Report

Quick reference for modulator's terminology and the modification codes used in this run.

Total counts of the objects modulator assembled and tested in this run — fragmentforms, genes, metagenes, segregating SNPs — and the reads supporting them.

Fragmentforms

131

Genes

15

Metagenes

15

Max ZN partitions / metagene

32

Assigned reads

54,045

Exact-chain reads

40,149

Truncation-assigned reads

13,896

Segregating SNPs

13

Haplotype blocks

3

Definitions for the overview metrics

Fragmentforms

Read-backed partial transcripts kept after assembly, read-support, and poly(A) filters. One fragmentform = one distinct (intron chain + clustered 3' end) supported by at least `min_reads` reads whose 3' ends fall within `apa_window` of each other (see input parameters `min_reads`, `apa_window`).

Genes

Distinct genes represented across the retained fragmentforms (reference gene where the fragmentform matched the annotation, otherwise an assembled locus).

Metagenes

Connected groups of genes whose exons overlap on the same strand, merged so their overlapping fragmentforms can be placed on separate ZN partitions. Every gene belongs to exactly one metagene, so the number of metagenes is always ≤ the number of genes (a gene overlapping nothing is its own metagene of one).

Max ZN partitions / metagene

The largest number of ZN partitions any single metagene needed in order to separate all of its overlapping genes and fragmentforms onto their own track, so their modification pileups are never pooled together.

Assigned reads

Total reads assigned to retained fragmentforms in the classification summary.

Exact-chain reads

Assigned reads that exactly match a retained fragmentform's canonical intron chain.

Truncation-assigned reads

Assigned reads absorbed into a retained fragmentform from suffix-compatible shorter chains (a read that is a 3' truncation of the fragmentform).

Segregating SNPs

Candidate SNP loci (read-backed variants passing the genotype discovery thresholds) retained for genotype-aware association testing (see input parameters `genotype.min_alt_reads`, `genotype.min_alt_frac`).

Haplotype blocks

Local read-backed SNP blocks retained for haplotype association testing.

↗

The highest-support fragmentforms retained after assembly and reference classification, ordered by assigned reads.

Per-sample fragmentform assignment totals and detection breadth derived from the final retained fragmentforms.

Sample-level retention across the primary read filters and fragmentform-assignment workflow.

Counts of unique genomic modification sites observed per gene and modification code in the ZN aggregation.

Where genes overlap on the same strand, modulator merges them into one metagene and resolves each read to a single gene by its fragmentform (ZT) assignment. This shows, per sample, how the reads in those overlapping-gene regions were used or scrapped (see input parameter multi_gene_action).

Read-backed loci whose fragmentforms overlap no annotated gene in the reference GTF. Each locus is formed by merging overlapping novel fragmentforms on one strand and is given a unique, deterministic, coordinate-anchored name (NOVEL_____).

No novel loci: every assembled fragmentform matched a reference gene.

Donor/acceptor dinucleotides of every assembled fragmentform intron, in fragmentform orientation. GT-AG is the major (U2) spliceosome; GC-AG is semi-canonical; AT-AC is the minor (U12) spliceosome; anything else is non-canonical and worth inspecting. Because the intron chains are read-derived, these are the junctions the reads actually support.

Summary Across All Measured Junctions

• CANONICAL_GT_AG: 1,102 (97.26%)
• SEMI_CANONICAL_GC_AG: 23 (2.03%)
• NONCANONICAL: 8 (0.71%)

The exact coordinates and donor/acceptor dinucleotide of every junction — and the specific genes carrying non-canonical junctions — are in the per-gene summary and non-canonical tables below.

Each fragmentform's TES is a cleavage/polyadenylation site. For every one, the genomic sequence around it is read in sense (fragmentform) orientation and scanned for a polyadenylation signal (canonical AATAAA or a known variant hexamer, normally 10-30 nt upstream) plus a downstream U/GU-rich element. This both annotates the site and flags lower-confidence 3′ ends: a site with no PAS whose downstream genome is A-rich is an A-rich, PAS-less 3′ end. In oligo-dT cDNA methods that is the classic internal-priming signature (a poly-dT primer annealing to an internal A-stretch instead of the tail); in direct-RNA, where the adapter is ligated to the native 3′ end, the same pattern can instead reflect a genome-encoded A-run captured at a fragment end. Either way, treat these as lower-confidence APA calls to inspect rather than confirmed cleavage sites.

↗

Summary Across All Measured APA Sites

• 131 APA sites checked — 117 (89.3%) carry a polyadenylation signal
• 42 canonical AATAAA; 75 a variant hexamer
• 0 A-rich, PAS-less 3′ ends (no PAS + A-rich downstream genome — possible internal priming or genome-encoded A-run)

Positions where the modification stoichiometry differs between the fragmentforms of a gene (across all detected mod codes; sample-stratified Cochran-Mantel-Haenszel with Benjamini-Hochberg FDR). The table shows sites at FDR<0.05 and |Δ|≥10% only — the exact thresholds this run used. Both are tunable — see input parameters classify_diffs.min_effect and classify_diffs.fdr.

↗

↗

↗

↗

↗

↗

Sites are placed in a three-level tree. Level 1 (bucket): does the base exist in the mature RNA of both forms — PRIVATE (base only in the higher form) vs SHARED. Level 2 (event) for shared sites: is the distinguishing structural change ON the base's exon (SHARED_LOCAL → ALT_DONOR / ALT_ACCEPTOR / ALT_POLYA_SITE / IPA_NO_EXTENSION) or ELSEWHERE in the transcript (SHARED_DISTAL)? SHARED_DISTAL now splits into DISTAL_APA (the forms differ in their 3' end / poly(A) site) vs DISTAL_SPLICING (same 3' end, they differ in an internal / 5' splicing choice). Level 3 (direction): which form is the more-modified one, worded per event (e.g. LONGER_EXON_HIGHER, PROXIMAL_HIGHER). A 4th bucket, UNEXPLAINABLE, holds sites with no confident structural cause. Every bucket, event and direction is ALWAYS listed — with a zero count where nothing landed there — so the taxonomy is reported completely. PRIVATE sites are detected by a COVERAGE-INDEPENDENT structural scan (every modified site checked against all fragmentform models of its gene) — so they are found even when the form lacking the base has no coverage there, which the differential test cannot see; SHARED / UNEXPLAINABLE come from the between-fragmentform differential test (BH-FDR + the >10% effect rule, tunable via classify_diffs.min_effect).

