{"slug":"pipeline-rnaseq-2","title":"pipeline-rnaseq","summary":"Execute ENCODE RNA-seq pipeline from FASTQ to gene quantification and signal tracks. Child of pipeline-guide. Provides Nextflow execution with Docker and cloud deployment. Use when processing RNA-seq data with STAR alignment, RSEM/Kallisto quantification, or generating expression","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-09-22T13:25:44.840475Z","repo":{"url":"https://github.com/ammawla/encode-toolkit","stars":21,"forks":5,"license":"AGPL-3.0","updatedAt":"2026-09-27T01:22:42Z"},"bodyHtml":"
\n

name: pipeline-rnaseq\ndescription: \"Execute ENCODE RNA-seq pipeline from FASTQ to gene quantification and signal tracks. Child of pipeline-guide. Provides Nextflow execution with Docker and cloud deployment. Use when processing RNA-seq data with STAR alignment, RSEM/Kallisto quantification, or generating expression matrices. Trigger on: RNA-seq pipeline, gene expression, STAR alignment, RSEM quantification, transcript quantification, TPM, FPKM, RNA processing, run RNA-seq.\"

\n

ENCODE RNA-seq Pipeline

\n

When to Use

\n\n

Execute the ENCODE RNA-seq processing pipeline from raw FASTQ files through splice-aware\nalignment, gene/transcript quantification, and strand-specific signal track generation.\nThis skill provides a complete Nextflow DSL2 implementation following ENCODE uniform\nanalysis standards.

\n

Overview

\n

RNA-seq measures transcriptome-wide gene expression by sequencing cDNA derived from\ncellular RNA. The ENCODE pipeline processes RNA-seq data through quality control,\nsplice-aware alignment with STAR (2-pass mode), gene and transcript quantification\nwith RSEM, optional fast pseudoalignment with Kallisto, and generation of strand-specific\nsignal tracks as bigWig files.

\n

Key design decisions: STAR 2-pass mode for maximum splice junction sensitivity, RSEM\nfor accurate gene/transcript/isoform quantification including multi-mapped reads,\nstranded library protocol (dUTP/rf-stranded) as the ENCODE standard, and paired-end\nsequencing with a minimum of 30 million uniquely mapped reads per replicate.

\n

Key Literature

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ReferenceJournalYearDOIRelevance
Dobin et al. \"STAR: ultrafast universal RNA-seq aligner\"Bioinformatics201310.1093/bioinformatics/bts635Splice-aware aligner (~12,000 citations)
Li & Dewey \"RSEM: accurate transcript quantification from RNA-Seq data\"BMC Bioinformatics201110.1186/1471-2105-12-323Gene/transcript quantification (~6,000 citations)
Bray et al. \"Near-optimal probabilistic RNA-seq quantification\"Nature Biotechnology201610.1038/nbt.3519Fast pseudoalignment (~4,000 citations)
Wang et al. \"RSeQC: quality control of RNA-seq experiments\"Bioinformatics201210.1093/bioinformatics/bts356RNA-seq QC suite (~3,500 citations)
ENCODE Project Consortium \"Expanded encyclopaedias\"Nature202010.1038/s41586-020-2493-4ENCODE Phase 3 standards
Frankish et al. \"GENCODE 2021\"Nucleic Acids Research202110.1093/nar/gkaa1087Gene annotation reference
\n

Pipeline Stages

\n
FASTQ\n  ├─> FastQC (raw reads)\n  └─> Trim Galore (+ FastQC on the trimmed reads)\n        ├─> Kallisto (optional) ──────────> kallisto/<sample>/abundance.tsv\n        └─> STAR (2-pass)\n              ├─> transcriptome BAM ─> RSEM ─> <sample>.genes.results / .isoforms.results\n              ├─> bedGraph str1/str2 ─> bedGraphToBigWig ─> signal/<sample>_{plus,minus}.bw\n              └─> genome BAM ─> RSeQC (infer_experiment, read_distribution,\n                                       geneBody_coverage, inner_distance)\n\nMultiQC <── FastQC (raw + trimmed), trimming reports, STAR Log.final.out,\n            RSEM .stat/, RSeQC infer_experiment + read_distribution\n  └─> qc/multiqc/multiqc_report.html\n
\n

The Kallisto abundances, the gene body coverage and the inner distance files are\npublished but are not part of the MultiQC report; read those files directly.

