Learn R Programming

MethScope

MethScope identifies Most Recurrent Methylation Patterns (MRMPs) as the basis to encode latent representations for interpretatble analysis of DNA mehtylation data, in particular single cell and spatial methylome. The MRMPs embeddings will support automatic cell annotation, bulk deconvolution, unsupervisded clustering, and cancer cell of origin prediction.

Installation

Install the released version of usethis from CRAN:

install.packages("MethScope")

or install from github

devtools::install_github("zhou-lab/MethScope")

Usage

MethScope performs rapid cell annotation, deconvolution and other tasks for single cell and spatial DNA methylome. Please check out our documentation for more details:

Explore the MethScope Website

Citation

If you use MethScope, kindly cite (coming soon):

Hongxiang Fu, Chin Nien Lee, Cameron Cloud, Hao Xu, Yanxiang Deng, Wanding Zhou, MethScope: Ultra-fast Analysis of Sparse DNA Methylome via Recurrent Pattern Encoding.

Copy Link

Version

Install

install.packages('MethScope')

Version

1.0.0

License

MIT + file LICENSE

Maintainer

Hongxiang Fu

Last Published

December 9th, 2025

Functions in MethScope (1.0.0)

GenerateReference

Generate reference pattern labels (no default writing)
filter_cell

Filter final prediction to reduce noise
smooth_matrix

Smooth cell by pattern matrix to reduce noise
PredictCellType

Predict cell type annotation from the trained model
PlotF1

Generate F1 score barplot for each class
confidence_score_top95

Produce confidence score based on top 95 percent for XGBoost prediction
confidence_score

Produce confidence score for XGBoost prediction
PlotUMAP_fixedwindow

Generate UMAP for the final prediction based on fixed window eg.100kb bin widows
GenerateInput

Generate pattern level data for cell type annotation
PlotUMAP

Generate UMAP for the final prediction based on cell patterns
imputeRowMean

Impute missing value for 100K window matrix
PlotConfusion

Generate confusion table for the final prediction
nnls_deconv

Estimate cell type relative proportion
Input_training

Train XGBoost model to predict cell type