Node reference

Methylation nodes.

Illumina methylation arrays (27K, 450K, EPIC v1, EPIC v2, Mouse). Load IDAT reads raw intensities with the mepylome engine — methylprep only supplies manifests and annotation — and applies raw-signal preprocessing chosen by its prep parameter: raw, illumina, noob or swan. Load Methyl Beta takes a vendor beta matrix instead. Then Methyl QC (detection p, intensity, control strip, SNP heatmap, call rates), Methyl Normalize (NaN-aware between-array quantile, or none), Methyl Filter, DMP (Welch or limma-style moderated t), a 3+ group F-test, bumphunter-style DMRs, BED / ENCODE cCRE feature enrichment with a forest plot, CpG-bias-corrected GOmeth, and Manhattan / QQ / volcano / heatmap / locus plots.

12 nodesCategory · methylationin 4 templates

Where things go

Phenotype attaches with Merge Cell Meta (a sample sheet keyed by sample ID), not through the loader. Manifests and the extended probe annotation are resources — see the Resource Center.

Read before trusting an old result

BMIQ and noob are not Methyl Normalize options. They were removed in v0.14.4 because the implementations behind those labels were not BMIQ or NOOB and corrupted beta values silently. Real NOOB and SWAN operate on raw signal and live in Load IDAT’s prep. Methylation results produced with the old options should be re-run.

SWAN is not deterministic. It uses random sampling, so two runs differ slightly. Prefer illumina or noob when reproducibility matters.

Run Methyl Filter before DMP. SNP-associated, sex-chromosome, cross-reactive and low-call-rate probes are removed there; DMP without it is dominated by artefacts.

Load IDAT

loadIDAT

Load Illumina methylation IDAT (450K / EPIC v1 / EPIC v2 / Mouse). EPIC v2 auto-routes to mepylome (methylprep ceiling). IDAT pairs (_Grn + _Red) required. The array manifest is installed once through the Resource Center — a run never downloads. Detects Fasoo DRM lock and refuses with a clear message.

Node LibraryMethylation

OutAnnData (samples × probes, beta + M-values)

Parameters (8)

Name	Type	Default	Description
`idat_dir`	directory	—	IDAT directory (recursive).
`array_type`	select	`auto`	auto/450k/epic/epicv2/mouse.
`collapse_replicates`	checkbox	`true`	Collapse EPIC v2 replicate probes.
`replicate_method`	select	`mean`	mean/median/first.
`prep`	select	`illumina`	Raw-signal preprocessing applied before beta values are computed: 'illumina' (Illumina's own scaling — the default and the historical behaviour) / 'noob' (real out-of-band background correction) / 'swan' (subset-quantile within-array normalization, corrects Type I vs Type II design bias) / 'raw' (no preprocessing). Blank = 'illumina'.
`sample_sheet`	file (.csv, .tsv, .txt)	—	Optional sample sheet.
`value_type`	select	`beta`	"beta" or "m".
`beta_file`	file (.csv, .tsv, .txt)	—	Optional fallback path to a pre-processed beta matrix if IDAT loading fails.

Load Methyl Beta Matrix

loadMethylBeta

Load pre-processed beta-value matrix (CSV/TSV/TXT). Used when raw IDAT is unavailable. Auto-detects orientation. Vendor percent (0-100) auto-divided. EPIC v2 replicates collapsed.

Node LibraryMethylation

OutAnnData (samples × probes)

Parameters (8)

Name	Type	Default	Description
`beta_file`	file (.csv, .tsv, .txt)	—	Beta matrix file (.csv/.tsv/.txt, .gz OK).
`separator`	select	`auto`	auto/comma/tab.
`orientation`	select	`auto`	auto/probes_x_samples/samples_x_probes.
`array_type`	select	`auto`	450k/epic/epicv2/mouse.
`collapse_replicates`	checkbox	`true`	Collapse EPIC v2 replicate probes (same CpG measured multiple times).
`replicate_method`	select	`mean`	Aggregation for replicate probes: mean / median / first.
`value_type`	select	`beta`	"beta" or "m".
`sample_sheet`	file (.csv, .tsv, .txt)	—	Optional CSV/TSV whose row IDs match sample column names in the beta matrix.

Methyl QC

methylQC

Per-sample / per-probe call rates with a built-in QC plot. "label" or "filter" mode.

