Fig. 2: Deep neural network architecture and model comparison.
From: Accurate cross-species 5mC detection for Oxford Nanopore sequencing in plants with DeepPlant
a Overview of the signal features used by DeepPlant and the triple-encoder architecture. b, c Bi-LSTM and Transformer encoder architectures applied in DeepPlant. d Accuracy progression during CHH methylation detection training, comparing the performance of Bi-LSTM and Transformer encoders across different k-mer lengths. 9-, 13-, and 51-mer denote the lengths of model feature contexts surrounding target C at CHH sites. e Quantitative evaluation of CHH methylation detection accuracy by different models on single chromosomes using 43× O. sativa and 35× A. thaliana nanopore data. Pearson correlations were calculated between nanopore and corresponding BS-seq data. Source data are provided as a Source Data file.