bucket	event	direction	n_sites	pct
PRIVATE	SKIPPED_EXON	IN_INCLUDED_EXON	13	6.6%
PRIVATE	INTRONIC_POLYA	IN_IPA_ISOFORM	24	12.2%
PRIVATE	THREE_PRIME_EXTENSION	IN_DISTAL_EXTENSION	55	28.1%
PRIVATE	ALT_LAST_EXON	IN_ALT_LAST_EXON	50	25.5%
SHARED_LOCAL	ALT_DONOR	LONGER_EXON_HIGHER	0	0.0%
SHARED_LOCAL	ALT_DONOR	LONGER_EXON_LOWER	0	0.0%
SHARED_LOCAL	ALT_ACCEPTOR	LONGER_EXON_HIGHER	1	0.5%
SHARED_LOCAL	ALT_ACCEPTOR	LONGER_EXON_LOWER	0	0.0%
SHARED_LOCAL	ALT_EXON	LONGER_EXON_HIGHER	1	0.5%
SHARED_LOCAL	ALT_EXON	LONGER_EXON_LOWER	0	0.0%
SHARED_LOCAL	ALT_POLYA_SITE	PROXIMAL_HIGHER	20	10.2%
SHARED_LOCAL	ALT_POLYA_SITE	DISTAL_HIGHER	19	9.7%
SHARED_LOCAL	RETAINED_INTRON	INTRON_RETAINED_HIGHER	3	1.5%
SHARED_LOCAL	RETAINED_INTRON	INTRON_RETAINED_LOWER	0	0.0%
SHARED_LOCAL	IPA_EXTENSION	EXTENDS_TO_PA_HIGHER	8	4.1%
SHARED_LOCAL	IPA_EXTENSION	EXTENDS_TO_PA_LOWER	0	0.0%
SHARED_LOCAL	IPA_NO_EXTENSION	SPLICED_FORM_HIGHER	0	0.0%
SHARED_LOCAL	IPA_NO_EXTENSION	POLYA_FORM_HIGHER	0	0.0%
SHARED_DISTAL	DISTAL_APA	PROXIMAL_HIGHER	0	0.0%
SHARED_DISTAL	DISTAL_APA	DISTAL_HIGHER	0	0.0%
SHARED_DISTAL	DISTAL_SPLICING	SAME_3PRIME_END	2	1.0%
UNEXPLAINABLE	FIVE_PRIME_UNCERTAIN	—	0	0.0%
UNEXPLAINABLE	INTRON_READ_ARTIFACT	—	0	0.0%
UNEXPLAINABLE	NO_MODEL	—	0	0.0%
UNEXPLAINABLE	UNRESOLVED	—	0	0.0%

Level 1-2: bucket definitions

PRIVATE

The modified base exists (exonically) ONLY in the more-modified fragmentform — it is intronic or absent in the other form's assembled model, so the difference is structurally trivial: the base is physically missing from one isoform. (Called from transcript-model coordinates, guarded against the 5' blind spot.)

SHARED_LOCAL

The base is exonic in BOTH fragmentforms, and the structural difference between them lies ON / adjacent to the base's own exon — a shifted splice site (ALT_DONOR/ALT_ACCEPTOR), an alternative poly(A) site, a retained intron, an intronic-polyadenylation extension, or a nearby differential junction. This holds even when the base's exon is the terminal exon in one form.

SHARED_DISTAL

The base is exonic in both, in an IDENTICAL local context; the fragmentforms differ only ELSEWHERE in the transcript, so the modification difference tracks isoform identity rather than any feature at the base.

UNEXPLAINABLE

The difference cannot be attributed to structure: the 5' blind spot (direct-RNA truncation), an intron / soft-clip read artifact, or too few covered isoforms to compare.

Level 2: event definitions (all events)

SKIPPED_EXON

The base sits in a cassette exon spliced INTO the higher form and OUT of the lower form.

INTRONIC_POLYA

The base is in the retained-intron 3'UTR of an intronic-polyadenylation isoform — it does not exist in the spliced full-length form.

THREE_PRIME_EXTENSION

The base is in a 3'UTR EXTENSION of a SHARED last exon (the two forms share the last-exon acceptor but the higher form extends to a distal poly(A) site) — tandem APA, not an alternative last exon. Distinct from ALT_LAST_EXON, which is a mutually-exclusive (different-acceptor) last exon.

ALT_LAST_EXON

The base is in a mutually-exclusive alternative last exon (DIFFERENT acceptor) used by the higher form.

ALT_DONOR

The base's exon uses a different 5' splice site (donor) in the two forms, changing that exon's length (structural_delta_nt).

ALT_ACCEPTOR

The base's exon uses a different 3' splice site (acceptor) in the two forms, changing that exon's length (structural_delta_nt).

ALT_EXON

The base's exon uses a DIFFERENT 5' splice site (donor) AND a different 3' splice site (acceptor) between the two forms — a mutually-exclusive alternative exon that shares neither boundary, not a single shifted splice site.

ALT_POLYA_SITE

Same last exon, different poly(A) cleavage sites (tandem APA); the base is in the shared body of the last exon.

RETAINED_INTRON

The base's own exon spans an intron in one form that the other form splices out — intron retention right at the base (both forms share the same 3' end).

IPA_EXTENSION

The base's own exon is EXTENDED in one form because it reads into the downstream intron and polyadenylates there (intronic polyadenylation WITH extension), while the other form splices on to a more distal 3' end. The base sits in the intron-derived extension. Contrast IPA_NO_EXTENSION, where the poly(A) form cleaves right at the splice donor with no extension.

IPA_NO_EXTENSION

Intronic polyadenylation with NO exon extension: one form SPLICES at the base's exon boundary while the other POLYADENYLATES at that SAME coordinate (no reading into the intron). The base's exon is identical in both forms and lies within the EJC/cleavage footprint (~24 nt) of the decision; whether it is spliced-and-EJC-marked or cleaved-and-polyadenylated tracks its modification. Contrast IPA_EXTENSION, where the poly(A) form's exon extends PAST the donor into the intron and the base sits in that extension.

DISTAL_APA

The base's local exon is identical in both forms, but the forms end at DIFFERENT poly(A) sites (a distal alternative-polyadenylation choice); methylation tracks which 3' end the isoform uses.

DISTAL_SPLICING

The base's local exon is identical AND the forms share the same 3' end; they differ only in an internal / 5' splicing choice ELSEWHERE, so methylation tracks that distal splicing decision.

FIVE_PRIME_UNCERTAIN

The base falls in (or 5' of) the 5'-most exon, where direct-RNA 5' truncation makes the assembled model unreliable — no confident structural call.

INTRON_READ_ARTIFACT

The more-modified form does not structurally contain the base (it is intronic there) — likely an intron / soft-clip read, not a real exonic modification.

NO_MODEL

Fewer than two adequately-covered fragmentforms at the site, so there is nothing to compare.

UNRESOLVED

Passes the filters but fits no structural rule.

Level 3: direction definitions (all directions)

IN_INCLUDED_EXON

PRIVATE / SKIPPED_EXON: the base lives in a cassette exon spliced IN to the carrier form and spliced OUT of the other.

IN_IPA_ISOFORM

PRIVATE / INTRONIC_POLYA: the base lives in the retained-intron 3'UTR of the carrier's intronic-polyadenylation isoform; it does not exist in the spliced form.

IN_DISTAL_EXTENSION

PRIVATE / THREE_PRIME_EXTENSION: the base lives in the carrier's 3'UTR EXTENSION to a DISTAL poly(A) site on a SHARED last exon (tandem APA); absent from the shorter form.

IN_ALT_LAST_EXON

PRIVATE / ALT_LAST_EXON: the base lives in a mutually-exclusive alternative last exon (different acceptor) used only by the carrier form.

LONGER_EXON_HIGHER

The more-modified fragmentform carries the LONGER version of the base's exon (structural_delta_nt = the length difference).

LONGER_EXON_LOWER

The LESS-modified fragmentform carries the longer version of the base's exon (i.e. the more-modified form has the shorter one).

PROXIMAL_HIGHER

The more-modified fragmentform uses the PROXIMAL (upstream, shorter-3'UTR) poly(A) site.

DISTAL_HIGHER

The more-modified fragmentform uses the DISTAL (downstream, longer-3'UTR) poly(A) site.

INTRON_RETAINED_HIGHER

The more-modified fragmentform is the one that RETAINS the intron (its exon spans it); the other splices it out.