\n

Stage Summary

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
StageToolInputOutputReference
1. QC & TrimmingFastQC, Trim GaloreRaw FASTQTrimmed FASTQ + FastQC reportsreferences/01-qc-trimming.md
2. AlignmentSTAR (2-pass)Trimmed FASTQGenome BAM + Transcriptome BAM + bedGraphreferences/02-star-alignment.md
3. QuantificationRSEM, KallistoTranscriptome BAM / trimmed FASTQGene/transcript counts, TPM, FPKMreferences/03-quantification.md
4. Signal TracksbedGraphToBigWigSTAR bedGraphStrand-specific bigWigreferences/04-signal-tracks.md
5. QC MetricsRSeQC, MultiQCGenome BAM, logsStrandedness, read distribution, gene body coveragereferences/05-qc-metrics.md
\n

Input Requirements

\n

Required Files

\n\n

There is no sample sheet: samples are the pairs that --reads matches, and the sample\nID is the shared prefix of each pair. The gene annotation is not a workflow parameter\nand there is no --gtf; the GTF is consumed when the STAR and RSEM references are built\n(references/02 and references/03), so the annotation is fixed by the index you pass in.

\n

Library Strandedness

\n

--strandedness takes one value for the whole run — reverse (default), forward or\nnone — and is validated before the first task. It drives three things at once: the RSEM\n--strandedness flag, the kallisto strand flag, and which STAR bedGraph becomes which\nsignal track.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Protocol--strandednessRSEM receivesKallisto receivesSignal tracks
dUTP (ENCODE standard), Illumina TruSeq Strandedreverse--strandedness reverse--rf-stranded<sample>_plus.bw, <sample>_minus.bw
Directional ligation (some legacy protocols)forward--strandedness forward--fr-stranded<sample>_plus.bw, <sample>_minus.bw
SMARTer / SMART-Seq2 and other unstranded kitsnone--strandedness noneno strand flag<sample>_unstranded.bw
\n

There is no per-sample strandedness and the workflow does not detect it. RSeQC\ninfer_experiment.py runs as a post-hoc check and writes\nqc/rseqc/<sample>.infer_experiment.txt. If the library type is unknown, run a first\npass, read that file (references/05 explains the output), and rerun with the correct\n--strandedness — the RSEM counts, the kallisto abundances and the signal tracks all\ndepend on it, so a wrong value has to be corrected by rerunning, not by post-processing.

\n

QC Thresholds

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
MetricThresholdProduced by
Total sequenced reads>=30M PE readsfastqc/, star/<sample>.Log.final.out
Uniquely mapped reads>=70% of input readsstar/<sample>.Log.final.out
Multi-mapped reads<10%star/<sample>.Log.final.out
Strandedness agreement>90% for a stranded libraryqc/rseqc/<sample>.infer_experiment.txt
Exonic rate>60%qc/rseqc/<sample>.read_distribution.txt
Gene body coverageRelatively uniform (5'/3' bias <1.5)qc/rseqc/<sample>.geneBody_coverage.geneBodyCoverage.txt
\n

Not computed by this workflow: rRNA rate, library duplication rate (beyond the\nsequence-level estimate inside the FastQC report), detected-gene counts, and saturation\ncurves. references/05-qc-metrics.md gives the commands to run those by hand on the\npublished BAM and RSEM output.

\n

Read Depth Guidelines

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ApplicationMinimum Reads (PE)RecommendedNotes
Gene-level expression20M30MENCODE minimum
Transcript-level expression40M60MIsoform resolution requires more depth
Differential expression20M per sample30M per sample3+ biological replicates per condition
Novel junction discovery60M100M+STAR 2-pass mode benefits from depth
Fusion detection50M80M+Chimeric reads are rare; needs a separate STAR run (references/02)
\n

Execution

\n

The versions the workflow runs are the ones in scripts/Dockerfile: STAR 2.7.11b,\nRSEM 1.3.3, kallisto 0.50.1, samtools 1.19, RSeQC 5.0.3, Trim Galore 0.6.10, cutadapt 4.6,\nMultiQC 1.21 and FastQC 0.12.1. The conda environment in bioinformatics-installer\n(environments/rnaseq-env.yml) is a separate manual route pinned to the same versions of\nSTAR, RSEM, kallisto, samtools, RSeQC, Trim Galore, FastQC and MultiQC; it leaves cutadapt\nto the Trim Galore package, adds salmon and subread, and does not carry\nbedGraphToBigWig, which the signal-track step needs.

