Skip to content

Test Datasets

This page provides a set of basic test files. To download any file, simply click its blue file name.

Biofiles Test Datasets

Gene Annotation

  • gencode_with_m6a.gff3.gz
    GENCODE gene annotation augmented with m6A modification features, bgzip-compressed GFF3 for feature/gene track rendering.
    Build: refbox or bgzip (block-compress the GFF3)
  • gencode_with_m6a.gff3.gz.tbi
    Tabix index for the GFF3 above, enabling fast range queries.
    Build: refbox or tabix -p gff gencode_with_m6a.gff3.gz

Reference Genome

  • GRCh38.genome.fa.gz
    GRCh38 reference genome sequence, bgzip-compressed FASTA.
    Build: refbox or bgzip GRCh38.genome.fa
  • GRCh38.genome.fa.gz.fai
    FASTA index for the genome, for random sequence access.
    Build: refbox or samtools faidx GRCh38.genome.fa.gz
  • GRCh38.genome.fa.gz.gzi
    bgzip index for the compressed FASTA, for coordinate-to-offset lookup.
    Build: refbox or samtools faidx GRCh38.genome.fa.gz (creates .gzi for bgzipped FASTA)

Alignments (BAM)

Coverage Signal (bigWig)

  • hg38.caRNA.bw
    hg38 chromatin-associated RNA signal, bigWig coverage track.
    Build: deeptools bamCoverage (or bedGraphToBigWig)

Variants (VCF)

  • hg38.clinvar.vcf.gz
    ClinVar variants on hg38, bgzip-compressed VCF for variant track rendering.
    Build: refbox or bgzip hg38.clinvar.vcf
  • hg38.clinvar.vcf.gz.tbi
    Tabix index for the ClinVar VCF.
    Build: refbox or tabix -p vcf hg38.clinvar.vcf.gz

m6A Sites (bigBed)

  • m6A_single_base.BigBed
    Single-base-resolution m6A sites (v2), bigBed feature track.
    Build: refbox or bedToBigBed

Repeat Regions (BED)

  • repeats.bed
    Repeat-region annotations, plain BED.
    Build: refbox or (source annotation export)
  • repeats.bed.gz
    Repeat-region annotations, bgzip-compressed and coordinate-sorted BED.
    Build: refbox or sort -k1,1 -k2,2n repeats.bed | bgzip
  • repeats.bed.gz.tbi
    Tabix index for the repeats BED.
    Build: refbox or tabix -p bed repeats.bed.gz

Genome 3D Structure (g3d)

  • GM12878.g3d
    GM12878 Genome 3D Structure file, for rendering Genome 3D Structure.
    Build: g3dtools

In Silico Mutagenesis (ISM)

Custom Reference Test Dataset (sacCer3)

A complete custom reference transcriptome + genome test dataset.

Reference Genome

  • sacCer3.genome.fa.gz
    sacCer3 (S. cerevisiae) reference genome, bgzip-compressed FASTA.
    Build: refbox or bgzip sacCer3.genome.fa
  • sacCer3.genome.fa.gz.fai
    FASTA index for the genome.
    Build: refbox or samtools faidx sacCer3.genome.fa.gz
  • sacCer3.genome.fa.gz.gzi
    bgzip index for the compressed genome FASTA.
    Build: refbox or samtools faidx sacCer3.genome.fa.gz
  • sacCer3.chrom.sizes
    Chromosome sizes for sacCer3.
    Build: refbox or samtools faidx then cut cols 1–2

Reference Transcriptome

Gene Annotation

RNAcentral Annotation

RBrowser Index (A transcript-coordinate-based index)

  • annotation.rbi
    a transcript-coordinate-based index file (.rbi) for the custom reference, generated by refbox.
    Build: refbox

RBrowser Render Benchmark

Metric: wall-clock time from load start to render complete, and JS-heap growth, for each track/plugin type, loaded from remote DataHub over the network. Each type measured 3 times; values below are the average (min–max shown).

Generated: 2026-07-09T21:54:06.775Z

Test machine

Item Value
Laptop Alienware 18 Area-51 AA18250
CPU Intel(R) Core(TM) Ultra 9 275HX (24 logical cores)
RAM 31 GB
OS Windows (win32 10.0.26200)
Browser Chrome 150.0.7871.101 (Playwright channel:'chrome', headed, real GPU)
Node v20.19.6
RBrowser @rbrowser/main (packaged renderer) + app source src/renderer/*
Data source remote — data.rbrowser.org DataHub
Locus VAMP3-201 — chr1:7,766,295-7,786,432 (GRCh38)

Test files

Type File Size
BAM ENCFF827RKM.bam (+ index 2.18 MB) 1.48 GB + 2.18 MB idx
BigWig hg38.caRNA.bw 472.0 MB
BigBed m6A_single_base_v2.BigBed 21.2 MB
r2s (RNA 2D) ENST00000054666.r2s 110.2 KB
g3d (Genome 3D) GM12878.g3d 19.2 MB

BAM/BigWig are large whole-genome files but are read via their index (BAI / bbi R-tree), so only the bytes overlapping the locus are fetched — not the full file.

Method

  • Real Chrome 150.0.7871.101 driven by Playwright loads a standalone harness page that imports the actual RBrowser renderer modules (no app code modified) and drives each one against the real remote URLs above.
  • Memory = JS-heap (JSHeapUsedSize via CDP) growth during the measured op: a forced GC is taken as baseline immediately before the op, and the heap is sampled every 40 ms during it to capture the peak. Δheap = peak − baseline. For BAM/BigWig/BigBed the baseline is after browser boot, so Δheap is the channel's marginal cost; for r2s/g3d it is the render's cost over a fresh page.
  • Cold network, warm engine: the HTTP cache is disabled (CDP Network.setCacheDisabled), so every run performs a real network fetch of the region bytes. One extra warm-up iteration per type is run and discarded to absorb one-time GPU-shader/JIT compilation, so the 3 recorded runs measure steady-state render cost with a genuinely cold data fetch each time.

Results (average of 3 runs)

Type Render time (avg) min–max JS-heap Δ (avg) Records
BAM 858.8 ms 807.1–904.8 ms 3.9 MB 852
BigWig 255 ms 250.6–262.8 ms 1.2 MB 1200
BigBed 228.5 ms 217.3–236.3 ms 0.9 MB 133
r2s (RNA 2D) 371.9 ms 366.8–377 ms 2.1 MB —
g3d (Genome 3D) 780.9 ms 754.4–803.8 ms 18.6 MB —