We compared scMeformer to four alternative models: the CNN model, the cluster model, the scMeformer model but with only the DNA module, and the scMeformer model but with only the CpG module. The CNN model also consists of three modules, including a DNA module, a CpG module, and a fully connected network. However, in the CNN model, both the DNA and CpG modules employ a convolutional neural network (the same as described in the DNA module of scMeformer) rather than the transformer encoders. The cluster model first clusters cells into clusters based on DNAm levels of CpGs within non-overlapping 100kb bins. For each cell in a cluster, the methylation state for a CpG site not covered by any reads is imputed by known methylation states of this CpG site in cells of the same cluster.
Do you have any questions about this protocol?
Post your question to gather feedback from the community. We will also invite the authors of this article to respond.
Tips for asking effective questions
+ Description
Write a detailed description. Include all information that will help others answer your question including experimental processes, conditions, and relevant images.