\n

Every index flag is shown in the examples below because the defaults are bare names\nresolved in the launch directory (GRCh38_star_index, GRCh38_rsem_index/GRCh38,\ngencode.v38.kallisto.idx, hg38_RefSeq.bed for --genome GRCh38). The run stops\nbefore the first task if any of them is missing. --kallisto_index is the one exception:\nit is not read when --skip_kallisto is set.

\n

Quick Start (local, Docker)

\n
nextflow run scripts/main.nf -profile local \\\n    --reads 'fastq/*_R{1,2}.fq.gz' \\\n    --genome GRCh38 \\\n    --star_index /ref/GRCh38_star_index \\\n    --rsem_index /ref/GRCh38_rsem_index/GRCh38 \\\n    --kallisto_index /ref/gencode.v38.kallisto.idx \\\n    --rseqc_bed /ref/hg38_RefSeq.bed \\\n    --strandedness reverse \\\n    --outdir results/\n
\n

SLURM HPC

\n
nextflow run scripts/main.nf -profile slurm \\\n    --container /path/to/pipeline-rnaseq.sif \\\n    --slurm_queue normal \\\n    --reads 'fastq/*_R{1,2}.fq.gz' \\\n    --genome GRCh38 \\\n    --star_index /ref/GRCh38_star_index \\\n    --rsem_index /ref/GRCh38_rsem_index/GRCh38 \\\n    --kallisto_index /ref/gencode.v38.kallisto.idx \\\n    --rseqc_bed /ref/hg38_RefSeq.bed \\\n    --outdir results/\n
\n

Cloud

\n
# Google Cloud Batch\nnextflow run scripts/main.nf -profile gcp \\\n    --container us-docker.pkg.dev/<project>/<repo>/pipeline-rnaseq:1.0.0 \\\n    --gcp_project <project> \\\n    --gcp_workdir gs://<bucket>/work \\\n    --reads 'gs://<bucket>/fastq/*_R{1,2}.fq.gz' \\\n    --genome GRCh38 \\\n    --star_index gs://<bucket>/ref/GRCh38_star_index \\\n    --rsem_index gs://<bucket>/ref/GRCh38_rsem_index/GRCh38 \\\n    --kallisto_index gs://<bucket>/ref/gencode.v38.kallisto.idx \\\n    --rseqc_bed gs://<bucket>/ref/hg38_RefSeq.bed \\\n    --outdir gs://<bucket>/results\n\n# AWS Batch\nnextflow run scripts/main.nf -profile aws \\\n    --container <account>.dkr.ecr.<region>.amazonaws.com/pipeline-rnaseq:1.0.0 \\\n    --aws_queue <job-queue> \\\n    --aws_workdir s3://<bucket>/work \\\n    --reads 's3://<bucket>/fastq/*_R{1,2}.fq.gz' \\\n    --genome GRCh38 \\\n    --star_index s3://<bucket>/ref/GRCh38_star_index \\\n    --rsem_index s3://<bucket>/ref/GRCh38_rsem_index/GRCh38 \\\n    --kallisto_index s3://<bucket>/ref/gencode.v38.kallisto.idx \\\n    --rseqc_bed s3://<bucket>/ref/hg38_RefSeq.bed \\\n    --outdir s3://<bucket>/results\n
\n

--outdir only sets where results are published; Google Batch and AWS Batch stage every\ntask through the work directory, and the workflow stops with an error if it or the\nproject/queue is missing.