Node LibraryMethylation

InAnnData (methylation)

OutAnnData (obs/var QC columns), PNG image (optional)

Parameters (15)

Name	Type	Default	Description
`min_call_rate_sample`	number	`0.95`	Min sample call rate.
`min_call_rate_probe`	number	`0.95`	Min probe call rate.
`mode`	select	`label`	"label" or "filter".
`show_plot`	checkbox	`true`	Render a QC plot. Drawn before filtering, so failing samples stay visible.
`plot_type`	select	`violin`	"violin" (one panel per metric with the call-rate threshold line), "scatter" (call rate vs mean beta), or "histogram".
`metrics`	text	`call_rate,mean_beta,median_beta`	Comma-separated metrics to plot. Add "probe_call_rate" for a probe-level panel.
`color_by`	obs column	—	Sample metadata column for colouring. Empty falls back to QC_pass.
`group_colors`	color map	—	Canonical categorical color overrides for the active color_by owner. Empty color_by uses obs['QC_pass']; null preserves the existing plot palette.
`detection_convention`	select	`minfi_mean`	Which detection-p convention plot_type='detection_p' renders. These are two DIFFERENT STATISTICS on the same data, each with its own cutoff, and the cutoffs are NOT interchangeable. 'minfi_mean' = per-sample MEAN of minfi's detectionP against a 0.01 sample cutoff (the F1000 workflow / ewastools idiom, barplot(colMeans(detP)) + abline(h=0.01)). 'poobah_fraction' = per-sample FRACTION of probes failing pOOBah at probe p>0.05, against a 0.05 sample cutoff (the sesame method with the wateRmelon pfilter / meffil sample rule). Ignored unless plot_type='detection_p'.
`control_types`	text	`BISULFITE CONVERSION I,BISULFITE CONVERSION II`	Comma-separated Illumina Control_Type names to panel in plot_type='control_strip'. Case-insensitive; unknown names are skipped with a warning; 'all' shows every type on the array. Ignored unless plot_type='control_strip'. Typical EPIC v2 types: BISULFITE CONVERSION I, BISULFITE CONVERSION II, EXTENSION, HYBRIDIZATION, NEGATIVE, NON-POLYMORPHIC, NORM_A, NORM_C, NORM_G, NORM_T, RESTORATION, SPECIFICITY I, SPECIFICITY II, STAINING, TARGET REMOVAL (the exact set is array-specific — the node reads it from the manifest).
`snp_cmap`	select	`RdYlBu_r`	Colormap used only by plot_type='snp_heatmap'. Channel and PASS/FAIL diagnostic colors remain fixed because they encode physical or QC meaning. Options: RdYlBu_r, viridis, plasma, coolwarm, Spectral_r, Blues.
`font_size`	number	—	Matplotlib font size override. Empty = matplotlib default.
`publication_quality`	checkbox	`false`	300 dpi output when on, else 96.
`figsize_w`	number	`12`	Figure width (inches).
`figsize_h`	number	`4`	Figure height (inches).

Methyl Normalize

methylNormalize

NaN-aware between-array quantile normalization in beta space, or no-op. Raw-signal NOOB/SWAN preprocessing belongs in Load IDAT.

Node LibraryMethylation

InAnnData

OutAnnData (normalized X, layers normalized_beta)

Parameters (2)

Name	Type	Default	Description
`method`	select	`quantile`	quantile or none. Retired bmiq/noob values fail with a migration message.
`convert_to_m`	checkbox	`false`	Place M-values in X.

Methyl Filter

methylFilter

Drop SNP-associated, sex chromosome, cross-reactive, low call rate probes. Mandatory before DMP.

Node LibraryMethylation

InAnnData

OutAnnData (probes filtered)

Parameters (8)

Name	Type	Default	Description
`remove_snp_probes`	checkbox	`true`	rs* + manifest SNP set.
`remove_sex_chr`	checkbox	`true`	Drop chrX/chrY.
`remove_ch_probes`	checkbox	`true`	Drop 'ch.' non-CpG (CpH) probes — ChAMP filterNoCG parity.
`remove_masked_probes`	checkbox	`false`	Drop probes flagged by the Zhou-lab MASK_general recommendation (needs var['mask_general']).
`remove_cross_reactive`	checkbox	`false`	Bundled or user CSV blacklist.
`cross_reactive_file`	file (.csv, .tsv, .txt)	—	Optional probe_id CSV.
`min_call_rate`	number	`0.95`	Drop probes below this.
`max_detp_frac_failed`	number	`0.0`	Maximum fraction of samples in which a probe may fail detection-p. 0 = drop a probe failing in any sample. Applied only when Load IDAT wrote the per-probe detection columns.

