prediction and percentile_rank

[wangjiawen2013]

Hi,
I have tested netTCR2.2 using the sample data on https://services.healthtech.dtu.dk/services/NetTCR-2.2/, but I don't know how to interpret the results. What do "prediction" and "percentile_rank" mean ? which one could represent the binding specificity ? and why some "percentile_rank" are "NaN" ?

This is the output:

[JonasNilsson1]

Hi,

The 'prediction' column refers to the raw prediction score from the model, while 'percentile_rank' refers to a rank score calculated based on a set of background TCRs. The lower the percentile rank, the more confident the prediction is.

Regarding the NaNs, these correspond to the peptide KLVALGINAV, for which the NetTCR-2.2 method does not have a peptide-specific model, so it uses a pan-specific model for the predictions. For these peptides, percentile rank scores are not available, hence the NaNs.

Hope this clarifies things

Best
Jonas