\n

Pipeline Parameters

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ParameterDefaultDescription
--readsrequiredGlob matching the FASTQ pairs, e.g. 'fastq/*_R{1,2}.fq.gz'. Quote it
--genomeGRCh38GRCh38 or mm10; selects the default index names and nothing else
--outdir./resultsDirectory results are published to
--single_endfalseTreat --reads as single files; kallisto then runs with the fixed --single -l 200 -s 20, and RSeQC inner_distance.py is skipped
--strandednessreversereverse, forward or none; one value for the whole run
--skip_kallistofalseSkip KALLISTO_QUANT; --kallisto_index is then not read
--star_indexGRCh38_star_index (mm10_star_index)STAR genome directory
--rsem_indexGRCh38_rsem_index/GRCh38 (mm10_rsem_index/mm10)RSEM reference prefix, not a directory; every file starting with it is staged
--kallisto_indexgencode.v38.kallisto.idx (gencode.vM27.kallisto.idx)kallisto index file; must be built with kallisto 0.50.1 (index version 13), not with 0.48 or earlier
--rseqc_bedhg38_RefSeq.bed (mm10_RefSeq.bed)BED12 gene model used by all four RSeQC modules
--chrom_sizes<star_index>/chrNameLength.txtChromosome sizes for bedGraphToBigWig
\n

Infrastructure parameters (nextflow.config)

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
ParameterDefaultDescription
--containerencode-toolkit/pipeline-rnaseq:1.0.0Image built from scripts/Dockerfile. Pass a registry image for gcp/aws, or a .sif file for slurm
--max_cpus, --max_memory, --max_time16, 64.GB, 24.hUpper bounds applied to every process
--slurm_queue, --slurm_accountnormal, noneSLURM partition and account
--gcp_project, --gcp_workdirnone (both required for -profile gcp)Google Cloud project and gs:// work directory
--gcp_location, --gcp_diskus-central1, 200.GBGoogle Batch region and per-task disk
--aws_queue, --aws_workdirnone (both required for -profile aws)AWS Batch job queue and s3:// work directory
--aws_region, --aws_cli_pathus-east-1, /home/ec2-user/miniconda/bin/awsAWS region, and the AWS CLI path inside the Batch AMI
\n

Cloud Cost Estimates

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
PlatformInstanceCost/SampleTime/SampleNotes
GCPn1-highmem-8~$3-62-4 hoursSTAR index loading dominates; the gcp profile uses spot VMs
AWSr5.2xlarge~$3-62-4 hoursr-series for STAR memory; spot recommended
Local8 cores, 36 GB$03-6 hoursDocker required
SLURM8 cores, 36 GBVaries2-4 hoursSingularity; pass the .sif with --container
\n

Memory note: nextflow.config asks for 36 GB for STAR_ALIGN (multiplied by the\nattempt number on a retry, capped by --max_memory). The local executor refuses the task\non a machine with less, so a 32 GB host is not enough. If 36 GB is out of reach, rebuild the STAR index\nwith a larger --genomeSAsparseD (which shrinks the loaded index at some cost in\nmapping speed) or move to a bigger machine. --limitGenomeGenerateRAM is a\ngenomeGenerate option and is not a parameter of this workflow.

\n

Output Directory Structure

\n
results/\n  fastqc/                                        # FastQC HTML/zip for raw and trimmed reads\n  trimmed/\n    <sample>_R1_val_1.fq.gz, <sample>_R2_val_2.fq.gz\n    <sample>_R1.fq.gz_trimming_report.txt        # one per input file\n  star/\n    <sample>.Aligned.sortedByCoord.out.bam\n    <sample>.Aligned.sortedByCoord.out.bam.bai\n    <sample>.Aligned.toTranscriptome.out.bam     # RSEM input\n    <sample>.Log.final.out\n    <sample>.SJ.out.tab\n    <sample>.ReadsPerGene.out.tab                # STAR gene counts\n    <sample>.Signal.UniqueMultiple.str1.out.bg   # plus str2 unless --strandedness none\n    <sample>.Signal.Unique.str1.out.bg           # same signal, unique mappers only\n  rsem/\n    <sample>.genes.results                       # gene_id, TPM, FPKM, expected_count\n    <sample>.isoforms.results                    # transcript_id, TPM, FPKM, IsoPct\n    <sample>.stat/                               # RSEM model statistics, read by MultiQC\n  kallisto/                                      # absent with --skip_kallisto\n    <sample>/abundance.tsv\n    <sample>/abundance.h5                        # only from a kallisto build with HDF5 support\n    <sample>/run_info.json\n  signal/\n    <sample>_plus.bw, <sample>_minus.bw          # or <sample>_unstranded.bw\n  qc/\n    rseqc/\n      <sample>.infer_experiment.txt\n      <sample>.read_distribution.txt\n      <sample>.geneBody_coverage.*\n      <sample>.inner_distance.*                  # paired-end only\n    multiqc/\n      multiqc_report.html\n      multiqc_data/\n  pipeline_info/\n    timeline.html, report.html, trace.txt        # Nextflow execution reports\n
\n

