QUANEXAS
Start learning
  1. Home
  2. Reference
  3. Practice datasets

Reference

Practice datasets

Real public data, ready to practise on. Every file comes unchanged from an official source (NCBI, ENA, UniProt, the PDB or Bioconductor), with its citation, licence and checksum. The code labs in the study topics load these files directly.

  • 10 hosted datasets
  • 31 files, 11.0 MB in total
  • 8 large datasets with download commands

Viral genome · Level 1

SARS-CoV-2 reference genome

Severe acute respiratory syndrome coronavirus 2 (isolate Wuhan-Hu-1)

The 29,903-base reference genome of the virus that causes COVID-19, with its gene annotation in two formats. Small enough to read in one go, real enough to analyse properly.

Good for practising

  • Reading FASTA by hand and with Biopython
  • GC content and base composition
  • Extracting genes from GFF3 coordinates
  • Translating coding sequences such as the spike gene
FileFormatSizeContents
NC_045512.2.fastaComplete genome sequence. FASTA 30 KB 1 record, 29,903 bases
NC_045512.2.gff3Genes, coding sequences and mature peptides, with 1-based coordinates. GFF3 14 KB 59 features, 11 genes
NC_045512.2.gbThe same record as a GenBank flat file, with features and translations. GenBank 77 KB 29,903 bases, 207 annotated features
Cite
Wu F, et al. A new coronavirus associated with human respiratory disease in China. Nature 579, 265–269 (2020). doi:10.1038/s41586-020-2008-3

Gene and protein · Level 1

Human β-globin (HBB) gene and protein

Homo sapiens

The messenger RNA and protein for haemoglobin subunit beta, the gene behind sickle-cell disease. The coding sequence runs from base 51 to 494 and translates to the 147-residue UniProt protein.

Good for practising

  • Transcription and translation
  • Finding a coding sequence in a GenBank record
  • Predicting the effect of the sickle-cell variant (c.20A>T, p.Glu7Val)
FileFormatSizeContents
NM_000518.5.fastaHBB mRNA (RefSeq NM_000518.5), written as DNA. FASTA 700 B 1 record, 628 bases
NM_000518.5.gbThe annotated mRNA record, including exon and CDS features. GenBank 22 KB 628 bases, 48 annotated features
P68871.fastaHBB protein from UniProtKB/Swiss-Prot (P68871). FASTA 236 B 1 record, 147 amino acids
Cite
NCBI Reference Sequence NM_000518.5; UniProt Consortium. UniProt: the Universal Protein Knowledgebase. Nucleic Acids Research (2025). doi:10.1093/nar/gkae1010

Genome region and annotation · Level 1

Human β-globin locus on chromosome 11 (GRCh38)

Homo sapiens

An 80,000-base stretch of human chromosome 11 holding the β-globin gene cluster (HBE1, HBG2, HBG1, HBD and HBB), the pseudogene HBBP1 and the locus control region, with NCBI’s annotation of the same region. Every globin gene here is on the minus strand.

Good for practising

  • Reading a eukaryotic GFF3 annotation: genes, transcripts, exons and CDS
  • Converting chromosome coordinates into positions in a region file
  • Splicing and translating a minus-strand gene
  • Finding a known variant, the sickle-cell base at 5,227,002, on the reference genome

The FASTA file starts at chromosome position 5,220,001, so chromosome position p is index p − 5,220,001 of its sequence. The GFF3 file keeps whole-chromosome coordinates, as real annotation files do.

FileFormatSizeContents
NC_000011.10_5220001-5300000.fastaChromosome 11 positions 5,220,001 to 5,300,000 (RefSeq NC_000011.10, GRCh38.p14). FASTA 79 KB 1 record, 80,000 bases
NC_000011.10_5220001-5300000.gff3Every line of NCBI’s chromosome 11 annotation that overlaps the region, unchanged. Coordinates are whole-chromosome positions. GFF3 122 KB 328 features, 6 genes
Cite
Schneider VA, et al. Evaluation of GRCh38 and de novo haploid genome assemblies demonstrates the enduring quality of the reference assembly. Genome Research 27, 849–864 (2017). doi:10.1101/gr.213611.116
Used in
Genes, genomes and chromosomes, Mutations and genetic variation, Gene expression and regulation, File formats and coordinates in practice (in preparation)

Bacterial genome · Level 1

Escherichia coli K-12 MG1655 genome and annotation

Escherichia coli str. K-12 substr. MG1655

The 4.6 Mb reference genome of the most studied laboratory bacterium, with 4,506 annotated genes, compressed with gzip as genomes usually are. Two of its lac operon proteins come from UniProt, to check translations against.

Good for practising

  • Reading gzipped files
  • Genome-wide base composition and Chargaff’s rules
  • Gene lengths and coding density from GFF3
  • Extracting a minus-strand gene (lacZ) and checking it against UniProt
  • Hydropathy plots of a membrane protein (LacY)
  • k-mer counting and ORF finding at genome scale
FileFormatSizeContents
NC_000913.3.fasta.gzComplete chromosome sequence. FASTA (gzip) 1.3 MB 1 record, 4,641,652 bases
NC_000913.3.gff3.gzGene, CDS, RNA and other features. GFF3 (gzip) 378 KB 9,523 features, 4,506 genes
P00722.fastaβ-galactosidase, LacZ, from UniProtKB/Swiss-Prot (P00722). FASTA 1.1 KB 1 record, 1,024 amino acids
P02920.fastaLactose permease, LacY, from UniProtKB/Swiss-Prot (P02920). FASTA 523 B 1 record, 417 amino acids
Cite
Blattner FR, et al. The complete genome sequence of Escherichia coli K-12. Science 277, 1453–1462 (1997). doi:10.1126/science.277.5331.1453

Sequencing reads · Level 3

E. coli evolution experiment: Illumina reads and reference

Escherichia coli B str. REL606 lineage

The first 10,000 read pairs (2 × 150 bp, Illumina HiSeq 2500) from run SRR2584863 of Richard Lenski’s long-term evolution experiment, plus the REL606 ancestor genome to align them to. The same run is used in Data Carpentry’s genomics lessons.

Good for practising

  • Reading FASTQ and decoding Phred quality scores
  • Spotting poor-quality reads and N calls at the start of a run
  • Read trimming
  • Aligning reads to a reference and calling variants

These are the first reads in the run file, not a random sample. Reads from the edge of a flow cell are often poorer, which makes them useful for practising quality control.

FileFormatSizeContents
SRR2584863_1.first10k.fastq.gzRead 1 of the first 10,000 pairs, unchanged from ENA. FASTQ (gzip) 1.1 MB 10,000 reads of 150 bases
SRR2584863_2.first10k.fastq.gzRead 2 of the same pairs. FASTQ (gzip) 1.2 MB 10,000 reads of 150 bases
NC_012967.1.fasta.gzReference genome of the ancestral strain REL606. FASTA (gzip) 1.3 MB 1 record, 4,629,812 bases
Cite
Tenaillon O, et al. Tempo and mode of genome evolution in a 50,000-generation experiment. Nature 536, 165–170 (2016). doi:10.1038/nature18959
Used in
Read quality control and trimming (in preparation), Read mapping, SAM and BAM (in preparation), Variant calling and annotation (in preparation)

Bulk RNA-seq · Level 4

pasilla RNA-seq gene counts

Drosophila melanogaster (S2-DRSC cells)

Read counts for 14,599 genes in seven samples: four untreated and three in which the splicing regulator pasilla was knocked down by RNAi. A classic first differential expression dataset.

Good for practising

  • Loading a count matrix with pandas
  • Library sizes and normalisation
  • Sample quality control with PCA
  • Differential expression and the effect of sequencing type (single-end or paired-end) as a batch factor

Sample names in the counts table (untreated1 …) match the annotation file names without the “fb” suffix (untreated1fb …).

FileFormatSizeContents
pasilla_gene_counts.tsvGenes in rows, samples in columns, raw counts. TSV 487 KB 14,599 rows × 8 columns
pasilla_sample_annotation.csvCondition, library type and read numbers for each sample. CSV 527 B 7 rows × 6 columns
Cite
Brooks AN, et al. Conservation of an RNA regulatory map between Drosophila and mammals. Genome Research 21, 193–202 (2011). doi:10.1101/gr.108662.110
Used in
Gene expression and regulation, Data analysis with pandas, NumPy and plots (in preparation), Differential expression and enrichment (in preparation)

3D structures · Level 1

DNA, RNA and nucleosome 3D structures

Synthetic DNA, Saccharomyces cerevisiae tRNA, and Xenopus laevis histones with human DNA

Four landmark crystal structures: a full turn of right-handed B-DNA, left-handed Z-DNA at atomic resolution, the folded L-shape of a transfer RNA and a nucleosome, DNA wrapped around eight histone proteins. These files drive the 3D viewers in the study topics.

Good for practising

  • Reading PDB coordinate files
  • Measuring base-pair and helix geometry
  • Finding hydrogen bonds between bases
  • Comparing DNA helix forms
FileFormatSizeContents
1BNA.pdbB-DNA dodecamer CGCGAATTCGCG (Drew et al. 1981), 1.9 Å. PDB 75 KB 486 atoms, chains A, B
2DCG.pdbLeft-handed Z-DNA, d(CGCGCG) (Wang et al. 1979), 0.9 Å. PDB 59 KB 269 atoms, chains A, B
1EHZ.pdbYeast tRNA-Phe with its modified bases (Shi and Moore 2000), 1.93 Å. PDB 223 KB 1,661 atoms, chain A
1KX5.pdbNucleosome core particle: 147 bp of DNA around a histone octamer (Davey et al. 2002), 1.9 Å. PDB 1.4 MB 13,625 atoms, chains I, J, A, B, C, D, E, F, G, H
Cite
Drew HR, et al. Structure of a B-DNA dodecamer: conformation and dynamics. PNAS 78, 2179–2183 (1981). doi:10.1073/pnas.78.4.2179
Used in
DNA, RNA and the central dogma, Genes, genomes and chromosomes, Structural bioinformatics (in preparation)

3D structures · Level 1

Protein 3D structures: ubiquitin, myoglobin and haemoglobin

Homo sapiens (ubiquitin, haemoglobin) and Physeter catodon (sperm whale myoglobin)

Four protein crystal structures that cover every level of protein structure: ubiquitin, with both helices and sheets; myoglobin with oxygen bound to its heme; the four-chain haemoglobin molecule; and sickle-cell haemoglobin, in which a mutant valine sticks into a neighbouring molecule.

Good for practising

  • Reading PDB files with Biopython
  • Secondary structure, backbone hydrogen bonds and Ramachandran plots
  • Measuring contacts, such as the bonds to a heme iron
  • Solvent accessibility and the hydrophobic core
  • Comparing a mutant protein with the normal one
FileFormatSizeContents
1UBQ.pdbHuman ubiquitin (Vijay-Kumar et al. 1987), 1.8 Å. PDB 77 KB 602 atoms, chain A
1MBO.pdbSperm whale myoglobin with O₂ bound to the heme (Phillips 1980), 1.6 Å. PDB 157 KB 1,277 atoms, chain A
4HHB.pdbHuman deoxyhaemoglobin: two α and two β chains (Fermi et al. 1984), 1.74 Å. PDB 463 KB 4,558 atoms, chains A, B, C, D
2HBS.pdbSickle-cell deoxyhaemoglobin S, two molecules in contact (Harrington et al. 1997), 2.05 Å. PDB 851 KB 9,104 atoms, chains A, B, C, D, E, F, G, H
Cite
Berman HM, et al. The Protein Data Bank. Nucleic Acids Research 28, 235–242 (2000). doi:10.1093/nar/28.1.235
Used in
Proteins: from amino acids to 3D structure, Structural bioinformatics (in preparation)

Variants and genotypes (VCF) · Level 1

Human β-globin variants: ClinVar and 1000 Genomes

Homo sapiens

Two real VCF files for the β-globin region of chromosome 11 (GRCh38). ClinVar’s interpretations of 2,105 variants, from the sickle-cell allele to the mutations behind β-thalassaemia, and the genotypes of all 3,202 people in the high-coverage 1000 Genomes data at 176 variants around HBB, with the population of each unrelated person.

Good for practising

  • Reading VCF headers, INFO fields and genotype columns
  • Naming variants in HGVS and converting between genomic and coding positions on the minus strand
  • Predicting variant consequences and comparing them with ClinVar
  • Allele frequencies, Hardy–Weinberg proportions and differences between populations
  • Transitions, transversions and CpG mutation hotspots

Both files were cut from the full chromosome 11 files with tabix-style range requests, keeping the original header and every overlapping record. ClinVar calls the chromosome 11 and the 1000 Genomes file chr11. Multi-allelic sites are split into one record per allele. The 1000 Genomes panel lists the 2,504 unrelated people; the other 698 people in the VCF are their relatives.

FileFormatSizeContents
clinvar_20261019_chr11_5220001-5300000.vcf.gzEvery record of ClinVar’s 19 October 2026 GRCh38 release that overlaps chromosome 11 positions 5,220,001 to 5,300,000, unchanged. VCF (gzip) 124 KB 2,105 variant records
1kGP_high_coverage_chr11_5225001-5230000.vcf.gzPhased genotypes of 3,202 people at every variant in chromosome 11 positions 5,225,001 to 5,230,000, from the 1000 Genomes high-coverage phased panel, unchanged. VCF (gzip) 75 KB 176 variant records, genotypes for 3,202 people
integrated_call_samples_v3.20130502.ALL.panelThe 2,504 unrelated people of 1000 Genomes phase 3, with population, continental group and sex. TSV 54 KB 2,504 people, 4 columns
Cite
Byrska-Bishop M, et al. High-coverage whole-genome sequencing of the expanded 1000 Genomes Project cohort including 602 trios. Cell 185, 3426–3440 (2022); Landrum MJ, et al. ClinVar. Nucleic Acids Research 53, D1313–D1321 (2025). doi:10.1016/j.cell.2022.08.004
Used in
Mutations and genetic variation, Gene expression and regulation, Variant calling and annotation (in preparation), File formats and coordinates in practice (in preparation)

3D structures · Level 1

Protein–DNA complexes: p53, the Lac repressor and TBP

Homo sapiens (p53, TBP) and Escherichia coli (Lac repressor)

Three crystal structures of proteins gripping DNA: the DNA-binding domain of the tumour suppressor p53, whose most common cancer mutations sit at its DNA-binding surface; the Lac repressor on its operator; and the TATA-binding protein bending a promoter.

Good for practising

  • Measuring protein–DNA contacts in PDB files
  • Major-groove and minor-groove recognition
  • Relating cancer mutations to structure
  • Allostery: an inducer pocket far from the DNA
FileFormatSizeContents
1TSR.pdbThe p53 core domain bound to DNA (Cho et al. 1994), 2.2 Å. PDB 514 KB 5,444 atoms, chains E, F, A, B, C
1EFA.pdbA dimeric Lac repressor bound to a symmetric operator and the anti-inducer ONPF (Bell and Lewis 2000), 2.6 Å. PDB 683 KB 7,776 atoms, chains D, E, A, B, C
1CDW.pdbHuman TBP core bound to the adenovirus major late promoter TATA box (Nikolov et al. 1996), 1.9 Å. PDB 215 KB 2,071 atoms, chains B, C, A
Cite
Berman HM, et al. The Protein Data Bank. Nucleic Acids Research 28, 235–242 (2000). doi:10.1093/nar/28.1.235
Used in
Mutations and genetic variation, Gene expression and regulation, Structural bioinformatics (in preparation)

For real projects

Large datasets, straight from the source

Too big to host here, and better downloaded from the official archive. Each entry gives a working command for a terminal.

Sequencing reads · about 375 MB (two gzipped FASTQ files)

Full Lenski sequencing run SRR2584863

All 1.55 million read pairs behind the hosted sample above, for a complete QC, alignment and variant-calling exercise.

curl -O https://ftp.sra.ebi.ac.uk/vol1/fastq/SRR258/003/SRR2584863/SRR2584863_1.fastq.gz
curl -O https://ftp.sra.ebi.ac.uk/vol1/fastq/SRR258/003/SRR2584863/SRR2584863_2.fastq.gz

Reference genome · about 880 MB gzipped, 3.1 GB uncompressed

Human reference genome GRCh38 (Ensembl)

The primary assembly: chromosomes plus unlocalised scaffolds, without alternate haplotypes. The usual choice for read alignment.

curl -O https://ftp.ensembl.org/pub/current_fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz

Bulk RNA-seq · a few MB as an R package

airway RNA-seq (GEO GSE52778)

Airway smooth muscle cells from four donors, untreated or treated with dexamethasone; the dataset used in the DESeq2 tutorials.

# In R
if (!require("BiocManager")) install.packages("BiocManager")
BiocManager::install("airway")
library(airway)
data(airway)

Single-cell RNA-seq · about 7.6 MB

PBMC 3k single-cell RNA-seq (10x Genomics)

About 2,700 peripheral blood mononuclear cells; the standard first dataset in the Seurat and Scanpy tutorials.

curl -O https://cf.10xgenomics.com/samples/cell-exp/1.1.0/pbmc3k/pbmc3k_filtered_gene_bc_matrices.tar.gz
tar -xzf pbmc3k_filtered_gene_bc_matrices.tar.gz

Human variation · about 205 MB

1000 Genomes Project, chromosome 22 genotypes

Phase 3 genotypes for 2,504 people from 26 populations on chromosome 22, the smallest autosome file.

curl -O https://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/ALL.chr22.phase3_shapeit2_mvncall_integrated_v5b.20130502.genotypes.vcf.gz

Clinical variants · about 190 MB, updated weekly

ClinVar variants (GRCh38)

Variants with reported clinical significance, for annotation and interpretation exercises.

curl -O https://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/clinvar.vcf.gz

Variant benchmark · about 150 MB VCF plus BED

Genome in a Bottle HG001 benchmark

High-confidence variant calls for sample NA12878, used to measure the accuracy of a variant-calling pipeline.

BASE=https://ftp-trace.ncbi.nlm.nih.gov/ReferenceSamples/giab/release/NA12878_HG001/latest/GRCh38
curl -O $BASE/HG001_GRCh38_1_22_v4.2.1_benchmark.vcf.gz
curl -O $BASE/HG001_GRCh38_1_22_v4.2.1_benchmark.bed