What is the difference between the splicing-index, FIRMA, MiDAS, Linear Regression and ASPIRE methods?
Answer: There are multiple algorithm choices for alternative exon analysis in AltAnalyze. Some of these only work for single probeset analyses (e.g., exon arrays), while others are only available for reciprocal exon-junction analyses.
The methods splicing-index, FIRMA and MiDAS are used to compare the expression of a single probeset to the gene expression, whereas LinearRegression and ASPIRE are used when you have two probeset measuring the exclusion and inclusion of an exon (e.g, two junctions). All of these produce a score which is basically equivalent to a fold change, except for MiDAS which produces a p-value. The "fold-changes", however, have different scales depending on the method use. Each scoring method has a corresponding p-value which is calculated based on an f-test between the individual sample scores in the two groups. Some methods, like FIRMA, tend to produce much larger scores than others, such as splicing-index. Likewise, scores for ASPIRE typically range from -1 to 1 while Linear Regression ranges from -10 to 10.
Negative scores for these methods actually indicate that the experimental group expression is higher than the control. For example, a splicing-index score of -1 indicates that the probeset is upregulated in experimental versus control. This is the opposite of what most people would assume, but is the standard convention for these algorithms. If an exon is called "upregulated" it indicates that there is increased exon inclusion of the indicated exon in one or several alternative isoforms in the experimental versus control samples. The same is true for the "upregulation" call for reciprocal splicing algorithms (ASPIRE, Linear Regression) in that the inclusion junction (aka column exons1) is increased in the experimental conditions relative to the control and the exclusion junction is decreased.
For details on these methods, see the associated documentation.