DMP Analysis

dmpAnalysis

Differentially Methylated Position — vectorized Welch's t-test or limma-style moderated t-test. reference_group / test_group are dropdowns populated from obs categories (v0.9.0-beta.42+).

Node LibraryMethylation

InAnnData (filtered, normalized, with phenotype)

OutAnnData (uns["dmp_results"])

Parameters (9)

Name	Type	Default	Description
`condition_col`	obs column	—	obs column with two groups.
`reference_group`	obs value	—	Baseline group.
`test_group`	obs value	—	Comparison group.
`test`	select	`welch`	"welch" or "moderated".
`use_m_values`	checkbox	`true`	Use M-values.
`covariates`	text	—	Comma-separated obs columns to adjust for (age, sex, slide, cell-type proportions). Non-empty switches the node to a per-probe linear model that tests the condition coefficient. Numeric columns enter as-is; anything else is treated as a factor and treatment-coded. A column perfectly confounded with the condition is REFUSED with an explanation, never silently dropped.
`block`	obs column	—	A single obs column to hold constant — subject ID for paired samples, or batch / slide. Added to the design alongside covariates.
`alpha`	number	`0.05`	BH FDR threshold.
`delta_beta_threshold`	number	`0.1`	Min \|delta_beta\|.

Multi-Group DM (3+ groups)

methylMultiGroupDM

Multi-group (3+) differential methylation — omnibus moderated F-test (empirical-Bayes shrinkage, limma-style) plus a reference contrast per group. Pure numpy, not R limma. Runs the whole 3+ group comparison in one node and keeps every result. Leave downstream_group blank for the omnibus/union view (best for a 3-group heatmap) or pick a group to promote that pairwise contrast to Volcano/DMR. For exactly two groups use DMP Analysis.

Node LibraryMethylation

InAnnData (filtered, normalized, with a 3+ group phenotype)

OutAnnData (uns["multigroup_*"] + promoted uns["dmp_results"])

Parameters (7)

Name	Type	Default	Description
`condition_col`	obs column	—	obs column with 3 or more groups.
`reference_group`	obs value	—	Baseline for every contrast.
`method`	select	`moderated_f`	"moderated_f" (eBayes) or "plain_anova".
`use_m_values`	checkbox	`true`	Use M-values.
`alpha`	number	`0.05`	BH FDR (omnibus + contrasts).
`delta_beta_threshold`	number	`0.1`	Min \|delta_beta\|.
`downstream_group`	obs value	—	View promoted downstream: blank = omnibus/union, or a group for that contrast.

DMR Analysis

dmrAnalysis

Differentially Methylated Regions (Bumphunter-style) — groups adjacent same-direction DMPs within max_gap_bp.

Node LibraryMethylation

InAnnData (with DMP results, chr/pos in var)

OutAnnData (uns["dmr_results"])

Parameters (6)

Name	Type	Default	Description
`min_probes`	number	`3`	Min CpGs per DMR.
`max_gap_bp`	number	`1000`	Max bp gap between probes.
`require_same_direction`	checkbox	`true`	All probes agree in direction.
`smooth_window`	number	`0`	Rolling-mean window (probes).
`alpha`	number	`0.05`	DMP FDR threshold.
`delta_beta_threshold`	number	`0.1`	Min \|delta_beta\| per probe.

DMP Feature Enrichment

dmpFeatureEnrichmentsince v0.15.0

Fisher enrichment of significant DMPs against local BED or cached ENCODE SCREEN cCRE classes, with BH FDR and a 95% CI forest plot. Uses non-significant probes from the actual DMP-tested universe as background. Custom BED and ENCODE modes are local/cache-only; genome-build mismatch fails and no implicit liftover or download occurs.