INTRON_RETAINED_LOWER

The LESS-modified fragmentform is the one that retains the intron; the more-modified one splices it out.

EXTENDS_TO_PA_HIGHER

The more-modified fragmentform is the one that reads into the intron and polyadenylates there (its exon is the extended, intronic-polyadenylation exon).

EXTENDS_TO_PA_LOWER

The LESS-modified fragmentform is the one that reads into the intron and polyadenylates there.

SPLICED_FORM_HIGHER

IPA_NO_EXTENSION: the more-modified fragmentform is the one that SPLICES at the base's exon boundary (base retained upstream of a deposited EJC), where the other form polyadenylates.

POLYA_FORM_HIGHER

IPA_NO_EXTENSION: the more-modified fragmentform is the one that POLYADENYLATES (cleaves/terminates) at the base's exon boundary, where the other form splices onward.

SAME_3PRIME_END

Both fragmentforms end at the same 3' position (same last exon / poly(A) site); the more-modified one differs only in an internal or 5' splicing choice further upstream.

↗

SHARED_LOCAL — 52 sites

The base is exonic in BOTH fragmentforms, and the structural difference between them lies ON / adjacent to the base's own exon — a shifted splice site (ALT_DONOR/ALT_ACCEPTOR), an alternative poly(A) site, a retained intron, an intronic-polyadenylation extension, or a nearby differential junction. This holds even when the base's exon is the terminal exon in one form.

ALT_DONOR — 0 sites

The base's exon uses a different 5' splice site (donor) in the two forms, changing that exon's length (structural_delta_nt).

LONGER_EXON_HIGHER — 0 sites

The more-modified fragmentform carries the LONGER version of the base's exon (structural_delta_nt = the length difference).

No sites of this direction in this run.

LONGER_EXON_LOWER — 0 sites

The LESS-modified fragmentform carries the longer version of the base's exon (i.e. the more-modified form has the shorter one).

No sites of this direction in this run.

ALT_ACCEPTOR — 1 site

The base's exon uses a different 3' splice site (acceptor) in the two forms, changing that exon's length (structural_delta_nt).

LONGER_EXON_HIGHER — 1 site

The more-modified fragmentform carries the LONGER version of the base's exon (structural_delta_nt = the length difference).

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	PCID2	a (m6A)	chr13	113177975	-	LONGER_EXON_HIGHER	32.0	7	TANDEM_APA	0.8611	36	5	TANDEM_APA	0.0	76	0.8611	T4_NEAR_BINARY	HI_HYPER	0.8541	9.352000e-194

Top 1 isoform architecture map(s) — exon/intron tracks, site marked

↗

Top 1 site figure(s) — per-sample stoichiometry & pooled coverage

↗

LONGER_EXON_LOWER — 0 sites

The LESS-modified fragmentform carries the longer version of the base's exon (i.e. the more-modified form has the shorter one).

No sites of this direction in this run.

ALT_EXON — 1 site

The base's exon uses a DIFFERENT 5' splice site (donor) AND a different 3' splice site (acceptor) between the two forms — a mutually-exclusive alternative exon that shares neither boundary, not a single shifted splice site.

LONGER_EXON_HIGHER — 1 site

The more-modified fragmentform carries the LONGER version of the base's exon (structural_delta_nt = the length difference).

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	ASGR2	a (m6A)	chr17	7114229	-	LONGER_EXON_HIGHER	145.0	5	TANDEM_APA	0.2697	178	13	TANDEM_APA	0.0263	38	0.2433	T1_MARGINAL	HI_HYPO	0.2601	0.000531

Top 1 isoform architecture map(s) — exon/intron tracks, site marked

↗

Top 1 site figure(s) — per-sample stoichiometry & pooled coverage

↗

LONGER_EXON_LOWER — 0 sites

The LESS-modified fragmentform carries the longer version of the base's exon (i.e. the more-modified form has the shorter one).

No sites of this direction in this run.

ALT_POLYA_SITE — 39 sites

Same last exon, different poly(A) cleavage sites (tandem APA); the base is in the shared body of the last exon.

PROXIMAL_HIGHER — 20 sites

The more-modified fragmentform uses the PROXIMAL (upstream, shorter-3'UTR) poly(A) site.

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	MXD1	a (m6A)	chr2	69938439	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.9310	58	2	REFERENCE	0.1754	114	0.7556	T4_NEAR_BINARY	HI_HYPER	0.6388	4.139000e-10
2	MXD1	a (m6A)	chr2	69940417	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.8261	46	2	REFERENCE	0.2136	103	0.6125	T3_STRONG	HI_HYPER	0.4397	5.311000e-04
3	MXD1	a (m6A)	chr2	69938415	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.7879	33	2	REFERENCE	0.2069	87	0.5810	T3_STRONG	HI_HYPER	0.4582	7.435000e-03
4	MXD1	a (m6A)	chr2	69939760	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.9474	57	2	REFERENCE	0.4444	117	0.5029	T3_STRONG	HI_HYPER	0.2450	3.083000e-02
5	MXD1	a (m6A)	chr2	69938897	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.6226	53	2	REFERENCE	0.1273	110	0.4954	T2_MODERATE	HI_INTERMED	0.4705	2.013000e-08
6	MXD1	a (m6A)	chr2	69938700	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.7091	55	2	REFERENCE	0.2212	113	0.4879	T2_MODERATE	HI_HYPER	0.3015	1.933000e-02
7	MXD1	a (m6A)	chr2	69939779	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.7097	62	2	REFERENCE	0.2522	115	0.4575	T2_MODERATE	HI_HYPER	0.3869	2.328000e-03
8	MXD1	a (m6A)	chr2	69938505	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.9375	64	2	REFERENCE	0.5000	118	0.4375	T2_MODERATE	HI_HYPER	0.3254	6.425000e-03
9	MXD1	a (m6A)	chr2	69938378	+	PROXIMAL_HIGHER	2304.0	3	TANDEM_APA	0.4737	57	2	REFERENCE	0.0625	112	0.4112	T2_MODERATE	HI_INTERMED	0.3758	8.797000e-03
10	VGLL4	a (m6A)	chr3	11558059	-	PROXIMAL_HIGHER	1910.0	2	TANDEM_APA	0.5294	51	3	FULL_LENGTH	0.1190	42	0.4104	T2_MODERATE	HI_INTERMED	0.3763	2.780000e-04

Top 10 isoform architecture map(s) — exon/intron tracks, site marked

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

Top 10 site figure(s) — per-sample stoichiometry & pooled coverage

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

DISTAL_HIGHER — 19 sites

The more-modified fragmentform uses the DISTAL (downstream, longer-3'UTR) poly(A) site.

