using multidimensional scaling (MDS).  This plot places the samples into two-dimensional space with the distances between them representing the degree of similarity (closer representing more similar).  This dimensionality reduction can be useful for feature selection, feature extraction, and outlier detection.  This Wikipedia entry gives some background on multidimensional scaling; this overview and this article give more details.
   MDS plots are available only when analyzing more than two samples (or replicates).
6. QC.DimensionalityReduction.pca
   A projection of expression estimates across samples (or replicates) onto a smaller number of dimensions using principal component analysis (PCA).  It is a means of identifying patterns in highly dimensional data by highlighting the variables with the greatest variance across the samples.  This dimensionality reduction can be useful for feature selection, feature extraction, and outlier detection.  It is recommended to plot with the second and third principal components, as the first principal component will most probably be expression level which is unlikely to be informative.  This Wikipedia entry gives some background on principal component analysis, and this tutorial gives more details, as does this second tutorial.
7. QC.Dispersion
   A scatter plot comparing the mean counts against the estimated dispersion for a given level of features from a cuffdiff run.  This is a way to evaluate the quality of the model fitting to look for overdispersion in your RNA-Seq data.
8. QC.FPKM.SCV
   A plot of the squared coefficient of variation between empirical repicate FPKM values per sample, across the range of FPKM estimates.  This is a normalized measure of cross-replicate variability that can be useful for evaluating the quality your RNA-seq data.  Differences in CV² can result in lower numbers of differentially expressed genes due to a higher degree of variability between replicate FPKM estimates.
9. QC.JSDistanceHeatmap.Samples
   A heatmap of all pairwise similarities between conditions at either the sample or the replicate level, measured using Jensen-Shannon distance.  In this heatmap, "cooler" colors represent greater similarity (smaller JS distance) while "hotter" colors represent greater difference.  Similarities between conditions and/or replicates can provide useful insight into the relationship between various groupings of conditions and can aid in identifying outlier replicates that do not behave as expected.
10. QC.sig_diffExp_genes.txt (or _isoforms or _TSS or _CDS)
    A text listing of data for differentially expressed genes, isoforms, transcription start sites, or coding sequences, respectively.  Only entities that pass a significance threshold are included.  If you used a Cuffmerge merged.gtf when running Cuffdiff, you can use the attempt.to.merge.names parameter to include gene names alongside the tracking IDs.
11. QC.sig_promoter_data.txt, QC.sig_splicing_data.txt and QC.sig_relCDS_data.txt
    Text listings of promoter, splicing, and relCDS distribution-level test data, respectively.  Only entities that pass a significance threshold are included.  If you used a Cuffmerge merged.gtf when running Cuffdiff, you can use the attempt.to.merge.names parameter to include gene names alongside the tracking IDs.
12. stdout.txt (and stderr.txt)
    A log of output (and errors) produced during the database creation and plotting process.  In case of an error, check both of these files for more details.  The module has been designed to skip those plots where it encounters a problem along the way, continuing on to the next; if a given plot is missing, it should be noted in one of these files along with a reason if one could be determined.

Example Data

There is an example reusable database file available on our FTP site.  This was generated using the example data and workflow from the Differential analysis of gene regulation at transcript resolution with RNA-seq article referenced above, by Trapnell, et al.

Requirements

CummeRbund.QcReport requires R 2.15.  When installing this module, GenePattern will automatically check for the presence of this exact version of R and will not proceed without it.  See the section of our Administrator's Guide on the R Installer plug-in for details.  Installing this module requires a number of supporting R packages from CRAN and Bioconductor; it will also check for their presence and install any that are missing in the process.  These packages will be installed in a separate area specific to GenePattern and will not affect any other R library on the machine.

Please install R2.15.3 instead of R2.15.2 before installing the module. The GenePattern team has confirmed test data reproducibility for this module using R2.15.3 compared to R2.15.2 and can only provide limited support for other versions. The GenePattern team recommends R2.15.3, which fixes significant bugs in R2.15.2, and which must be installed and configured independently as discussed in Using Different Versions of R and Using the R Installer Plug-in. These sections also provide patch level fixes that are necessary when additional installations of R are made and considerations for those who use R outside of GenePattern.

Task Type:
RNA-seq

CPU Type:
any

Operating System:
any

Language:
R