kycg-examples

Runnable examples and gallery figures for kycg — the C command-line reimplementation of knowYourCG. Each example is a self-contained script that turns the bundled example data into an enrichment table and publication-ready figures, so the kycg docs are reproducible rather than illustrative.

Prerequisites

kycg and cinderplot on your PATH:

conda install -c zhou-lab -c conda-forge kycg cinderplot

Or point the scripts at source builds with the KYCG / CINDERPLOT env vars.

Layout

data/      example inputs (query .cg + knowledgebase .cm), self-contained
scripts/   one runnable script per worked example
results/   the kycg test output table (regenerated by the scripts)
figures/   rendered gallery figures (PDF for print, PNG for the web)

Examples

Recovering tissue identity from methylation (array)

scripts/tissue_tfbs_enrichment.sh asks whether kycg recovers known biology. A tissue's hypomethylated CpGs mark the regulatory DNA it holds open, so its Hypo probes from the MM285 tissue signature should enrich for that tissue's lineage transcription factors. Tested against ReMap TF binding sites (638 factors) they do — cleanly, with no tuning:

bash scripts/tissue_tfbs_enrichment.sh   # fetches MM285:TFBSrm into the store

fetal brain → neuronal TFs	hepatocyte → hepatic TFs

Fetal-brain hypomethylation returns the neuronal program (DLX1/2/5, OTX2, PAX6, LHX2/3, EOMES); hepatocyte returns the hepatic nuclear receptors (HNF4A, PPARA, RXRA, PROX1) and the liver circadian core. The signal tracks the cells present, not the organ name — fetal_liver returns GATA1 / TAL1 (fetal liver is the site of fetal hematopoiesis).

The query masks (data/<tissue>_hypo.cg) are bundled, so the script needs no R. scripts/make_tissue_queries.R documents how they were built from the signature (the probe lists are also bundled as data/<tissue>_hypo.txt).

Metagene methylation profile (array)

scripts/metagene_profile.sh reads out the classic gene methylation architecture. A real MM285 sample's betas are averaged over the 30 metagene bins of MetagenePC (kycg test's beta column), giving the textbook profile:

bash scripts/metagene_profile.sh          # fetches MM285:MetagenePC

A hypomethylation dip at the promoter/TSS, gene-body hypermethylation rising toward the 3′ end, and a return to the intergenic baseline downstream. The beta query (data/mm285_sample.cg, MM285.1.SigDF packed as a fraction .cg) is bundled; scripts/make_tissue_queries.R records the derivation.

ChromHMM enrichment of a single-cell methylome (sequencing)

scripts/chromhmm_enrichment.sh tests one query cell (data/onecell.cg) against the 18 ChromHMM chromatin states (data/chromhmm.cm), then renders the four core KnowYourCG plots with cinderplot — no R involved.

bash scripts/chromhmm_enrichment.sh

volcano	dot
bar	waterfall

The enrichment table it produces is results/res.tsv; sort and threshold on log10_p / neglog10_fdr (the p_value / fdr columns underflow to 0 for strong hits by design).

Data provenance

data/chromhmm.cm and data/onecell.cg are the example files shipped with the knowYourCG Bioconductor package (inst/extdata/), reused here so the examples run without installing R. Both index the same whole-genome CpG row space.

data/<tissue>_hypo.{cg,txt} are the fetal-brain / fetal-liver / hepatocyte Hypo probe sets from the MM285 tissue signature (sesameData MM285.tissueSignature), each packed into a binary mask over the MM285 probe ordering. scripts/make_tissue_queries.R regenerates them (needs R + sesame and the MM285 ordering); the .cg are bundled so the example itself needs neither. The knowledgebase TFBSrm is fetched on demand into the shared YAME store.