MethScope

Analysis of DNA methylomes via Most Recurrent Methylation Patterns (MRMPs) — annotate, deconvolve, impute, and embed bulk, single-cell & spatial data, uniquely robust on sparse & ultra-sparse inputs.

C CLI + R packageconda 0.1.1CRAN 1.0.3R ≥ 4.0AGPL-3.0no-GPU

conda install -c zhou-lab -c conda-forge methscope-cli↗📦 R package🗂 Pretrained models

Applications what MethScope does

🧬Sparse methylome annotationPredict cell type, sex, age and other traits from sparse methylomes, with confidence.predict → 🩸DeconvolutionEstimate cell-type proportions in mini-bulk / cfDNA mixtures (NNLS).deconv → 🧩Imputation & upscalingReconstruct CpG-level methylation from ultra-sparse input.upscale → 🗺️Clustering & embeddingMRMP representations for UMAP, clustering, and structure discovery.matrix →

Two ways to run same MRMP models, shell or R

⌨ Command line methscope-cli · no R runtime

# install (no R runtime)
conda install -c zhou-lab -c conda-forge methscope-cli
# annotate · deconvolve · impute
methscope predict query.cg model.ubjx > labels.tsv
methscope deconv  mixture.cg panel.refx > props.tsv
methscope upscale -o out.cg model.updecx query.cg

▸ runnable examples →

📦 R package interactive · plotting

# install (CRAN)
install.packages("MethScope")
# annotate + visualize in R
x <- GenerateInput(cg, mrmp)
m <- Zhou2025_HumanAtlas_P1000()
p <- PredictCellType(m, x)
PlotUMAP(x, p)

▸ runnable example →

Documentation go deeper

Prepare input files →BED / ALLC / bigwig → .cg Build & interpret MRMPs →recurrent methylation patterns Work with pretrained models →bundle / unbundle & formats Function reference →all exported functions