scCapsNet-mask: an automatic version of scCapsNet
Wang, L.; Zhang, J.; Cai, J.
Show abstract
SummaryRecently we developed scCapsNet, an interpretable deep learning cell type classifier for single cell RNA sequencing data, based on capsule network. However, the running process of scCapsNet is not fully automatic, in which a manual intervention is required for getting the final results. Here we present scCapsNet-mask, an updated version of scCapsNet that utilizes a mask to fully automate the running process of scCapsNet. scCapsNet-mask could constrain the internal parameter coupling coefficients and result in a one to one correspondence between the primary capsule and type capsule. Based on those bijective mapping between primary capsule and type capsule, the model could automatically extract the cell type related genes according to weight matrix connecting input and primary capsule, without a need for manual inspection of the relationship between primary capsules and type capsules. The scCapsNet-mask is evaluated on two single cell RNA sequence datasets. The results show that scCapsNet-mask not only retains the merits of the original scCapsNet with high classification accuracy and high interpretability, but also has the virtue of automatic processing.
Matching journals
The top 6 journals account for 50% of the predicted probability mass.
Similar papers in this journal
Similar papers in this journal
- scPADGRN: A preconditioned ADMM approach for reconstructing dynamic gene regulatory network using single-cell RNA sequencing data 96%
- Cell-type annotation with accurate unseen cell-type identification using multiple references 96%
- DeepHE: Accurately Predicting Human Essential Genes based on Deep Learning 96%
Similar papers in this journal
- SCSA: a cell type annotation tool for single-cell RNA-seq data 95%
- Deepprune: Learning efficient and interpretable convolutional networks through weight pruning for predicting DNA-protein binding 95%
- GMQN: A reference-based method for correcting batch effects as well as probes bias in HumanMethylation BeadChip 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.