Ablation study results comparing the performance of different model configurations across sequencing platforms. (A) Training loss curves across different models. The results shown are based on the training datasets from three sequencing platforms: CLR, ONT, and CCS. (B) Training true positive (TP) ratio count curves. The results shown are based on the training datasets from three sequencing platforms: CLR, ONT, and CCS. (C) Distribution of validation F1 score. (D) Distribution of validation true positive ratio.