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	ATG7	a (m6A)	chr3	11555082	+	DISTAL_HIGHER	2582.0	3	REFERENCE	0.9913	115	4	TANDEM_APA	0.0435	69	0.9478	T4_NEAR_BINARY	HI_HYPER	0.9409	5.311000e-04
2	RANBP3	a (m6A)	chr19	5917337	-	DISTAL_HIGHER	1182.0	9	FULL_LENGTH	0.8387	31	2	TANDEM_APA	0.1730	555	0.6657	T3_STRONG	HI_HYPER	0.6734	3.017000e-119
3	ASGR2	a (m6A)	chr17	7101415	-	DISTAL_HIGHER	1157.0	29	FULL_LENGTH	0.8837	43	11	TANDEM_APA	0.4688	32	0.4150	T2_MODERATE	HI_HYPER	0.4754	2.546000e-02
4	ATG7	a (m6A)	chr3	11554911	+	DISTAL_HIGHER	2582.0	3	REFERENCE	0.7573	103	1	TANDEM_APA	0.4178	1424	0.3394	T2_MODERATE	HI_HYPER	0.3905	1.505000e-07
5	C8A	a (m6A)	chr1	56917930	+	DISTAL_HIGHER	214.0	5	REFERENCE	0.8485	66	1	TANDEM_APA	0.5430	372	0.3055	T2_MODERATE	HI_HYPER	0.3086	2.975000e-08
6	RANBP3	a (m6A)	chr19	5917420	-	DISTAL_HIGHER	1182.0	9	FULL_LENGTH	0.9091	33	2	TANDEM_APA	0.6061	495	0.3030	T2_MODERATE	HI_HYPER	0.2957	2.966000e-17
7	RANBP3	a (m6A)	chr19	5917449	-	DISTAL_HIGHER	873.0	7	FULL_LENGTH	0.4340	53	6	TANDEM_APA	0.1346	52	0.2993	T2_MODERATE	HI_INTERMED	0.3383	2.428000e-04
8	IP6K2	a (m6A)	chr3	48693802	-	DISTAL_HIGHER	243.0	16	IPA	0.4255	47	20	IPA	0.1481	27	0.2774	T2_MODERATE	HI_INTERMED	0.3005	6.790000e-06
9	RANBP3	a (m6A)	chr19	5917441	-	DISTAL_HIGHER	1182.0	10	FULL_LENGTH	0.4118	34	4	TANDEM_APA	0.1493	268	0.2625	T2_MODERATE	HI_INTERMED	0.2490	8.228000e-08
10	ALCAM	a (m6A)	chr3	105574527	+	DISTAL_HIGHER	466.0	4	TANDEM_APA	0.2708	48	6	TANDEM_APA	0.0526	38	0.2182	T1_MARGINAL	HI_HYPO	0.2509	4.537000e-02

Top 10 isoform architecture map(s) — exon/intron tracks, site marked

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

Top 10 site figure(s) — per-sample stoichiometry & pooled coverage

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

RETAINED_INTRON — 3 sites

The base's own exon spans an intron in one form that the other form splices out — intron retention right at the base (both forms share the same 3' end).

INTRON_RETAINED_HIGHER — 3 sites

The more-modified fragmentform is the one that RETAINS the intron (its exon spans it); the other splices it out.

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	ASGR1	a (m6A)	chr17	7173773	-	INTRON_RETAINED_HIGHER	234.0	7	FULL_LENGTH	0.7647	51	6	FULL_LENGTH	0.2400	75	0.5247	T3_STRONG	HI_HYPER	0.4964	1.256000e-05
2	PCID2	a (m6A)	chr13	113177947	-	INTRON_RETAINED_HIGHER	451.0	6	FULL_LENGTH	0.5588	68	5	TANDEM_APA	0.2078	77	0.3510	T2_MODERATE	HI_INTERMED	0.3845	1.457000e-07
3	ASGR1	a (m6A)	chr17	7173978	-	INTRON_RETAINED_HIGHER	530.0	7	FULL_LENGTH	0.3043	46	6	FULL_LENGTH	0.0390	77	0.2654	T2_MODERATE	HI_HYPO	0.3042	5.311000e-04

Top 3 isoform architecture map(s) — exon/intron tracks, site marked

↗

↗

↗

Top 3 site figure(s) — per-sample stoichiometry & pooled coverage

↗

↗

↗

INTRON_RETAINED_LOWER — 0 sites

The LESS-modified fragmentform is the one that retains the intron; the more-modified one splices it out.

No sites of this direction in this run.

IPA_EXTENSION — 8 sites

The base's own exon is EXTENDED in one form because it reads into the downstream intron and polyadenylates there (intronic polyadenylation WITH extension), while the other form splices on to a more distal 3' end. The base sits in the intron-derived extension. Contrast IPA_NO_EXTENSION, where the poly(A) form cleaves right at the splice donor with no extension.

EXTENDS_TO_PA_HIGHER — 8 sites

The more-modified fragmentform is the one that reads into the intron and polyadenylates there (its exon is the extended, intronic-polyadenylation exon).

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	C8A	a (m6A)	chr1	56886077	+	EXTENDS_TO_PA_HIGHER	1345.0	6	IPA	0.9348	46	5	REFERENCE	0.0484	62	0.8864	T4_NEAR_BINARY	HI_HYPER	0.8895	1.098000e-61
2	SERAC1	a (m6A)	chr6	158114798	-	EXTENDS_TO_PA_HIGHER	315.0	8	IPA	0.7949	39	1	REFERENCE	0.0452	155	0.7497	T3_STRONG	HI_HYPER	0.7606	9.774000e-52
3	ALCAM	a (m6A)	chr3	105524469	+	EXTENDS_TO_PA_HIGHER	452.0	3	IPA	0.5079	63	4	TANDEM_APA	0.0000	40	0.5079	T3_STRONG	HI_INTERMED	0.5181	5.311000e-04
4	ALCAM	a (m6A)	chr3	105524376	+	EXTENDS_TO_PA_HIGHER	452.0	3	IPA	0.4667	60	6	TANDEM_APA	0.0385	26	0.4282	T2_MODERATE	HI_INTERMED	0.4382	5.311000e-04
5	ALCAM	a (m6A)	chr3	105524462	+	EXTENDS_TO_PA_HIGHER	452.0	3	IPA	0.3103	58	2	FULL_LENGTH	0.0000	107	0.3103	T2_MODERATE	HI_HYPO	0.3187	5.311000e-04
6	C8A	a (m6A)	chr1	56885962	+	EXTENDS_TO_PA_HIGHER	1345.0	6	IPA	0.1957	46	5	REFERENCE	0.0000	63	0.1957	T1_MARGINAL	HI_HYPO	0.1936	5.311000e-04
7	C8A	a (m6A)	chr1	56885990	+	EXTENDS_TO_PA_HIGHER	1345.0	6	IPA	0.1154	52	5	REFERENCE	0.0000	61	0.1154	T1_MARGINAL	HI_HYPO	0.1127	1.859000e-03
8	C8A	a (m6A)	chr1	56883640	+	EXTENDS_TO_PA_HIGHER	101.0	3	IPA	0.1014	138	5	REFERENCE	0.0000	56	0.1014	T1_MARGINAL	HI_HYPO	0.1005	6.373000e-03

Top 8 isoform architecture map(s) — exon/intron tracks, site marked

↗

↗

↗

↗

↗

↗

↗

↗

Top 8 site figure(s) — per-sample stoichiometry & pooled coverage

↗

↗

↗

↗

↗

↗

↗

↗

EXTENDS_TO_PA_LOWER — 0 sites

The LESS-modified fragmentform is the one that reads into the intron and polyadenylates there.

No sites of this direction in this run.

IPA_NO_EXTENSION — 0 sites

Intronic polyadenylation with NO exon extension: one form SPLICES at the base's exon boundary while the other POLYADENYLATES at that SAME coordinate (no reading into the intron). The base's exon is identical in both forms and lies within the EJC/cleavage footprint (~24 nt) of the decision; whether it is spliced-and-EJC-marked or cleaved-and-polyadenylated tracks its modification. Contrast IPA_EXTENSION, where the poly(A) form's exon extends PAST the donor into the intron and the base sits in that extension.