Node LibraryMethylation

InAnnData (with DMP results and chr/pos), Local BED3+ annotation

OutAnnData (managed enrichment table + provenance), PNG forest plot

Parameters (22)

Name	Type	Default	Description
`feature_source`	select	`custom_bed`	custom_bed or encode_ccre; both require a local bed_path.
`bed_path`	file (.bed, .gz, .tsv, .txt)	—	Required local or cached zero-based, half-open BED file.
`feature_column`	text	—	Optional zero-based class column or BED-header name; blank uses the source default.
`feature_checksum`	text	—	Optional expected SHA-256; mismatch fails closed.
`genome_build`	select	—	Probe build from metadata, or hg19/hg38/mm10/mm39.
`feature_genome_build`	select	—	BED build from its header; ENCODE defaults to hg38.
`probe_position_base`	select	`1`	Illumina positions are normally one-based.
`direction`	select	`all`	all, hyper, hypo, or split.
`alternative`	select	`greater`	Fisher alternative: greater, less, or two-sided.
`fdr_scope`	select	`per_direction`	BH correction per direction or globally.
`fdr_alpha`	number	`0.05`	BH FDR threshold.
`forest_x_min`	number	`0.01`	Log-axis display minimum; stored statistics are not clipped.
`forest_x_max`	number	`100.0`	Log-axis display maximum; stored statistics are not clipped.
`max_forest_rows`	number	`60`	Rows rendered; full result table remains stored.
`forest_all_color`	color	`#4C78A8`	Forest point/interval color for the combined all-DMP direction.
`forest_hyper_color`	color	`#D64F4F`	Forest point/interval color for hypermethylated DMP enrichment rows.
`forest_hypo_color`	color	`#3B75AF`	Forest point/interval color for hypomethylated DMP enrichment rows.
`title`	text	`DMP genomic-feature enrichment`	Forest plot title.
`dpi`	number	`150`	Forest PNG DPI, clamped to 72-600.
`feature_release`	text	—	Optional annotation release provenance.
`feature_url`	text	—	Optional provenance only; never fetched.
`feature_download_date`	text	—	Optional local annotation download date.

Methyl Visualization

methylVisualization

Genomic-coordinate Manhattan, capped/raw QQ and volcano, top/random/most-variable heatmap up to 10,000 probes, and direction-aware annotation distributions plus existing methylation plots.

Node LibraryMethylation

InAnnData (methylation)

OutPNG image

Plots (10)beta_distnorm_comparisonbeta_by_typeqq_plotmanhattanheatmapsample_pcavolcanocpg_boxplotfeature_distribution

Parameters (29)

Name	Type	Default	Description
`plot_type`	text	`beta_dist`	10 plot types: beta_dist, norm_comparison, beta_by_type, qq_plot, manhattan, heatmap, sample_pca, volcano, cpg_boxplot, feature_distribution.
`n_top_probes`	number	`50`	Heatmap probes, up to 10,000.
`selection_mode`	select	`top_dmp`	top_dmp, seeded random, or most_variable.
`random_seed`	number	`0`	Reproducible random-probe seed.
`variability_metric`	text	`sd`	SD or MAD for most-variable selection.
`min_observed_fraction`	number	`0.8`	Minimum finite sample fraction per selected probe.
`cluster_probe_limit`	number	`2000`	Disable row clustering above this count; sample clustering may continue.
`dendrogram`	checkbox	`true`	Heatmap only — cluster probes and samples hierarchically (clustermap). Off gives the plain rank-ordered heatmap.
`color_by`	text	—	Sample color column. Used by sample_pca (point colour) and, as a column colour bar, by the clustered heatmap.
`beta_type_stage`	select	`auto`	beta_by_type only — which matrix to draw: "auto" (normalized if present), "raw", or "normalized".
`cpg_probes`	text	—	cpg_boxplot only — comma-separated probe ids. Empty auto-picks the top hits by adjusted p-value.
`n_cpg`	number	`6`	cpg_boxplot only — how many CpGs to auto-pick (1–12). Ignored when cpg_probes is set.
`group_by`	text	—	cpg_boxplot only — grouping column. Empty falls back to the DMP condition column.
`show_points`	checkbox	`true`	cpg_boxplot only — overlay jittered per-sample points.
`x_axis_mode`	select	`genomic`	Manhattan cumulative genomic coordinates or legacy index.
`y_axis_mode`	select	`capped`	QQ/Manhattan/volcano display: capped or raw; p-values stay unchanged.
`y_cap`	number	—	Optional explicit -log10(p) display cap.
`feature_dist_split`	select	`all_vs_sig`	All-vs-significant or Hyper/Hypo direction.
`feature_dist_axis`	select	`both`	CpG island, TSS-distance proxy, or both.
`feature_dist_y_mode`	text	`auto`	Count, within-direction %, or within-category 100%.
`include_overall_panel`	checkbox	`false`	Show overall Hyper/Hypo counts.
`merge_shore_shelf`	checkbox	`false`	Merge north/south shore and shelf categories.
`feature_hyper_color`	text	`#f97316`	Hyper bar color for direction-split feature_distribution panels, including number_of_dmps.
`feature_hypo_color`	text	`#2563eb`	Hypo bar color for direction-split feature_distribution panels, including number_of_dmps.
`all_color`	color	`#94a3b8`	All-probe bar color for feature_distribution all_vs_sig mode.
`sig_color`	color	`#dc2626`	Significant-DMP bar color for feature_distribution all_vs_sig mode.
`x_label_rotation`	number	`45`	feature_distribution annotation-panel X tick rotation in degrees, clamped to 0-90.
`delta_beta_threshold`	number	`0.1`	Volcano cutoff.
`n_label`	number	`10`	Top hits annotated (volcano).