Common Pitfalls

\n

1. Insufficient Memory for STAR

\n

STAR loads the whole genome index into memory. STAR_ALIGN is configured for 36 GB, so a\n32 GB machine will not schedule the task at all under -profile local. On shared HPC\nsystems, check the per-job memory limit before submitting.

\n

2. Wrong Strandedness Setting

\n

Using incorrect strandedness results in near-zero gene counts. If you see uniformly low\ncounts, read qc/rseqc/<sample>.infer_experiment.txt and rerun with the matching\n--strandedness. ENCODE dUTP libraries are reverse stranded.

\n

3. Using FPKM for Cross-Sample Comparison

\n

FPKM values are not comparable across samples because they depend on total library\ncomposition. Use TPM (comparable across samples) or raw counts with DESeq2/edgeR\nnormalization for differential expression.

\n

4. Ignoring Multi-Mapped Reads

\n

RSEM uses an expectation-maximization algorithm to probabilistically assign multi-mapped\nreads. This is critical for gene families and repetitive elements. Do not pre-filter\nmulti-mappers before RSEM quantification.

\n

5. Assuming rRNA Contamination Was Checked

\n

High rRNA contamination (>10%) indicates failed rRNA depletion and reduces effective\nsequencing depth. This workflow does not measure it — run the manual check in\nreferences/05-qc-metrics.md against star/<sample>.Aligned.sortedByCoord.out.bam before\ntrusting the quantifications.

\n

6. Not Using 2-Pass Mode for Novel Junctions

\n

STAR 1-pass mode only uses annotated splice junctions. 2-pass mode first discovers novel\njunctions then re-maps, critical for non-model organisms or samples with extensive\nalternative splicing. This workflow always runs --twopassMode Basic.

\n

Pipeline Scripts

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
FileDescription
scripts/main.nfNextflow DSL2 pipeline
scripts/nextflow.configExecution profiles (local/slurm/gcp/aws)
scripts/DockerfileDocker build with STAR, RSEM, Kallisto, RSeQC
\n

ENCODE Data Integration

\n

After running on your own data, compare with ENCODE reference:

\n
# Find matching ENCODE RNA-seq experiments\nencode_search_experiments(\n    assay_title=\"total RNA-seq\",\n    organ=\"pancreas\",\n    biosample_type=\"tissue\"\n)\n\n# Download ENCODE gene quantifications for comparison\nencode_batch_download(\n    download_dir=\"/data/encode_reference/\",\n    output_type=\"gene quantifications\",\n    assay_title=\"total RNA-seq\",\n    organ=\"pancreas\",\n    assembly=\"GRCh38\"\n)\n\n# Download ENCODE signal tracks for browser visualization\nencode_search_files(\n    file_format=\"bigWig\",\n    assay_title=\"total RNA-seq\",\n    organ=\"pancreas\",\n    output_type=\"signal of unique reads\"\n)\n
\n

Pitfalls & Edge Cases

\n\n

Walkthrough: Processing ENCODE RNA-seq from FASTQ to Gene Quantification

\n

Goal: Process raw RNA-seq FASTQ files through the ENCODE pipeline to generate gene expression quantifications (TPM/FPKM) and signal tracks.\nContext: The ENCODE RNA-seq pipeline uses STAR 2-pass alignment and RSEM quantification, producing both gene-level and transcript-level expression estimates.