SPLICED_FORM_HIGHER — 0 sites

IPA_NO_EXTENSION: the more-modified fragmentform is the one that SPLICES at the base's exon boundary (base retained upstream of a deposited EJC), where the other form polyadenylates.

No sites of this direction in this run.

POLYA_FORM_HIGHER — 0 sites

IPA_NO_EXTENSION: the more-modified fragmentform is the one that POLYADENYLATES (cleaves/terminates) at the base's exon boundary, where the other form splices onward.

No sites of this direction in this run.

SHARED_DISTAL — 2 sites

The base is exonic in both, in an IDENTICAL local context; the fragmentforms differ only ELSEWHERE in the transcript, so the modification difference tracks isoform identity rather than any feature at the base.

DISTAL_APA — 0 sites

The base's local exon is identical in both forms, but the forms end at DIFFERENT poly(A) sites (a distal alternative-polyadenylation choice); methylation tracks which 3' end the isoform uses.

PROXIMAL_HIGHER — 0 sites

The more-modified fragmentform uses the PROXIMAL (upstream, shorter-3'UTR) poly(A) site.

No sites of this direction in this run.

DISTAL_HIGHER — 0 sites

The more-modified fragmentform uses the DISTAL (downstream, longer-3'UTR) poly(A) site.

No sites of this direction in this run.

DISTAL_SPLICING — 2 sites

The base's local exon is identical AND the forms share the same 3' end; they differ only in an internal / 5' splicing choice ELSEWHERE, so methylation tracks that distal splicing decision.

SAME_3PRIME_END — 2 sites

Both fragmentforms end at the same 3' position (same last exon / poly(A) site); the more-modified one differs only in an internal or 5' splicing choice further upstream.

Column definitions

example_rank

1-based rank of the site within its leaf (bucket / event / direction) by hi_lo_delta, then FDR, then the shallower form's coverage. The rankNN__ example figures follow this rank, so row N of the table is figure N.

gene_name

Gene label assigned to the reported row or site.

mod_code

Modification code reported by modkit or downstream aggregation.

chrom

Reference chromosome or contig containing the reported feature.

start0

0-based inclusive start coordinate of the reported modification site.

strand

Genomic strand of the reported feature.

hi_ZN

ZN partition index of the higher-stoichiometry isoform in the classified contrast.

hi_arch

3' architecture of the higher-stoichiometry isoform versus the gene's longest-3'UTR anchor (IPA / TANDEM_APA / FULL_LENGTH / DISTAL_EXT / REFERENCE / AMBIGUOUS).

hi_frac

Pooled modified fraction (stoichiometry) of the higher isoform at the site.

hi_cov

Pooled valid coverage (reads) of the higher isoform at the site.

lo_ZN

ZN partition index of the lower-stoichiometry isoform in the classified contrast.

lo_arch

3' architecture of the lower-stoichiometry isoform versus the gene's longest-3'UTR anchor.

lo_frac

Pooled modified fraction (stoichiometry) of the lower isoform at the site.

lo_cov

Pooled valid coverage (reads) of the lower isoform at the site.

hi_lo_delta

hi_frac minus lo_frac: the pooled stoichiometry difference between the two fragmentforms this classification compares. This is the contrast the structural event explains; the table and the example figures are ranked by it (then FDR, then coverage).

stoich_tier

Magnitude of the between-isoform stoichiometry gap: T1_MARGINAL (.10–.25) / T2_MODERATE / T3_STRONG / T4_NEAR_BINARY (≥.75).

hi_stoich_level

Absolute modification level of the favored fragmentform: HI_HYPER (≥.66) / HI_INTERMED / HI_HYPO (<.33).

effect_max_abs_frac_diff

Maximum absolute difference in modified fraction across tested fragmentform partitions, Mantel-Haenszel-weighted over EVERY sample covering both forms (the sample-stratified effect that matches the primary p-value, NOT the sample-pooled fraction). The old sample-pooled value is kept separately as effect_max_abs_frac_diff_pooled.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

example_rank	gene_name	mod_code	chrom	start0	strand	direction	structural_delta_nt	hi_ZN	hi_arch	hi_frac	hi_cov	lo_ZN	lo_arch	lo_frac	lo_cov	hi_lo_delta	stoich_tier	hi_stoich_level	effect_max_abs_frac_diff	p_adj_bh
1	ASGR1	a (m6A)	chr17	7173830	-	SAME_3PRIME_END	NaN	10	FULL_LENGTH	0.2000	20	9	FULL_LENGTH	0.0000	24	0.2000	T1_MARGINAL	HI_HYPO	0.1915	0.040550
2	RANBP3	a (m6A)	chr19	5917474	-	SAME_3PRIME_END	NaN	9	FULL_LENGTH	0.2308	39	7	FULL_LENGTH	0.0385	52	0.1923	T1_MARGINAL	HI_HYPO	0.2162	0.001439

Top 2 isoform architecture map(s) — exon/intron tracks, site marked

↗

↗

Top 2 site figure(s) — per-sample stoichiometry & pooled coverage

↗

↗

Every fragmentform's mature (spliced) mRNA is scanned for sequence-defined cis-elements anchored to its 3' end (PAS, AU-rich element, CPE, GU-rich element, rG4), start codon (Kozak, uORF, 5'TOP, m6Am) and stop codon (readthrough context). Each element is then joined to EVERY overlapping modification with no mod-code bias. Only mature-mRNA (exonic) elements are scanned; 5'-anchored elements (Kozak/uORF/TOP/m6Am) sit where direct-RNA read coverage is sparsest, so their modification calls are coverage-limited.

↗

Summary Across All Sequence Elements

• 1,175 element instances across 9 element types
• 332 carry ≥1 modification — mod codes seen: 17596, 17802, 19227, 19228, 19229, 69426, a, m (no code is filtered out)

Direct per-read poly(A) tail-length estimates from the dorado basecaller (pt:i tag), grouped by fragmentform (isoform). Tail length is mechanistically tied to m6A, RNA stability, and translation, so it is reported as a first-class readout: its distribution across fragmentforms, whether the fragmentforms of a gene differ in tail length, and whether modification at a site tracks with a read's tail length.

↗

Summary Across All Poly(A) Tail Measurements

• 131 fragmentforms with a tail-length distribution; overall median 109 nt
• 11 genes with differential tail length between their fragmentforms (FDR<0.05)
• 1 modification sites where tail length differs by modification state WITHIN (sample, fragmentform) strata (sample-stratified van Elteren test, FDR<0.05; confounded sites excluded)

Differential tail length between fragmentforms of a gene

Column definitions

gene_name

Gene label assigned to the reported row or site.

n_fragmentforms_tested

Number of the gene's fragmentforms with enough tailed reads to compare.

n_reads

Total reads contributing to the reported test.

effect_median_range_nt

Spread (nt) of per-fragmentform median tail lengths within a gene — how differently its isoforms are tailed.

min_median_tail

Smallest per-fragmentform median tail length (nt) among the gene's fragmentforms.

max_median_tail

Largest per-fragmentform median tail length (nt) among the gene's fragmentforms.

test_name

Statistical test used: the sample-stratified van Elteren test (stratified Wilcoxon / Kruskal-Wallis). The sample-pooled Mann-Whitney U / Kruskal-Wallis statistic is retained separately as test_name_pooled / p_value_pooled.

