Explore the data behind our papers in your browser
Each app runs entirely in your browser, with no login, no install, and nothing sent to a server. Open one and go straight to the gene, the cell, or the clone you care about.
Pediatric ALL treatment resistance
Why a cancer with one of the lowest mutation burdens needs one of the most complex treatment regimens. Compare what bulk sequencing reports against what single cells actually carry, watch allele frequencies move under prednisolone and daunorubicin, walk the single-cell phylogenies of 165 genomes, and follow the clones that appear only after treatment.
Data: Pang, Prieto et al., single-cell exome and whole-genome sequencing of pediatric acute lymphoblastic leukemia.
Pediatric AML target discovery
Screen 1,704 surface proteins for marrow-sparing immunotherapy targets across 96,627 cells from 28 children with AML and 2 healthy bone-marrow donors. Set your own thresholds for efficacy, patient breadth and normal-marrow sparing and watch the candidate list narrow, then profile any target across the 49 leukemic states, normal tissue, and independent adult cohorts. Any of the 21,895 genes the atlas quantifies can be mapped onto the embedding, not only the screened surface proteins.
Data: single-cell RNA-seq of pediatric AML and healthy marrow, with bulk validation in TARGET and Beat AML.
Developing cerebellum atlas
The developing mouse cerebellum at single-cell resolution: 40,253 cells spanning thirteen timepoints from embryonic day 10 to postnatal day 10, across granule, Purkinje, GABAergic, glial and progenitor populations. Map any of 15,624 genes onto the atlas, follow expression across thirteen timepoints, and rank genes by how specific they are to one population.
Data: Carter RA et al., Current Biology 28, 2910-2920 (2018). Raw reads at ENA PRJEB23051.
Where the matrices live
Each browser includes the processed tables it needs. The primary sequencing data lives in the public archives below, not on this site.
Pediatric ALL
- Processed tables behind this browser Mutation burden for every bulk sample and single cell, the RAS allele frequencies, the ex vivo drug-response matrices, the 115 single-cell genotypes, read counts and clone calls, the six phylogenies in Newick, and the gene-level recurrence and AlphaMissense scores. 90 KB.
Sequencing data are not yet deposited. Raw reads are going to NCBI SRA and dbGaP under controlled access, because they come from identifiable patient material, with processed matrices in Mendeley Data. Accessions will appear here when the paper is published.
Pediatric AML
- Single-cell count matrices 30 samples, scRNA-seq and CITE-seq, at the ALSF Single-cell Pediatric Cancer Atlas, project SCPCP000007
- Bulk RNA-seq for survival validation TARGET-AML through the Genomic Data Commons, dbGaP phs000465
- Adult validation cohort The Beat AML public release
- Processed tables behind this browser The 96,627-cell embedding, per-gene expression in each grid square, the 1,704-gene surfaceome screen, and the safety and combination tables. 6.4 MB.
- Per-bin expression for all 21,895 atlas genes The full gene × grid-bin matrix the browser maps, as 343 files of 64 genes each, about 147 KB apiece. The browser fetches one only when you pick a gene it holds. Mitochondrial, ribosomal and haemoglobin genes are not in the atlas's own gene list and so are not here.
Developing cerebellum
- Aligned BAMs at ENA Thirteen timepoints in duplicate, about 595 GB, accession PRJEB23051
- Processed tables behind this browser The per-cell embedding and annotation, plus mean expression and detection rate in every cell type and timepoint for all 21,657 genes. 2.1 MB.
No count matrix has ever been deposited for this dataset. ENA holds aligned BAMs only, with no FASTQs and no processed matrices, so raw integer counts have to be regenerated by running cellranger bamtofastq on those BAMs and then cellranger count. Everything the browser shows is derived from log-normalized expression and is labelled as such.
Built to outlive their hosting
Every app is compiled to WebAssembly, so the analysis runs inside your browser rather than on a server we have to keep alive. That means they load from plain static files, cost nothing to run, and will keep working for as long as the page exists. The first visit downloads the R runtime and takes a few seconds; after that it is cached.
Every number shown comes from the published source data, which is bundled with each app. Where a measurement was only made for a subset of genes, the app says so rather than leaving a blank panel.