PELEUS was designed and trained on validated data from the TRACERx study. It is the largest longitudinal patient data set with 3,200 tumour multi-regions (both primary and met sites) collected from 795 NSCLC patients over five years.
The predicted clonal neoantigens are further validated prospectively with patient samples (over 120 for different types to date). Also, deep learning algorithms are being used to enhance the prediction of neoantigen immunogenicity. So I think this will improve as times goes on. The next update will be in 2H with higher-dose cNeT (process two) and in combo with an anti-PD-1 (both process one and two).