Fig. 1: Overview of the pipeline for CCLMoff.
From: A versatile CRISPR/Cas9 system off-target prediction tool using language model
The high-throughput off-target data is encoded as the sgRNA-target site pair and concatenated by a predefined token [SEP]. The sgRNA-target site pairs go through the Input Embedding layer and feed into 12 Transformer Blocks initialized by RNAcentral. The [CLS] of the final hidden layer is employed for classification using the multilayer perceptron to predict the sgRNA-target site pair score.