\n

Step 1: Find RNA-seq experiment

\n
encode_get_experiment(accession=\"ENCSR000CPR\")\n
\n

Expected output:

\n
{\n  \"accession\": \"ENCSR000CPR\",\n  \"assay_title\": \"total RNA-seq\",\n  \"biosample_summary\": \"K562\",\n  \"assembly\": [\"GRCh38\"],\n  \"bio_replicate_count\": 2,\n  \"status\": \"released\"\n}\n
\n

Step 2: List FASTQ files

\n
encode_list_files(experiment_accession=\"ENCSR000CPR\", file_format=\"fastq\")\n
\n

Expected output (a JSON array of file records; fields abridged):

\n
[\n  {\"accession\": \"ENCFF200RN1\", \"file_format\": \"fastq\", \"output_type\": \"reads\", \"file_size_human\": \"3.2 GB\", \"biological_replicates\": [1], \"status\": \"released\"},\n  {\"accession\": \"ENCFF201RN2\", \"file_format\": \"fastq\", \"output_type\": \"reads\", \"file_size_human\": \"3.3 GB\", \"biological_replicates\": [1], \"status\": \"released\"}\n]\n
\n

Step 3: Download and name the FASTQs so a read-pair glob can find them

\n
encode_download_files(file_accessions=[\"ENCFF200RN1\", \"ENCFF201RN2\"], download_dir=\"/data/rnaseq/fastq\")\n
\n

ENCODE names every FASTQ after its accession (ENCFF200RN1.fastq.gz), with no _R1/_R2\nin the name, so the two files of a pair share no prefix and the --reads glob cannot pair\nthem. Link them into the shape the glob expects. Which mate an accession is comes from the\nENCODE file record on encodeproject.org, which carries paired_end (1 or 2) and\npaired_with; the MCP file tools do not return those two fields:

\n
cd /data/rnaseq/fastq\nln -s ENCFF200RN1.fastq.gz k562_rep1_R1.fq.gz\nln -s ENCFF201RN2.fastq.gz k562_rep1_R2.fq.gz\n
\n

Step 4: Run the RNA-seq pipeline

\n
nextflow run scripts/main.nf -profile local \\\n    --reads '/data/rnaseq/fastq/k562_*_R{1,2}.fq.gz' \\\n    --genome GRCh38 \\\n    --star_index /ref/GRCh38_star_index \\\n    --rsem_index /ref/GRCh38_rsem_index/GRCh38 \\\n    --kallisto_index /ref/gencode.v38.kallisto.idx \\\n    --rseqc_bed /ref/hg38_RefSeq.bed \\\n    --strandedness reverse \\\n    --outdir results/\n
\n

Key pipeline steps:

\n
    \n
  1. FastQC on the raw reads, quality and adapter trimming with Trim Galore (which also runs FastQC on the trimmed reads)
    2. \n
  2. STAR 2-pass alignment (splice-aware), writing the genome BAM, the transcriptome BAM, gene counts and bedGraphs
    3. \n
  3. RSEM gene and isoform quantification (TPM, FPKM, expected counts)
    4. \n
  4. Kallisto transcript quantification (optional, skipped with --skip_kallisto)
    5. \n
  5. Signal track generation (bedGraph to bigWig)
    6. \n
  6. RSeQC (infer_experiment, read_distribution, geneBody_coverage, inner_distance) and MultiQC
    7. \n
\n

Step 5: Validate output quality

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
MetricThresholdWhere to read it
Uniquely mapped rate>= 70%star/<sample>.Log.final.out
Strandedness agreement> 90% and matching --strandednessqc/rseqc/<sample>.infer_experiment.txt
Exonic rate> 60%qc/rseqc/<sample>.read_distribution.txt
Replicate correlation>= 0.9compute yourself from the TPM column of rsem/<sample>.genes.results
\n

Step 6: Use expression data with ENCODE epigenomic data

\n

Compare gene expression with enhancer marks:

\n
encode_search_experiments(assay_title=\"Histone ChIP-seq\", biosample_term_name=\"K562\", target=\"H3K27ac\", organism=\"Homo sapiens\")\n
\n

Interpretation: Genes with high TPM AND nearby H3K27ac peaks have validated enhancer-gene connections. Low expression despite nearby enhancer marks suggests poised or tissue-specific regulation.