p_value

Nominal p-value from the reported hypothesis test.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

gene_name	n_fragmentforms_tested	n_reads	effect_median_range_nt	min_median_tail	max_median_tail	test_name	p_value	p_adj_bh
ASGR2	32	24757	131.86	73.0	147.5	van_elteren_stratified_kw	1.154365e-34	1.500675e-33
RIOK3	8	2103	61.25	128.0	162.0	van_elteren_stratified_kw	4.886421e-15	3.176174e-14
SERAC1	8	638	94.24	123.0	197.5	van_elteren_stratified_kw	1.851207e-12	8.021898e-12
RANBP3	10	2495	84.53	98.0	160.0	van_elteren_stratified_kw	3.867428e-10	1.256914e-09
ASGR1	12	9274	59.72	76.0	108.0	van_elteren_stratified_kw	6.519172e-08	1.694985e-07
ALCAM	6	1208	66.43	110.0	154.5	van_elteren_stratified_kw	1.200105e-07	2.600227e-07
MXD1	4	549	49.82	128.0	174.0	van_elteren_stratified_kw	2.948305e-07	5.475424e-07
IP6K2	20	4915	93.77	88.0	125.0	van_elteren_stratified_kw	3.272203e-06	5.317331e-06
ATG7	6	1979	65.68	100.5	121.5	van_elteren_stratified_kw	8.402891e-04	1.213751e-03
PCID2	8	2926	113.16	102.0	137.0	van_elteren_stratified_kw	1.488638e-03	1.935230e-03
C8A	7	1250	63.19	95.5	120.0	van_elteren_stratified_kw	2.863655e-02	3.384320e-02
NHSL1	4	296	62.35	134.0	151.0	van_elteren_stratified_kw	2.432413e-01	2.635114e-01
VGLL4	4	996	43.25	122.0	136.0	van_elteren_stratified_kw	5.366938e-01	5.366938e-01

Top 10 per-gene tail-distribution figure(s) — tail length by fragmentform

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

Poly(A) tail length vs modification state

Column definitions

gene_name

Gene label assigned to the reported row or site.

mod_site_id

Canonical modification-site identifier in `chrom:start-end:strand:mod` format.

target_mod_code

Modification code tested at the reported site.

n_modified

Reads classified as modified for the target mod code.

n_unmodified

Number of reads unmodified at the target site.

median_tail_modified

Median tail length (nt) of reads modified at the target site.

median_tail_unmodified

Median tail length (nt) of reads unmodified at the target site.

effect_median_diff_nt

Median tail of modified reads minus unmodified reads at a site (negative = modification associates with shorter tails).

test_name

Statistical test used: the sample-stratified van Elteren test (stratified Wilcoxon / Kruskal-Wallis). The sample-pooled Mann-Whitney U / Kruskal-Wallis statistic is retained separately as test_name_pooled / p_value_pooled.

p_value

Nominal p-value from the reported hypothesis test.

p_adj_bh

Benjamini-Hochberg false-discovery-rate adjusted p-value.

n_strata_informative

Number of strata (by sample — by sample x fragmentform for the tail-length tests) that carried enough within-stratum information (both outcome states / >=2 covered rows) to contribute to the sample-stratified test. The primary p-value uses the stratified test only when this is >=2; with fewer strata it falls back to the exact pooled test, because a single-stratum statistic is an anti-conservative asymptotic approximation.

n_forms_comparable

Number of fragmentforms with >=3 reads in BOTH the modified and unmodified state (the forms in which a within-fragmentform tail comparison is possible).

n_forms_concordant

Of the comparable fragmentforms, how many shift in the SAME direction as (and with at least a third of) the reported within-stratum effect. When this is a minority, the site is flagged tailmod_confounded and excluded from the headline count.

tailmod_confounded

True when fewer than half of the comparable fragmentforms reproduce the effect within-form — i.e. the modified-vs-unmodified tail gap is likely the modification proxying for a differently-tailed isoform rather than a within-isoform effect. Confounded sites are excluded from the headline significant count.

effect_median_diff_nt_pooled

The OLD pooled median(mod)-median(unmod) over all reads at the site, ignoring sample/fragmentform structure. Retained for transparency; can be badly inflated by the isoform confound — do not rank on it.

p_value_pooled

The OLD sample-POOLED p-value (all reads pooled into one table, ignoring sample structure). Retained for transparency; it can be Simpson's-paradox inflated, so the primary p_value is sample-stratified. Do not rank on it.

gene_name	mod_site_id	target_mod_code	n_modified	n_unmodified	median_tail_modified	median_tail_unmodified	effect_median_diff_nt	test_name	p_value	p_adj_bh	n_strata_informative	n_forms_comparable	n_forms_concordant	tailmod_confounded	effect_median_diff_nt_pooled	p_value_pooled
PCID2	chr13:113177838-113177839:-:17802	17802 (pseudouridine)	116	2646	76.0	119.0	-36.41	van_elteren_stratified_wilcoxon	4.694429e-10	2.525603e-07	19	5	4	False	-43.0	4.553074e-10
ALCAM	chr3:105576724-105576725:+:a	a (m6A)	29	972	84.0	119.0	-34.74	van_elteren_stratified_wilcoxon	9.808703e-04	2.638541e-01	7	2	2	False	-35.0	1.317212e-03
PCID2	chr13:113179929-113179930:-:19228	19228 (2'-O-methylcytidine, Cm)	84	2699	98.5	116.0	-20.89	van_elteren_stratified_wilcoxon	2.162687e-03	3.878418e-01	13	4	2	True	-17.5	6.957694e-02
SERAC1	chr6:158110919-158110920:-:19227	19227 (2'-O-methyluridine, Um)	21	418	129.0	135.0	-65.88	van_elteren_stratified_wilcoxon	3.032704e-02	4.535605e-01	3	4	3	False	-6.0	4.328035e-01
ALCAM	chr3:105574988-105574989:+:a	a (m6A)	37	1092	146.0	118.0	64.12	van_elteren_stratified_wilcoxon	8.860629e-03	4.535605e-01	7	3	2	False	28.0	5.723250e-02
NHSL1	chr6:138422562-138422563:-:m	m (5mC)	11	165	118.0	151.0	-62.84	van_elteren_stratified_wilcoxon	6.875922e-03	4.535605e-01	4	2	1	True	-33.0	3.192715e-02
NHSL1	chr6:138433212-138433213:-:a	a (m6A)	50	29	121.5	198.0	-58.17	van_elteren_stratified_wilcoxon	6.611794e-03	4.535605e-01	6	2	1	True	-76.5	4.260362e-03
NHSL1	chr6:138433217-138433218:-:m	m (5mC)	72	11	139.0	198.0	-48.85	van_elteren_stratified_wilcoxon	1.221090e-02	4.535605e-01	3	1	1	False	-59.0	7.849729e-02
ALCAM	chr3:105575054-105575055:+:17802	17802 (pseudouridine)	17	1102	82.0	119.0	-45.03	van_elteren_stratified_wilcoxon	1.606869e-02	4.535605e-01	4	1	1	False	-37.0	8.888935e-03
PCID2	chr13:113185534-113185535:-:17596	17596 (inosine)	19	2445	66.0	114.0	-42.46	van_elteren_stratified_wilcoxon	1.267511e-02	4.535605e-01	5	3	2	False	-48.0	1.746495e-02
ALCAM	chr3:105575672-105575673:+:a	a (m6A)	36	1037	157.0	118.0	36.98	van_elteren_stratified_wilcoxon	2.589492e-02	4.535605e-01	7	2	2	False	39.0	1.542594e-02
SERAC1	chr6:158110602-158110603:-:17802	17802 (pseudouridine)	19	421	124.0	135.0	-34.78	van_elteren_stratified_wilcoxon	3.249502e-02	4.535605e-01	5	2	1	True	-11.0	2.876344e-01
ASGR2	chr17:7100327-7100328:-:a	a (m6A)	18	119	85.5	109.0	-33.92	van_elteren_stratified_wilcoxon	2.565176e-02	4.535605e-01	5	2	1	True	-23.5	1.048861e-01
NHSL1	chr6:138431040-138431041:-:a	a (m6A)	123	68	155.0	134.0	30.61	van_elteren_stratified_wilcoxon	1.820910e-02	4.535605e-01	15	3	2	False	21.0	2.945101e-02
ALCAM	chr3:105520069-105520070:+:17802	17802 (pseudouridine)	22	806	103.5	122.5	-28.85	van_elteren_stratified_wilcoxon	3.078759e-02	4.535605e-01	4	1	1	False	-19.0	2.150280e-02
ASGR2	chr17:7102133-7102134:-:19228	19228 (2'-O-methylcytidine, Cm)	77	24323	114.0	92.0	26.48	van_elteren_stratified_wilcoxon	4.653569e-03	4.535605e-01	14	5	4	False	22.0	9.798189e-03
NHSL1	chr6:138431582-138431583:-:17596	17596 (inosine)	72	84	141.0	155.5	-23.70	van_elteren_stratified_wilcoxon	1.222315e-02	4.535605e-01	16	3	3	False	-14.5	2.878498e-02
PCID2	chr13:113177999-113178000:-:a	a (m6A)	59	2197	101.0	118.0	-23.47	van_elteren_stratified_wilcoxon	1.420185e-02	4.535605e-01	11	3	2	False	-17.0	1.605854e-02
ALCAM	chr3:105524543-105524544:+:a	a (m6A)	31	39	167.0	144.0	22.64	van_elteren_stratified_wilcoxon	7.795652e-03	4.535605e-01	3	1	1	False	23.0	4.537920e-03
ALCAM	chr3:105576161-105576162:+:19227	19227 (2'-O-methyluridine, Um)	107	933	104.0	119.0	-20.26	van_elteren_stratified_wilcoxon	2.501843e-02	4.535605e-01	10	3	2	False	-15.0	4.517431e-02

