Autoencoder based local T cell repertoire density can be used to classify samples and T cell receptors
Fig 2
A. Edit distances distribution. Histogram of edit distances between all pairs of CDR3 sequences from two samples. Few sequences have 2 or fewer mismatches. B. Models glossary. The table describes the model and model acronym used in all the following figures and in the text. C. E accuracy. All accuracies of E for the different datasets with no mismatches (fully matched, red), 1 mismatch (orange), and 2 mismatches (blue). D. VE accuracy. All accuracies of VE for the different datasets (except for datasets without Vs). E. Encoder and CDR3 distances correlation. Pairwise Euclidean distances in the set of CDR3 one-hot vectors (Xs) in the x axis and pairwise Euclidean distances in the embedded set (Zs) predicted by E in the y axis. The legend contains the Spearman correlation between the axes. F. Encoder’s distances correlation when combined with distances. Pairwise Euclidean distances in the set of CDR3 one-hot vectors (Xs) in the x axis and pairwise Euclidean distances in the embedded set (Zs) predicted by EM in the y axis. G. EM accuracy. All accuracies of EM for the different datasets. H. VEM accuracy. All accuracies of VEM for the different datasets.