Slide deck
Backsc-SPLASH — Reference-Free Discovery in Barcoded Single-Cell Sequencing
A new method, sc-SPLASH, finds genes and cell types directly from raw single-cell sequencing reads, without a reference genome. In a freshwater sponge with no matching reference at all, it found five previously unknown secreted-protein genes, absent from every database checked.
Single-cell genomics · Nature Biotechnology, 22 Apr 2026
A Single-Cell Method That Needs No Reference Genome at All
sc-SPLASH finds genes and cell types directly from raw single-cell sequencing reads, without a reference genome.
sc-SPLASH · reference-free discovery1 / 18
What happened
Four numbers that carry the paper
- ~50×faster than UMI-tools, for the barcode and UMI preprocessing step (BKC)
- 11benchmark datasets on which sc-SPLASH ran faster and used less memory than both Cell Ranger and STARsolo
- 5new secreted-protein genes in a freshwater sponge, traced from an anchor absent from its genome assembly and from NCBI
- 47homologous YYD-repeat genes found across five tunicate genome assemblies
sc-SPLASH · reference-free discovery2 / 18
The argument · the method
Anchor and target: statistics, not a reference
A reference genome is not required to find real, regulated biology in sequencing data — only statistics is.
sc-SPLASH takes a fixed k-mer from each read, the anchor, and the varying k-mer that follows it, the target, and asks whether the distribution of targets for a given anchor depends on which sample (which cell, in barcoded data) it came from. Nothing in this process ever looks at a reference genome.
- Effect size, 0 to 1How separable the two target distributions are.
- A closed-form p-valueThe OASIS test keeps multiple testing under control without simulation.
sc-SPLASH · reference-free discovery3 / 18
The argument · the pipeline
Three stages, and only the first is new
- 1BKCBarcode and UMI preprocessing — the one new stage
- 2–3SPLASH2, reusedMerging counts, computing p-values, multiple-testing correction
- +Extendor, optionalAligning or searching it gives a gene name, splice site or Pfam domain
- Anchors are called significant at an effect size above 0.2 and a Benjamini–Yekutieli-corrected p-value below 0.05. The extendor step is one the reference-free part of the pipeline does not need and does not do.
sc-SPLASH · reference-free discovery4 / 18
What happened · preprocessing
BKC: about 50 times faster than UMI-tools
- ~50×faster than UMI-tools, stated in the paper's own abstract
- UMI-tools is the standard network-based method for correcting UMI sequencing errors and deduplicating reads — the job BKC is benchmarked against.
sc-SPLASH · reference-free discovery5 / 18
What happened · the benchmark
Faster and lighter on all 11 datasets
Less runtime and less memory than both Cell Ranger v8.0.1 and STARsolo v2.7.10b in every case (Figure 1c).
The benchmark compares two different jobs
Cell Ranger and STARsolo only align reads; Seurat or Scanpy is still needed afterwards for differential expression. sc-SPLASH's runtime already includes calling significant anchors. So the gap is, if anything, an understatement, not an overstatement — but the two are also not finishing the same job.
sc-SPLASH · reference-free discovery6 / 18
What happened · Tabula Sapiens
V-set tops every Pfam domain
Of every Pfam protein domain found in the Tabula Sapiens data, V-set (the immunoglobulin variable-region fold) has the highest average entropy, the highest average effect size, and the largest number of unaligned extendors (Figure 1d).
- ~0.90V-set's effect size, consistent with Figure 1d
- Its entropy was given as 2.16 in the earlier preprint, but the published figure's own axis stops short of that and places V-set near 1.8 — so the precise entropy is not carried here.
sc-SPLASH · reference-free discovery7 / 18
What happened · immune receptors
5,212 in-frame V(D)J sequences, across the Tabula Sapiens tissues
- 4,098from plasma cells
- 1,114from B cells
Where most were found Sequences Spleen 1,632 Lymph node 1,484 Salivary gland 409 Bone marrow 378 Blood 332 sc-SPLASH · reference-free discovery8 / 18
What happened · spatial data
A cancer mutation, spatially resolved
In a human squamous-cell-carcinoma Visium sample, the strongest signal is a mitochondrial MT-ND4 double mutation, expressed mainly, but not only, inside the tumour region.
- 0.757effect size, the strongest signal in the sample
- ChrM 11,413–11,466where the double mutation sits
sc-SPLASH · reference-free discovery9 / 18
What happened · splicing
The same splice choice, two phyla apart
An RPS24 exon, included in one cell type and excluded in its neighbour, in both animals.
