Appendix B — QC thresholds and their sources

This appendix lists every value EpiNexus 1.4 grades quality metrics against. Chapter 6 explains what each metric measures. EpiNexus has graded this way since version 1.2.7. Runs made earlier keep the grades they were given; Run again grades them again.

B.1 How a value is graded

Each metric has up to three values: a good value, a minimum and a fail value.

Table B.1: The four grades.
Grade When
Good At or better than the good value
Marginal Between the good value and the minimum
Below minimum Worse than the minimum
Fail Worse than the fail value
  • Only a Fail fails a sample. A sample Below minimum is kept, with a warning.
  • A fail value exists only where ENCODE sets an error level. A library with no usable fragments, or with peaks but no reads in them, also fails, whatever it is.
  • Values that no publication sets are EpiNexus’s own choices. They are marked “EpiNexus” in the tables below. They never fail a sample.
  • A metric without a value for a kind of sample is shown without a grade.
  • “Usable fragments” are fragments with good mapping quality (at the run’s minimum, 30 by default) that are not duplicates and not in a blacklisted region. With paired-end reads, a pair counts once. CUT&Tag and CUT&RUN samples keep their duplicates, so duplicates count there.

B.2 Which values a sample is graded against

The values depend on what was profiled. EpiNexus sorts each sample into a class from its assay and its target. The QC page shows the class under Graded as.

Table B.2: Classes of samples. The lists of narrow and broad marks are ENCODE’s (ENCODE Data Coordination Center 2026); EpiNexus adds H2AK119ub to the broad marks.
Assay Class Samples
ChIP-seq narrow histone mark H2A.Z, H3ac, H3K27ac, H3K4me2, H3K4me3, H3K9ac
ChIP-seq broad histone mark H3.3, H3K27me3, H3K36me3, H3K4me1, H3K79me2, H3K79me3, H3K9me1, H3K9me2, H4K20me1, H2AK119ub
ChIP-seq H3K9me3 H3K9me3 (graded on all mapped reads)
ChIP-seq transcription factor or other protein Any target that is not a histone mark
ChIP-seq histone mark not on ENCODE’s lists Other histone marks; graded as narrow marks
ChIP-seq target not given Samples without a target; graded as narrow marks
ChIP-seq control for narrow marks or factors; control for broad marks Inputs, graded with the values of the marks they serve; broad when any of them is broad
ATAC-seq ATAC-seq All ATAC-seq samples
CUT&Tag, CUT&RUN histone mark; transcription factor or other protein; target not given By the target
CUT&Tag, CUT&RUN IgG control IgG and other controls

B.3 ChIP-seq

Table B.3: ChIP-seq values. NSC and RSC have no minimum: a value between the good value and zero is Marginal.
Metric Narrow marks, factors, other targets Broad marks Source
Usable fragments (good / minimum / fail) 20 M / 10 M / 5 M 45 M / 35 M / 5 M (ENCODE Data Coordination Center 2026)
H3K9me3: mapped fragments (good / minimum / fail) — 45 M / 35 M / fewer than 100 (ENCODE Data Coordination Center 2026)
Mapping rate (good / minimum) 85 % / 70 % 85 % / 70 % EpiNexus
Duplicate rate (good / minimum) at most 20 % / above 40 % the same EpiNexus
NRF (good / minimum) 0.8 / 0.5 0.8 / 0.5 (ENCODE Project Consortium 2026b)
PBC1 (good / minimum) 0.9 / 0.5 0.9 / 0.5 (ENCODE Project Consortium 2026b)
PBC2 (good / minimum) 10 / 1 10 / 1 (ENCODE Project Consortium 2026b)
FRiP (good / minimum) factors: 5 % / 1 %; histone marks and other targets: not graded not graded 5 %: EpiNexus; 1 %: Landt et al. (2012)
NSC (good / fail) factors and narrow marks: 1.05 / negative; other targets: not graded, negative fails not graded; negative fails 1.05: Landt et al. (2012); negative: ENCODE Data Coordination Center (2026)
RSC (good / fail) factors and narrow marks: 0.8 / negative; other targets: not graded, negative fails not graded; negative fails 0.8: Landt et al. (2012); negative: ENCODE Data Coordination Center (2026)
Peaks (good / minimum) 10,000 / 3,000 10,000 / 3,000 EpiNexus

Controls. An input is graded on usable fragments with the values of the marks it serves, and on mapping rate, duplicates, NRF, PBC1 and PBC2. It has no peaks, FRiP, NSC or RSC.