\n

Integration with downstream skills

\n\n

Code Examples

\n

1. Find RNA-seq experiments for a tissue

\n
encode_search_experiments(assay_title=\"total RNA-seq\", organ=\"liver\", organism=\"Homo sapiens\")\n
\n

Expected output:

\n
{\n  \"results\": [\n    {\"accession\": \"ENCSR300RNA\", \"assay_title\": \"total RNA-seq\", \"biosample_summary\": \"liver tissue male adult (54 years)\", \"status\": \"released\"}\n  ],\n  \"total\": 35,\n  \"limit\": 25,\n  \"offset\": 0,\n  \"has_more\": true,\n  \"next_offset\": 25\n}\n
\n

2. Check for existing gene quantifications

\n
encode_list_files(experiment_accession=\"ENCSR300RNA\", file_format=\"tsv\", output_type=\"gene quantifications\", assembly=\"GRCh38\")\n
\n

Expected output (a JSON array of file records; fields abridged):

\n
[\n  {\"accession\": \"ENCFF400GEQ\", \"file_format\": \"tsv\", \"output_type\": \"gene quantifications\", \"assembly\": \"GRCh38\", \"file_size_human\": \"5.2 MB\", \"status\": \"released\"}\n]\n
\n

3. Download expression data

\n
encode_download_files(file_accessions=[\"ENCFF400GEQ\"], download_dir=\"/data/rnaseq/quantification\")\n
\n

Expected output (fields abridged):

\n
{\n  \"downloaded\": [\n    {\"accession\": \"ENCFF400GEQ\", \"file_path\": \"/data/rnaseq/quantification/ENCFF400GEQ.tsv\", \"file_size_human\": \"5.2 MB\", \"success\": true, \"md5_verified\": true}\n  ],\n  \"errors\": [],\n  \"summary\": {\"total_requested\": 1, \"successful\": 1, \"failed\": 0, \"total_size_human\": \"5.2 MB\"}\n}\n
\n

Integration

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
This skill produces...Feed into...Purpose
Gene expression (TPM/FPKM)peak-annotationValidate enhancer targets with expression data
Expression matrixgtex-expressionCompare cell-line vs. tissue expression
Differential expression resultscompare-biosamplesIdentify tissue-specific gene regulation
Signal tracks (bigWig)visualization-workflowDisplay expression signal in genome browser
Expression quantificationsdisease-researchConnect gene expression to disease phenotypes
Pipeline run parametersdata-provenanceRecord STAR/RSEM versions and settings
QC metricsquality-assessmentValidate against ENCODE RNA-seq standards
\n

Related Skills

\n\n

Presenting Results

\n

When reporting RNA-seq pipeline results:

\n\n

For the request: \"$ARGUMENTS\"

\n","files":[{"path":"references/01-qc-trimming.md","sizeBytes":3445,"isText":true},{"path":"references/02-star-alignment.md","sizeBytes":6043,"isText":true},{"path":"references/03-quantification.md","sizeBytes":5780,"isText":true},{"path":"references/04-signal-tracks.md","sizeBytes":5331,"isText":true},{"path":"references/05-qc-metrics.md","sizeBytes":6884,"isText":true},{"path":"references/literature.md","sizeBytes":12050,"isText":true},{"path":"scripts/Dockerfile","sizeBytes":2723,"isText":false},{"path":"scripts/main.nf","sizeBytes":11914,"isText":false},{"path":"scripts/nextflow.config","sizeBytes":4815,"isText":false},{"path":"SKILL.md","sizeBytes":28885,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"clean","suspicious":0,"notes":0,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-09-22T13:27:40.232241Z","sha256":"FC3ADDBA9F4547357728095975D0A73EF287BE1D54A53910EE5356C69DCC884E","sizeBytes":34308},"review":null,"source":{"repositoryUrl":"https://github.com/ammawla/encode-toolkit","path":"skills/pipeline-rnaseq","license":"AGPL-3.0","commit":"36836c8725fd4d20d9c851ce314f5151aea5f57c","subtreeSha":"2BC1490CADC85B49FA68EC9F66C80107D921258024D1FE349D4133911D011A0F","lastSyncedAt":"2026-09-29T20:56:54.045383Z"},"reviewedAt":"2026-09-22T13:31:09.958362Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/ammawla/encode-toolkit/tree/main/skills/pipeline-rnaseq"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install ammawla-encode-toolkit@llmmart"},{"target":"git","command":"git clone https://github.com/ammawla/encode-toolkit.git"}]}