The Splicing Efficiency pipeline

The splicing efficincy pipeline is similar to the RPKM pipeline in that it is a pipeline which is designed specifically for RNA-Seq data, and looks at the content of exons but in the context of whole transcripts or genes.

The purpose of the pipeline is to provide a way to measure the relative proportion of reads falling into introns and exons in a quantitative way. This proportion can be useful in a number of ways but is most likely to be used when the relative propotion of mature and unspliced transcript can be used to indirectly measure the efficiency of splicing, or the efficiency of the degradation of the mature message.

The pipeline generates a set of probes covering every gene in the genome, it also makes up a merged set of regions covering all of the exons of all of the splice forms of that gene. It can then count the number of reads falling into either the exonic or intronic parts of the gene and can base its calculation on these.

Options

The options you can set for this pipeline are:

1. The feature type to use as genes and transcripts for this analysis. This will default to gene and mRNA if these are present. Other types can be selected, but it is important that the transcript features use the Ensembl convention of naming transcripts by the feature name followed by a dash and a 3 digit number indicating the splice form. If this convention is not used then the pipeline will not be able to correctly match transcripts to their corresponding gene and the counts produced will be wrong.
2. The type of library you are quantitating. Some RNA-Seq libraries are strand specific and in these cases the pipeline can ignore reads coming from the wrong strand. You can also choose between strand specific libraries which produce reads on the same strand as the feature or the opposing strand.
3. Whether you want to count reads instead of bases. The basic counts for this quantitation are made in bases since many reads will be split over two or more exons and would otherwise be counted multiple times. If this option is selected the pipeline works out the longest read length from each dataset and divides the base counts by this value, rounding down any fractional values, to obtain a number of reads per transcript. You generally want to apply this transformation otherwise undue emphasis is placed on the set of reads which only have part of one read overlapping them.
4. Whether the results should be log2 transformed. Data analysis and visualisation of this type of data is often easier when performed on a log scale. If this option is selected then empty exons or introns will be given a count of 0.9 bases (or 0.9 reads if read length correction is applied). This count is applied before read length or total read count correction is applied.
5. Whether to correct for the length of each transcript. If this option is selected then the quantitated values are expressed per kilobase of transcript. This option is selected by default and makes sense when comparing the density of reads in exons and introns where you can say that a 1:1 ratio indicates the same read density in both contexts.