Top 10 per-site figure(s) — modified vs unmodified tail length

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

Replicate-aware comparisons between conditions, from the samplesheet's `condition` column. Counts (modification and isoform/APA/junction usage) use a beta-binomial likelihood-ratio test with dispersion shrinkage across features; poly(A) tail length is continuous and is compared with Welch's t-test across per-replicate medians. Reads are never pooled across replicates -- the biological unit is the replicate, and pooling would make trivial differences look overwhelmingly significant. Differential modification is reported two ways: a SITE-LEVEL table (counts summed across a site's fragmentforms -- the headline change at the modified position), and, when the per-transcript table is enabled, a PER-FRAGMENTFORM table that tests each ZN transcript partition on its own to resolve which fragmentform carries the change. Each analysis shows a top-by-effect figure with the individual replicate values, so effect size and replicate agreement are visible together. CAVEAT: a modification site that comes out 0% in one condition and X% in the other (an entire arm all-unmodified) has an ANTI-conservative raw p at the mu=0 boundary -- ~4-10x optimistic -- so for those most-extreme '0% -> X%' sites trust the FDR-adjusted p_adj_bh (family-wide control still holds), not the raw p_value.

Contrast: zikv_vs_mock

Reference condition: mock  ·  Test condition: zikv. Every delta is test − reference (mu_test − mu_reference / mean_tail_test − mean_tail_reference), so a POSITIVE value means higher in zikv (test) than in mock (reference), and a negative value means lower.

analysis	tested	FDR<0.05	effect_threshold	FDR<0.05 & above threshold
Differential modification (site-level, pooled across fragmentforms)	1430	15	|delta| >= 0.1	14
Differential modification per fragmentform (ZN partition)	1449	0	|delta| >= 0.1	0
Differential isoform usage	129	52	|delta| >= 0.1	13
Differential APA-site usage	45	30	|delta| >= 0.1	16
Differential junction usage	136	94	|delta| >= 0.1	38
Differential poly(A) tail length	72	0	|delta_nt| >= 10 nt	0

Differential modification (site-level, pooled across fragmentforms) — top by effect size

gene_name	chrom	start0	strand	mod_code	mu_reference	mu_test	delta	p_adj_bh
MXD1	chr2	69939760	+	a (m6A)	0.36648	0.80003	0.43355	0.004891
MXD1	chr2	69940417	+	a (m6A)	0.11995	0.49857	0.37863	0.010193
MXD1	chr2	69938439	+	a (m6A)	0.24042	0.59198	0.35156	0.013136
MXD1	chr2	69938255	+	a (m6A)	0.42733	0.72594	0.29861	0.027664
MXD1	chr2	69939779	+	a (m6A)	0.15699	0.45518	0.29819	0.014244
MXD1	chr2	69938700	+	a (m6A)	0.16530	0.46004	0.29475	0.014244
MXD1	chr2	69938505	+	a (m6A)	0.51495	0.80231	0.28736	0.014244
MXD1	chr2	69940436	+	a (m6A)	0.45998	0.74077	0.28079	0.024071
MXD1	chr2	69938147	+	a (m6A)	0.39995	0.68041	0.28046	0.022584
MXD1	chr2	69938415	+	a (m6A)	0.19672	0.47642	0.27970	0.038064
MXD1	chr2	69938351	+	a (m6A)	0.59622	0.87200	0.27578	0.014244
MXD1	chr2	69938378	+	a (m6A)	0.06668	0.28280	0.21612	0.019024
MXD1	chr2	69938777	+	a (m6A)	0.04580	0.22744	0.18164	0.031335
RIOK3	chr18	23481352	+	a (m6A)	0.21010	0.39083	0.18073	0.014244
RANBP3	chr19	5932489	-	17802 (pseudouridine)	0.00109	0.03705	0.03596	0.017039

↗

Differential modification per fragmentform (ZN partition) — top by effect size

↗

Differential isoform usage — top by effect size

gene_name	feature	mu_reference	mu_test	delta	p_adj_bh
RIOK3	RIOK3.RIOK3.G13.T1	0.78484	0.45513	-0.32971	0.000022
MXD1	MXD1.MXD1.G8.T2	0.43605	0.12304	-0.31301	0.000224
NHSL1	NHSL1.NHSL1.G9.T4	0.03448	0.33775	0.30327	0.000384
ALCAM	ALCAM.ALCAM.G1.T1	0.80350	0.56688	-0.23662	0.000148
MXD1	MXD1.MXD1.G8.T3	0.00000	0.17016	0.17016	0.000369
NHSL1	NHSL1.NHSL1.G9.T3	0.27586	0.11921	-0.15665	0.018601
SERAC1	SERAC1.SERAC1.G14.T1	0.36899	0.21806	-0.15092	0.009970
ALCAM	ALCAM.ALCAM.G1.T3	0.00000	0.14861	0.14861	0.000022
VGLL4	VGLL4.VGLL4.G15.T1	0.89026	0.75746	-0.13280	0.002555
RIOK3	RIOK3.RIOK3.G13.T3	0.01467	0.14076	0.12609	0.000148
MXD1	MXD1.MXD1.G8.T4	0.01744	0.13351	0.11607	0.002995
VGLL4	VGLL4.VGLL4.G15.T2	0.02980	0.14552	0.11572	0.000634
C8A	C8A.C8A.G5.T3	0.08361	0.19677	0.11315	0.001644
C8A	C8A.C8A.G5.T4	0.07357	0.16441	0.09084	0.003345
C8A	C8A.C8A.G5.T2	0.32776	0.24259	-0.08517	0.034853
SERAC1	SERAC1.SERAC1.G14.T8	0.00535	0.09031	0.08496	0.003345
ATG7	ATG7.ATG7.G4.T2	0.00914	0.09299	0.08385	0.000148
RIOK3	RIOK3.RIOK3.G13.T4	0.00489	0.08212	0.07723	0.000416
IP6K2	IP6K2.IP6K2.G7.T5	0.02656	0.10314	0.07658	0.000022
RANBP3	RANBP3.RANBP3.G12.T3	0.21181	0.13736	-0.07446	0.002866

