Currently TPMs from all samples are scaled to the same library size. To better reflect uneven library sizes between samples and allow this to weigh in on the contribution of the samples to the decision, a scaling vector should be allowed, with a value for each sample.