Tissue Exon included Exon excluded Human fetal intestine Epithelium Stromal cells Electric eel, electric organ Electrocytes The insulating septa around them sc-SPLASH · reference-free discovery10 / 18
What happened · a freshwater sponge
An anchor missing from the sponge's own reference genome
A 27-nucleotide anchor, absent from both the Spongilla lacustris assembly and NCBI, led to five newly named genes, Granrep1–5.
- 6.2entropy, the highest of any anchor in the dataset
- 667distinct targets
- 5newly named genes, Granrep1–5
sc-SPLASH · reference-free discovery11 / 18
What happened · a freshwater sponge
Granrep1–5 sit mainly in granulocytes and amoebocytes
- 47 / 74granulocytes profiled that express it
- 12 / 49amoebocytes that express it
- 30 / 34cells co-expressing it with the granulocyte marker Acp5, by RNA-FISH
sc-SPLASH · reference-free discovery12 / 18
What happened · a tunicate
A parallel repeat in a second phylum
The tunicate Ciona's own high-diversity anchor points to “YYD” genes, expressed in circulating haemocytes.
- 4.80–5.97entropy: the only repeat anchor above 4 among nine invertebrates surveyed
- 47homologous genes across five tunicate genome assemblies
- 12 hafter hatching: expression peaks at metamorphosis
- The 47, by assembly: 5 in C. robusta HT, 5 in C. robusta KH, 15 in C. intestinalis Plymouth, 18 in C. intestinalis Roscoff, 4 in C. savignyi (Supplementary Note).
sc-SPLASH · reference-free discovery13 / 18
The argument · a second look
A repeated protein in immune cells is not the same as a proven immune role
The parallel across two phyla is real; a causal immune function is not yet shown for either.
In the sponge Granrep expression Over development, gemmule (day 0) to day 12 Rises After immune challenge with lipopolysaccharide or a cyclic dinucleotide No evident difference, in a preliminary test - Both rest on the predecessor bioRxiv preprint's own text, not re-read in the paywalled final text.
sc-SPLASH · reference-free discovery14 / 18
What happened · the evidence
How Firmly Each Part Stands
How firm What Confirmed The method and its findings, peer-reviewed in Nature Biotechnology (PMID 42020848) Confirmed Less runtime and memory than both tools on all 11 datasets (Figure 1c) Confirmed BKC about 50 times faster than UMI-tools (the abstract) Confirmed V-set first among Pfam domains on entropy, effect size and unaligned extendors (Figure 1d) Reported, and superseded V-set's preprint entropy of 2.16; its effect size of about 0.90 matches the figure Confirmed, but not verifiable here Granrep's rise over development, and no evident change after immune challenge (preprint text) sc-SPLASH · reference-free discovery15 / 18
Timeline
Four Years, One Method Growing Up
- 24 Jun 2022The original SPLASH algorithm is first posted to bioRxiv.
- 17 Mar 2023SPLASH2, the fast, scalable engine sc-SPLASH's later stages reuse, is posted to bioRxiv.
- 7 Dec 2023SPLASH is published, peer-reviewed, in Cell.
- 23 Sep 2024SPLASH2 is published, peer-reviewed, in Nature Biotechnology.
- 24 Dec 2024The sc-SPLASH preprint extends SPLASH2 to barcoded single-cell and spatial data for the first time.
- 22 Apr 2026The peer-reviewed, condensed version is published as a Brief Communication in Nature Biotechnology.
- 25 Aug 2026Europe PMC records no independent citation, response or replication of the paper.
sc-SPLASH · reference-free discovery16 / 18
What others add
Somebody Else's Atlases, and No Response Yet
sc-SPLASH re-analyses each human dataset without touching its original conclusions.
Dataset Whose it is Tabula Sapiens atlas, >400,000 cells Tabula Sapiens Consortium (Science, 2022) Squamous-cell-carcinoma Visium sample Ji and colleagues (Cell, 2020) Fetal-intestine Visium sample Fawkner-Corbett and colleagues (Cell, 2021) - No response yet. Europe PMC records no independent citation, response or replication as of 25 August 2026, four months after publication.
sc-SPLASH · reference-free discovery17 / 18
Conclusion
What to Take Away
The speed and memory result is the firmest part
Less runtime and memory than both Cell Ranger and STARsolo on all 11 datasets, and BKC about 50 times faster than UMI-tools — read off the paper's own figures and abstract.
The biology is real; the immune story is not yet closed
Five new sponge genes and a parallel tunicate repeat family are genuine discoveries, both in immune-like cells. That they respond to immune challenge is not yet shown — the one preliminary test of that ran negative.
Four months old, and not yet independently checked
Peer-reviewed, with public code, but as of August 2026 no independent group has cited, replicated or disputed it — ordinary for a four-month-old Brief Communication, not a mark against it.
sc-SPLASH · reference-free discovery18 / 18