↗

Differential APA-site usage — top by effect size

gene_name	feature	mu_reference	mu_test	delta	p_adj_bh
MXD1	chr2:69942945:+	0.43605	0.12303	-0.31302	0.000138
NHSL1	chr6:138446827:-	0.03444	0.33776	0.30333	0.000301
ALCAM	chr3:105576900:+	0.92998	0.67096	-0.25902	0.000010
NHSL1	chr6:138422043:-	0.71032	0.48998	-0.22034	0.006302
MXD1	chr2:69940641:+	0.54651	0.74346	0.19695	0.002420
VGLL4	chr3:11556069:-	0.93910	0.78729	-0.15181	0.000417
RIOK3	chr18:23481478:+	0.16873	0.31746	0.14873	0.000536
ALCAM	chr3:105524960:+	0.00000	0.14854	0.14853	0.000010
RIOK3	chr18:23483137:+	0.80927	0.66262	-0.14665	0.000660
IP6K2	chr3:48693764:-	0.34080	0.46154	0.12074	0.000094
SERAC1	chr6:158109519:-	0.42781	0.30837	-0.11944	0.029048
MXD1	chr2:69939558:+	0.01745	0.13351	0.11606	0.002032
VGLL4	chr3:11557979:-	0.02967	0.14556	0.11589	0.000515
C8A	chr1:56883782:+	0.08362	0.19678	0.11316	0.001020
RANBP3	chr19:5916139:-	0.57490	0.47088	-0.10402	0.001426
RANBP3	chr19:5917321:-	0.33610	0.44000	0.10390	0.001280
C8A	chr1:56883684:+	0.07347	0.16432	0.09085	0.002350
C8A	chr1:56918106:+	0.32778	0.24259	-0.08519	0.023517
IP6K2	chr3:48688003:-	0.47814	0.39516	-0.08298	0.000788
ATG7	chr3:11557665:+	0.10448	0.03101	-0.07347	0.000417

↗

Differential junction usage — top by effect size

gene_name	feature	mu_reference	mu_test	delta	p_adj_bh
NHSL1	chr6:138429844-138430392:-	0.96552	0.66225	-0.30327	0.000111
NHSL1	chr6:138424817-138429710:-	0.96552	0.66225	-0.30327	0.000111
NHSL1	chr6:138496372-138499232:-	0.03448	0.33775	0.30327	0.000111
RIOK3	chr18:23473627-23474947:+	0.97555	0.71085	-0.26470	0.000003
RIOK3	chr18:23473484-23474947:+	0.02445	0.28915	0.26470	0.000003
C8A	chr1:56883682-56885926:+	0.81048	0.59030	-0.22018	0.000075
SERAC1	chr6:158150590-158158272:-	0.24064	0.45154	0.21090	0.001008
SERAC1	chr6:158150590-158155314:-	0.75936	0.54846	-0.21090	0.001008
SERAC1	chr6:158155352-158158272:-	0.75936	0.54846	-0.21090	0.001008
C8A	chr1:56906793-56907955:+	0.75920	0.56604	-0.19316	0.000154
C8A	chr1:56886168-56906666:+	0.75920	0.56604	-0.19316	0.000154
C8A	chr1:56908114-56912402:+	0.75920	0.56604	-0.19316	0.000154
C8A	chr1:56912626-56917564:+	0.75920	0.56604	-0.19316	0.000154
MXD1	chr2:69921766-69935350:+	1.00000	0.82984	-0.17016	0.000105
MXD1	chr2:69916221-69935350:+	0.00000	0.17016	0.17016	0.000105
MXD1	chr2:69916221-69921735:+	1.00000	0.82984	-0.17016	0.000105
NHSL1	chr6:138433681-138441982:-	0.27586	0.11921	-0.15665	0.011922
NHSL1	chr6:138496372-138545622:-	0.27586	0.11921	-0.15665	0.011922
NHSL1	chr6:138442115-138447000:-	0.27586	0.11921	-0.15665	0.011922
ALCAM	chr3:105524509-105532001:+	1.00000	0.85138	-0.14861	0.000003

↗

Differential poly(A) tail length — top by effect size

↗

Read-supported non-reference loci discovered in the cleaned tagged BAMs.

Associations between segregating SNP alleles and fragmentform-partition usage.

Top 4 SNP→transcript example(s)

↗

↗

↗

↗

Associations between segregating SNP alleles and target modification states on the same molecules.

Per-example allele-split stoichiometry figures

Top 12 SNP→modification example(s)

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

Per-fragmentform breakdown at each hit

Per-fragmentform breakdown at top cis-SNP-linked modification sites (split by SNP allele)

For each top hit, the target modification's stoichiometry is broken down across every fragmentform (ZN partition) of the gene AND split by SNP allele, computed directly from the per-read modification calls joined to the per-read SNP genotypes. Comparing ref vs alt within the same fragmentform is the direct allelic-effect read-out — it tells apart a variant that changes modification from one that merely shifts which fragmentform is used.

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

↗

A SNP whose alleles carry different modification rates gets explained on three axes: WHERE it sits relative to the modified base (at it / inside the 5-mer / inside the 9-mer / proximal / distal cis), whether the alt allele DISRUPTS or CREATES the modification's sequence motif, and whether the observed allelic direction matches what the motif predicts. A sequence motif is used only where the modification has one (the DRACH consensus for m6A); modifications without a comparable motif (e.g. pseudouridine, inosine) get positional context only.

↗

• 1,030 SNP x modification pairs classified
• 0 of the 0 significant (FDR<0.05) pairs are self-reporting artifacts — the variant and the modification are potentially the same physical event

For every pair of nearby modification sites seen on the SAME reads, does the state of one predict the state of the other? The 2x2 is tested conditional on reads covering both sites, and enrichment is judged against the expected co-occurrence given each site's own marginal stoichiometry — so a concordant call means the two are modified together MORE than their individual rates alone would predict, not merely that both are often modified. Significance, effect_abs_delta_mod_frac (|P(B mod|A mod) − P(B mod|A unmod)|), and odds_ratio/direction are all SAMPLE-STRATIFIED (Cochran–Mantel–Haenszel p, Mantel–Haenszel common odds ratio); the *_pooled columns are the single-2x2 companions and can be Simpson's-paradox inflated. jaccard_both_pooled is co-modified reads / reads modified at either.

12,024 co-localized mod-site pairs tested; 639 significant at BH-FDR < 0.05.

direction: 6,331 concordant · 2,169 mutually exclusive · 3,519 independent

modification codes: 17596 = inosine · 17802 = pseudoU · 19227 = Um · 19228 = Cm · 19229 = Gm · 69426 = Am · a = m6A · m = 5mC

↗