The reads should first be mapped to exons using the tool "Map aligned reads to exons for DEXSeq". Count files obtained like this should then be combined using the tool "Utilities / Define NGS experiment", which produces a count table and a phenodata file. Once the experimental groups have been indicated in the phenodata, the count table can be used by DEXSeq to analyze differential exon expression.
It is highly recommended to always have at least two biological replicates for each experiment condition. If this is not possible, you can still run the analysis by manually setting the dispersion factor with the parameter "Dispersion estimate". By default the dispersion estimate is set to 0.1, which is somewhere in-between what is usually observed for technical replicates (0.01) and human data (0.4). It is recommended to experiment with different values for this parameter.
The analysis output consists of the following files: