ConductScience.org, home

Hosted pipelines

Standard nf-core analyses of your sequencing reads, run at pinned versions. Each run will come with a manifest of exactly what ran: the pipeline and Nextflow versions, each container by digest, each tool's version, the reference files, and every input and output file by size and SHA-256.

  • So far this page covers the four analyses the service will run, the reference genome and software each one uses, and the benchmark that germline runs will be measured against. You can't submit runs yet, and none of our runs has been benchmarked yet.
  • Research use only. Not for diagnosis.

Versions pinned on 17 Nov 2026

The analyses

The two germline analyses come first. Both have run end to end on our own machine, with nf-core's small test data. RNA-seq gene expression and differential expression come next. They use the GENCODE 50 annotation of GRCh38, which we haven't added yet.

Germline variants, whole genome

Coming first

Calls germline SNVs and small indels in whole-genome reads with GATK HaplotypeCaller, on GRCh38.

Pipeline
nf-core/sarek 3.11.0, released 26 Oct 2026
Input
FASTQ reads, or aligned BAM or CRAM files (these start at variant calling)
Reference
GRCh38 no-alt analysis set
Settings
  • No base quality score recalibration (BQSR): it needs known sites such as dbSNP, which the reference files don't include
  • No filtering after HaplotypeCaller (CNNScoreVariants and FilterVariantTranches): calls come back unfiltered, with their quality annotations
  • Reads are trimmed with fastp

Germline variants, exome or panel

Coming first

Calls germline SNVs and small indels within your capture regions with GATK HaplotypeCaller, on GRCh38.

Pipeline
nf-core/sarek 3.11.0, released 26 Oct 2026
Input
FASTQ reads, or aligned BAM or CRAM files (these start at variant calling)
Also needs
A BED file of your capture regions
Reference
GRCh38 no-alt analysis set
Settings
  • No base quality score recalibration (BQSR): it needs known sites such as dbSNP, which the reference files don't include
  • No filtering after HaplotypeCaller (CNNScoreVariants and FilterVariantTranches): calls come back unfiltered, with their quality annotations
  • Reads are trimmed with fastp

RNA-seq gene expression

Coming next

Aligns RNA-seq reads with STAR and quantifies gene and transcript expression with Salmon, on GRCh38 with GENCODE's annotation.

Pipeline
nf-core/rnaseq 3.27.0, released 23 Sep 2026
Input
FASTQ reads
Reference
GRCh38 with the GENCODE 50 annotation, which we haven't added yet

Differential expression

Coming next

Compares gene counts between groups of samples with DESeq2, giving each gene's fold change and adjusted p-value.

Pipeline
nf-core/differentialabundance 2.0.0, released 23 Jun 2026
Input
A gene count table
Reference
GRCh38 with the GENCODE 50 annotation, which we haven't added yet
Settings
  • No gene set analysis

Steps in a germline run

Both germline analyses ran these 17 steps in our test runs from FASTQ. A run from BAM or CRAM starts at variant calling and runs 10 of them.

The steps of a germline run, with the tools each one uses and the inputs that run it
StepWhat it doesTools and licensesRuns from
Reads
FASTPTrims adapters from the reads.
  • fastp 1.1.0 (MIT)
FASTQ only
FASTQCReports read quality: base quality scores, base content, adapter content and overrepresented sequences.
  • FastQC 0.12.1 (GPL-3.0-or-later)
FASTQ only
Alignment
BWAMEM1_INDEXIndexes the reference genome for BWA.
  • BWA 0.7.19 (GPL-3.0-only)
FASTQ only
BWAMEM1_MEMAligns the reads to GRCh38 with BWA-MEM, then sorts them with samtools.
  • BWA 0.7.19 (GPL-3.0-only)
  • samtools 1.22.1 (MIT)
FASTQ only
GATK4_MARKDUPLICATESMarks read pairs that are duplicates of one DNA fragment, so variant calling ignores them, and writes CRAM.
  • GATK 4.7.0.0 (Apache-2.0)
  • samtools 1.24 (MIT)
FASTQ only
Variant calling
CREATE_INTERVALS_BEDSplits the calling regions into intervals that are called in parallel.
  • GNU Awk 5.3.0 (GPL-3.0-or-later)
FASTQ, BAM or CRAM
TABIX_BGZIPTABIX_INTERVAL_SPLITCompresses and indexes each interval's BED file.
  • htslib (bgzip and tabix) 1.24 (MIT AND BSD-3-Clause)
FASTQ, BAM or CRAM
TABIX_BGZIPTABIX_INTERVAL_COMBINEDCompresses and indexes the BED file of all the calling regions.
  • htslib (bgzip and tabix) 1.24 (MIT AND BSD-3-Clause)
FASTQ, BAM or CRAM
GATK4_HAPLOTYPECALLERCalls germline SNPs and indels by local re-assembly of haplotypes, in each interval separately.
  • GATK 4.7.0.0 (Apache-2.0)
FASTQ, BAM or CRAM
MERGE_HAPLOTYPECALLERMerges the intervals' calls into one VCF for each sample.
  • GATK 4.7.0.0 (Apache-2.0)
FASTQ, BAM or CRAM
Quality checks
MOSDEPTHMeasures read depth across the genome.
  • mosdepth 0.3.14 (MIT)
FASTQ only
SAMTOOLS_STATSCollects statistics from the aligned reads.
  • samtools 1.24 (MIT)
FASTQ only
BCFTOOLS_STATSCounts the calls by type and quality.
  • bcftools 1.23.1 (GPL-3.0-only)
FASTQ, BAM or CRAM
VCFTOOLS_SUMMARYCounts the calls under each filter.
  • VCFtools 0.1.17 (LGPL-3.0-only)
FASTQ, BAM or CRAM
VCFTOOLS_TSTV_COUNTGives the transition to transversion ratio by alternative allele count.
  • VCFtools 0.1.17 (LGPL-3.0-only)
FASTQ, BAM or CRAM
VCFTOOLS_TSTV_QUALGives the transition to transversion ratio by quality.
  • VCFtools 0.1.17 (LGPL-3.0-only)
FASTQ, BAM or CRAM
MULTIQCGathers every step's quality figures into one HTML report.
  • MultiQC 1.35 (GPL-3.0-or-later)
FASTQ, BAM or CRAM

Each step runs in a container pinned by digest. A container also holds the runtimes its tools need, such as Java, Perl and Python, under their own licenses.

Reference genome

Germline runs align to the GRCh38 no-alt analysis set (GenBank GCA_000001405.15), from NCBI's sequences for alignment pipelines.

The files of the GRCh38 no-alt analysis set
FileWhat it isSize
GCA_000001405.15_GRCh38_no_alt_analysis_set.fnaSequence (FASTA)3.14 GB
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.faiFASTA index7.8 kB
GCA_000001405.15_GRCh38_no_alt_analysis_set.dictSequence dictionaryMade with GATK 4.7.0.0 CreateSequenceDictionary23.4 kB
GCA_000001405.15_GRCh38_no_alt_analysis_set.bedCalling regions (BED)Made from the FASTA index with GNU Awk 5.3.0, as sarek 3.11.0 makes its calling regions4.89 kB
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.ambBWA index17.5 kB
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.annBWA index27.7 kB
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.bwtBWA index3.1 GB
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.pacBWA index775 MB
GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.saBWA index1.55 GB
SHA-256 of each file
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna9cce8b926416dd96b152deea85188495b75f7ac8d634cc723a017067be8702b7
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.fai502a1b8fb73ccd53285c28a0f12df90c818b4fe3de1e862ef47c593ef1a0a4b4
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.dict9b321bde7d702bbfe78e353dcb6bb9f163a7c109c1298e6e498b10ca716819ac
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.bed8bb31f29b3127b3aea3f20ead931d5a42d545b745919a43c696696510e390ac6
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.amb937fb382f79b8382cb3be2f13dc88d29147661919bb8c398d307b1bcc52c0890
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.ann71b231aee7bd25800756f1165d2cf9dd41501ca248a147de70c2bb83f2849f1c
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.bwtd32fa528b2bb021e624aeb42bb4eaf2348a7b874030748f35354287363e66e59
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.pac1b598382ddcf7e4ce8b120b22ba4c66ec837795e22b5a9caa21cc61797a6561a
  • GCA_000001405.15_GRCh38_no_alt_analysis_set.fna.sa6649a799aac70c12ceb453895a3ca7b5e347f7445276a0e51acf6d6c48380e3e

Benchmark

A germline analysis becomes available only once its calls pass these checks against Genome in a Bottle (GIAB) v4.2.1, NIST's benchmark variant calls for the reference samples HG001 to HG005. We use v4.2.1 because the published results behind the thresholds used it. Each whole-genome threshold is a published F1 on the same sample, less half of what dropping from 30x to 20x cost the same caller, so a run fails only if it falls further below the published result than that. A repeat run must give identical calls.

The checks a germline analysis must pass against Genome in a Bottle v4.2.1, with their thresholds and sources
CheckWhat is runThresholdWhere the threshold comes fromResult
Whole-genome analysis: HG002, chromosome 20HG002's chromosome 20 from reads at about 30x, scored with RTG vcfeval against GIAB v4.2.1 inside its benchmark regionsSNP F1 at least 99.13%, indel F1 at least 98.46%Not run yet
Whole-genome analysis: HG005, chromosome 20HG005's chromosome 20, scored the same waySNP F1 at least 99.14%, indel F1 at least 98.74%Not run yet
Exome analysis: HG001, HG002 and HG003GIAB's exome reads for the three samples, scored with hap.py against GIAB v4.2.1 inside the capture regionsSNV F1 at least 98.99%, 98.74% and 98.73%; indel F1 at least 89.84%, 91.99% and 92.46%Not run yet
ReproducibilityHG002's chromosome 20 run twice with the same inputs, versions and settingsEvery call's position, alleles, filter and genotype identicalNot run yet
Starting from BAM or CRAMHG002's chromosome 20 started from a BAM or CRAM at variant calling, scored the same wayThe HG002 thresholds, and within 0.10 percentage points of SNP F1 and 0.50 percentage points of indel F1 of the same run from FASTQNot run yet
Other sequencersHG002's chromosome 20 from HiSeq X, Element AVITI and DNBSEQ reads, scored the same wayHiSeq X: SNP F1 at least 99.18%, indel F1 at least 98.10%. Element AVITI: the HG002 thresholds. DNBSEQ: SNP F1 at least 99.07%, indel F1 at least 98.46%Not run yet

We keep a copy of GIAB's v4.2.1 files for the five samples, with GIAB's checksums, as release giab-2026-11-18-0.2.0, so every check scores against the same files.

About this data

  • Benchmark data: Genome in a Bottle, National Institute of Standards and Technology (NIST). The v4.2.1 benchmark's paper. Works of NIST employees are not subject to copyright protection in the United States.
  • Reference genome: GRCh38 no-alt analysis set from the National Center for Biotechnology Information (NCBI), Genome Reference Consortium assembly GRCh38. NCBI does not endorse this service.
  • Pipelines and their licenses:The tools inside each pipeline, such as GATK, STAR, Salmon and DESeq2, carry their own licenses.
  • Versions pinned on 17 Nov 2026. This page's data were taken on 18 Nov 2026.