Skip to main content
. 2023 May 17;14:2813. doi: 10.1038/s41467-023-38414-8

Fig. 2. The small-sample learning framework and performance of KprFunc.

Fig. 2

a The architecture of KprFunc, which incorporates PWD, SMO, PLR, DNN and MAML algorithms. b ROC curves and AUC values of tenfold cross-validations of KprFunc-i and other methods. c ROC curves and AUC values demonstrating the performance of KprFunc-i and KprFunc in distinguishing functional Kpr sites from other sites. d The confusion matrix of KprFunc-i under the medium threshold (Sp ≥ 90%). e The confusion matrix of KprFunc under the medium threshold (Sp ≥ 90%). f GO-based enrichment analysis of circadian proteins with functional Kpr sites predicted by KprFunc. E-ratio, enrichment ratio (one-sided hypergeometric test).