Fragment size. For ChIP-seq, EpiNexus warns when the median fragment size is outside 150 to 300 bp. This is EpiNexus’s own check. It gives a note, not a grade.

B.4 ATAC-seq

Table B.4: ATAC-seq values.
Metric Good Minimum Fail Source
Usable fragments 25 M 20 M 15 M (ENCODE Data Coordination Center 2026; ENCODE Project Consortium 2026a)
Mapping rate 95 % 80 % — (ENCODE Project Consortium 2026a)
NRF 0.9 0.7 — (ENCODE Project Consortium 2026a)
PBC1 0.9 0.7 — (ENCODE Project Consortium 2026a)
PBC2 3 1 — (ENCODE Project Consortium 2026a)
FRiP 30 % 20 % — (ENCODE Project Consortium 2026a)
Duplicate rate not graded up to 25 % above 25 %: Marginal — Grandi et al. (2022)
Mitochondrial reads at most 15 % above 30 % — 15 %: Grandi et al. (2022); 30 %: EpiNexus
Peaks 20,000 5,000 — EpiNexus

TSS enrichment is graded by ENCODE’s bands for the genome (ENCODE Project Consortium 2026a): “ideal” is Good, “acceptable” is Marginal and “concerning” is Below minimum.

Table B.5: ENCODE’s TSS enrichment bands. For other genomes, such as mm39, the score is shown without a grade.
Genome Concerning Acceptable Ideal
hg38 below 5 5 to 7 above 7
hg19 below 6 6 to 10 above 10
mm10 below 10 10 to 15 above 15

B.5 CUT&Tag and CUT&RUN

Table B.6: CUT&Tag and CUT&RUN values. Duplicates are not graded for these assays. “Unusually many peaks” often means the peak call failed.
Metric CUT&Tag CUT&RUN Source
Usable fragments, targets (good / minimum) 3 M / 1 M 3 M / 1 M CUT&Tag: Kaya-Okur et al. (2020) and Zheng et al. (2020); CUT&RUN: 3 M from EpiCypher (2026) and Meers et al. (2019), 1 M EpiNexus
Mapped reads, targets (fail) fewer than 0.2 M — Murray et al. (2026)
Usable fragments, IgG control (minimum) not graded 1 M EpiCypher (2026)
Mapping rate (good / minimum) 80 % / 60 % 85 % / 65 % EpiNexus
FRiP, histone marks and targets not given (good / minimum) 30 % / 10 % 30 % / 10 % 30 %: nf-core community (2026); 10 %: EpiNexus
FRiP, factors not graded not graded no published value
Peaks (good / minimum) 5,000 / 1,000 5,000 / 1,000 EpiNexus
Peaks, unusually many above 150,000: Marginal the same EpiNexus

B.6 Spike-in notes

The share of spike-in reads gives notes, never grades or failures (Chapter 10).

Table B.7: Spike-in notes. The first two do not apply to an E. coli spike-in: its DNA comes with the enzyme (CUT&Tag) or is added in a small fixed amount (CUT&RUN kits), not as measured chromatin.
Note When Source
Too few to normalise reliably spike-in reads under 0.5 % of the library Tallan et al. (2025)
Small differences change the result under 2 % Tallan et al. (2025)
Depth spent on the spike-in above 20 % EpiNexus
Many reads align to both genomes more than 5 % of the spike-in reads EpiNexus