Fig. 4: Disease-specific grouping of TCR repertoire samples via ultra-large-scale clustering. | Nature Communications

Fig. 4: Disease-specific grouping of TCR repertoire samples via ultra-large-scale clustering.

From: GIANA allows computationally-efficient TCR clustering and multi-disease repertoire classification by isometric transformation

Fig. 4

a Graphic representation for the similarities of the TCR-seq samples based on TCR co-clustering. Sample-wise count-sharing matrix was computed from the original TCR clustering results of the 1,213 reference samples. Spearman correlation matrix was calculated based on counts of co-clustered TCRs, with pairs having a correlation value ≤0.4 set to be zero. The resulting sparse matrix was used to generate the graph. Nodes with fewer than two connections were removed to visualize the sample groups. b ROC curves using disease fractions calculated from co-clustered TCRs. AUC values were labeled at the bottom right of each panel. 95% confidence intervals were calculated using 2,000 stratified bootstraps. Disease abbreviations: GBM for glioblastoma multiforme; RCC for renal clear cell carcinoma; MS for multiple sclerosis.

Back to article page