ballgown

Ballgown is a software tool that can be used with the Tuxedo suite and downstream Bioconductor tools. Its purpose is to enhance the statistical accuracy, experimental design flexibility, and computational speed of RNA-seq analyses. Using Ballgown, researchers can extract a subset of data to study and test all assembled transcripts for differential expression between two groups. This capability is particularly important for scientists trying to identify the genes expressed differently in two groups of individuals.

One of the advantages of using Ballgown is its ability to integrate with the Tuxedo suite. The Tuxedo suite is a widely-used pipeline for transcript assembly, quantification, and flexible differential expression analysis at transcript resolution. It consists of aligning reads using Bowtie and Tophat, assembling transcripts using Cufflinks, and performing differential expression analysis using Cuffdiff. Although the Tuxedo suite is a powerful toolset, it has some limitations. For example, Cuffdiff can only be used for two-group differential expression analyses, is computationally demanding, and produces strongly conservative estimates of statistical significance. Additionally, per-feature read counts are inappropriate for isoform-level analysis, as isoforms from the same gene may have a high degree of overlap, leading to ambiguous read counts.

Ballgown has been developed to address these limitations to connect the Tuxedo suite with other tools, such as EdgeR, DESeq, and Voom, which are present in Bioconductor. By doing so, Ballgown can provide more accurate statistical analysis, experimental design flexibility, and computational speed. Moreover, Ballgown can analyze isoform-level expression, which is impossible with per-feature read counts.

To demonstrate the effectiveness of Ballgown, researchers conducted a test involving subjects in the Finnish population. They randomly assigned subjects to one of two groups and tested all assembled transcripts for differential expression between those two groups. The results from Ballgown using linear models were found to be consistent with the exon-level results from Cuffdiff and EdgeR. This test shows that the default methods in Ballgown can work without a differential expression signal.