Methyl DMR Locus Plot

methylDMRLocusPlotsince v0.9.0-beta.41

Genome-browser-style locus plot for one DMR — per-sample beta + group means, DMR span shaded.

Node LibraryMethylation

InAnnData (with DMR results)

OutPNG image

Parameters (12)

Name	Type	Default	Description
`dmr_rank`	number	`1`	1 = top DMR by min_padj.
`padding_bp`	number	`5000`	Flanking bases.
`color_by`	obs column	—	Group column (same as DMP condition_col).
`group_colors`	color map	—	Canonical categorical color overrides for the obs column selected by color_by. null preserves the existing locus palette.
`chr`	text	—	Manual chromosome.
`start`	number	—	Manual region start (bp).
`end`	number	—	Manual region end (bp).
`figsize_w`	number	`10`	Figure width in inches.
`figsize_h`	number	`5`	Figure height in inches.
`alpha_point`	number	`0.35`	Transparency (0-1) of per-sample beta-value dots. Lower = more transparent; useful when many samples overlap.
`dmr_highlight_color`	color	`#dc2626`	Color of the selected DMR span and boundary lines.
`publication_quality`	checkbox	`false`	Generate high-resolution (300 DPI) publication-quality figure.

Methyl GOmeth

methylGOmethsince v0.9.0-beta.43

CpG-bias-corrected gene-set enrichment for DMP results (gometh-style two-step). Removes the bias toward CpG-rich genes that plain hypergeometric gives. v0.9.0-beta.43+ — gene names lazy-downloaded from zhou-lab.

Node LibraryMethylation

InAnnData (with DMP results, gene var column)

OutAnnData (uns["methyl_gometh_results"]) + PNG

Parameters (15)

Name	Type	Default	Description
`gene_set_library`	select	`KEGG_2021_Human`	gseapy/Enrichr library.
`organism`	select	`human`	"human" or "mouse".
`direction`	select	`all`	all, hyper, hypo, or split while preserving one tested-CpG universe.
`bias_correction`	checkbox	`true`	Apply gene-level Fisher correction.
`method`	select	`wallenius`	Which CpG-bias correction to use: 'wallenius' (missMethyl gometh — probability weighting by probe count plus a Wallenius non-central hypergeometric term test; the default and the published method) or 'two_stage' (gene-level Fisher selection then a plain hypergeometric term test; the pre-v0.27.1 behaviour). Ignored when bias_correction is false.
`gene_alpha`	number	`0.05`	Gene-level BH FDR.
`term_alpha`	number	`0.05`	Term-level BH FDR.
`min_overlap`	number	`2`	Min overlapping genes.
`max_term_size`	number	`1000`	Skip very generic terms.
`top_n`	number	`15`	Top terms in bar plot.
`figsize_w`	number	`10`	Bar plot width in inches.
`figsize_h`	number	`8`	Bar plot height in inches.
`term_sig_color`	color	`#dc2626`	Bar color for terms passing the term-level FDR cutoff.
`term_ns_color`	color	`#94a3b8`	Bar color for terms that do not pass the term-level FDR cutoff.
`publication_quality`	checkbox	`false`	Render at 300 DPI with publication font sizes (slower).

Bulk RNA-seq

ATAC-seq