Interesting work. The clonal frequency of the TCRs was included in the model. Tumours often have more clonal TCRs, whereas normal tissues generally do not, so I

Interesting work. The clonal frequency of the TCRs was included in the model. Tumours often have more clonal TCRs, whereas normal tissues generally do not, so I would imagine this would be an important feature (in fact including the TCRB prevalence gave better predictions than the 4-mer prevalence for breast cancer). Can't see in the paper that they ran the model without the frequency feature, or even ran it with just the frequency feature. Building a very simple model probably would have been a more salient comparator than shuffling the labels.