--aggregate
In the case of DSR tests, multiple mutually exclusive features may be grouped, and several interdependent features may satisfy the selection criteria. Such features may vary in opposite directions and create false patterns during clustering. To eliminate the dependencies, the --aggregate option selects one representative entry per group in an unbiased way, as follows. The feature with the largest average PSI value in the control samples is chosen, to determine the direction of variation, i.e. up or down-regulation. Then, to account for the effects of potentially multiple isoforms varying in the same direction, the additive contributions of all features varying in the same direction is calculated for each sample and shown in the display. In the ‘unsupervised’ case, the first sample is arbitrarily chosen as ‘control’ to determine the direction of variation, and the same procedure is applied.