Fig. 2

The parameter optimization of the pretrained protein language model for accurately predicting human Kla sites. A–C The heatmap of learning rate (x-axis) and batch size (y-axis) optimization in terms of Acc, MCC, and AUC metrics, respectively. Brighter yellow blocks correspond to higher metric values in the heatmap. D–F representing the performance of different learning rates, batch sizes, and sequence length. Yellow, green, and red lines indicate Acc, MCC, and AUC metrics. G ROC (receiver operating characteristic) curve and H PRC (precision-recall) curve plotting the overall prediction performance of PBertKla with the optimal parameter setting on the validation data (green line) and independent validation data (red line), respectively