When you click "Launch Pipeline", EpiNexus works through these steps automatically and the run page shows each one's progress. Steps that don't apply to your data are skipped, and the run page says why. Only a failed alignment stops a run; if another step fails, the others still finish.
Trims adapters (Trim Galore), aligns reads (Bowtie2), removes duplicates (kept in CUT&Tag and CUT&RUN target samples, where they are expected) and blacklisted regions, and calls peaks: MACS3 for ChIP-seq and ATAC-seq, SEACR or GoPeaks for CUT&Tag and CUT&RUN. Input and IgG controls are used as background. Writes signal tracks (bigWig). From BAM files, your own peak files are used instead.
Usable fragments, mapping rate, duplication, library complexity, FRiP, fragment sizes, TSS enrichment and spike-in reads, graded against published standards for each kind of sample (ENCODE and others): Good, Marginal, Below minimum or Fail. Only ENCODE's error level, or a library without usable data, fails a sample. With a plain-language interpretation.
Checks which peaks replicate: IDR for two transcription-factor or ATAC-seq replicates, overlap between replicates otherwise.
Labels each peak by where it sits in the genome (promoter, UTR, exon, intron, downstream, distal intergenic) and finds the nearest gene.
Connects each regulatory region to the gene(s) it most likely controls within 5 Mb, with the Activity-by-Contact (ABC) model's rules, for samples that measure activity (accessibility or H3K27ac).
Compares conditions (every pair, or the ones you choose) to find regions that gained or lost signal: DESeq2 (in R), csaw for broad marks, spike-in calibration when available. Sharp histone marks are tested on one peak call on the pooled libraries.
GREAT-style Gene Ontology enrichment of gained and lost regions and of each sample's peaks.
Reads H3K27ac and H3K27me3 changes together at each gene's promoter, and groups genes by what both marks did, for example H3K27ac gained and H3K27me3 lost.
Stitches nearby peaks (12.5 kb, ROSE), ranks them by signal and separates super-enhancers from typical enhancers.
Tests transcription-factor motifs from JASPAR 2024 for enrichment in your peaks (MEME AME), and in the regions that gained or lost signal against unchanged regions of the same GC content (MEME SEA and monaLisa).
Tornado heatmaps and average profiles around peaks for every sample and its input, and a sample-correlation matrix (deepTools).
Collects the tools' own QC reports into one MultiQC report. The run also records tool versions, commands and the reference genome used.
Compares the changed regions and each sample's peaks with public chromatin data: ENCODE's regulatory elements, chromatin states of hundreds of reference samples from many tissues, and genome-wide states. Checks each sample against the tissue you gave it. Starts by itself after new runs; older runs can be compared from the Regulatory Atlas page.
Methylation runs take their own path: bisulfite alignment (Trim Galore and Bismark; whether the libraries are RRBS or WGBS is detected from the reads, and RRBS reads are never deduplicated) → DMR calling (dmrseq, region-level FDR, on CpGs covered by at least 4 reads in every sample by default; you can change both values for a run) → MultiQC report. Results: differentially methylated regions with their nearest gene and CpG-island context (islands, shores, shelves, open sea), bisulfite QC (coverage, conversion, M-bias), sample plots and figures. When nothing is significant, the strongest candidate regions are shown and labelled. You can also start from Bismark coverage (.cov) files.
17 configured genomes across the major model organisms. The app lists the genomes installed on the service; ask us if the one you need isn't there yet.
Homo sapiens
GRCh38 · GENCODE v46Homo sapiens
GRCh37 · GENCODE v46lift37Mus musculus
GRCm39 · GENCODE vM35Mus musculus
GRCm38 · GENCODE vM25Rattus norvegicus
mRatBN7.2 · UCSC ncbiRefSeqRattus norvegicus
Rnor_6.0 · UCSC ncbiRefSeqDanio rerio
GRCz11 · UCSC ncbiRefSeqDrosophila melanogaster
BDGP6 · UCSC ncbiRefSeqCaenorhabditis elegans
WBcel235 · UCSC ncbiRefSeqSaccharomyces cerevisiae
R64 · UCSC ncbiRefSeqMacaca mulatta
Mmul_10 · MacaqueGallus gallus
GRCg6a · ChickenXenopus laevis
v10.1 · FrogSus scrofa
Sscrofa11.1 · PigBos taurus
ARS-UCD1.2 · CowCanis lupus familiaris
UU_Cfam_GSD_1.0 · DogArabidopsis thaliana
TAIR10 · Ensembl Plants 62When you create a project, EpiNexus lists the genomes installed on the service. If the one you need is missing — one of the genomes above, or another organism with a reference genome and gene annotation — write to info@epinexus.io and we'll look at adding it.
Genome sequences come from UCSC (human: the no-alt analysis sets of hg38 and hg19) and Ensembl Plants (Arabidopsis); gene annotations from GENCODE (human and mouse), UCSC ncbiRefSeq (other animals), or Ensembl Plants 62 (Arabidopsis). All references use chr-prefixed chromosome names. Downloads are checked against the providers' checksums where they publish them, and seven genomes (hg38, hg19, mm39, mm10, dm6, ce11, tair10) have a published blacklist of problem regions that EpiNexus removes. CpG islands come from UCSC (13 vertebrate genomes). Every run records the reference build it used.
The Regulatory Atlas uses ENCODE's registry of candidate regulatory elements (V4), chromatin states of 833 human reference samples (EpiMap) and of mouse fetal tissues (van der Velde et al. 2021), ENCODE's DNase accessibility of mouse samples, and genome-wide full-stack ChromHMM states. The mm39 atlas is converted from mm10 with NCBI's assembly alignment. The app's Data sources page lists every source with its version, licence and what to cite.