Data

The tables, grouped by what they support.

Every table behind the preprints, under a licence that allows reuse. Attribution to the identifier below is sufficient.

Archive identifier Pending deposit

The deposit has not been completed, so no identifier is quoted here. It is marked pending rather than estimated, which is the rule applied to every unfinished value on this site.

Base editing and off-target safety

Behind the editing design and the specificity result.

TableWhat it holds
ABE_GUIDE_TABLE.csvEvery guide that places the target letter in the editing window, with editor, docking sequence, genomic span, and the count of neighbouring letters the editor could change by mistake.
ABE_OFFTARGET_SUMMARY.csvPer-guide counts across the genome by mismatch distance, given both as raw sequence matches and as sites the editor could actually edit.
ABE_RESCUE_MODEL.csvPredicted current against the fraction of cells corrected, anchored to the correction rate that removed the disease signature in mice.
ABE_ACCESSIBILITY_TOPGENES.csvTwenty cardiac genes with the openness of their coding regions, included as the control that stops the accessibility measure being over-read.

The closed upregulation route

The measurement that killed it, and the arithmetic it failed.

TableWhat it holds
SCN5A_SPLICING_SUMMARY.csvThe non-productive fraction of message in heart against the brain positive control, by two independent methods, with sample counts and spread.
SCN5A_TRANSCRIPTS.csvAll 21 annotated forms of the gene with length, type, and whether each can produce working protein.
UPREGULATION_HEADROOM.csvPredicted recovery across boost factors under both candidate mechanisms, which is the table the route needed to clear and did not.

The database census

Behind the result that generalises beyond this gene.

TableWhat it holds
CLINVAR_FUNCTIONAL_CENSUS.csvEvery variant carrying functional evidence deposited by a laboratory other than the one bulk depositor, one row each, with classification, submitter counts and the evidence codes. This is the corrected build.
CLINVAR_ACCESS_ROUTES.csvSample variants across five genes and six depositors, with response sizes and which fields each retrieval route returns.
NTD_FUNCTIONAL_EVIDENCE_TABLE.csvVariants in this domain joined across functional scores, classification, population frequency, conservation and burial. This is the table that surfaced two variants with strong evidence and no reported carrier.

Route comparison

The ranking, with its precedent citations.

TableWhat it holds
MODALITY_COMPARISON.csvAll nine routes with what each acts on, predicted current, best precedent, what gates it, and whether it depends on the unresolved mechanism.
PE_VS_ABE_EFFICIENCY.csvPublished cardiac efficiency for prime editing against base editing, which is the comparison that keeps prime editing second despite its better design.

One table is withheld, and why

The per-gene breakdown of the census is not offered. The build that exists predates the correction described on the census page, and a corrected per-gene table has not been produced. Publishing the old one would put a file on this site that the project has already shown to be wrong, so it waits.

What is deliberately not here

Two tables that exist in the working set are not published. One records which findings would be new to a specific research group, and one weighs my own likely authorship position on work that has not happened. Both were written to plan correspondence, not to report a result. Publishing them would put private judgements about named people into the public record, and they support no claim on this site.

Reference sources

  • GRCh38 reference genome, no-alt analysis set, with checksum recorded in the deposit
  • RefSeq annotation, release recorded in the deposit
  • ClinVar full release, with the published checksum verified
  • GTEx expression and splice-junction data, release recorded per table
  • ENCODE heart and cardiomyocyte accessibility experiments, accessions listed per row

If you reuse the off-target scan

One thing that will otherwise cost you a day: the reference genome carries the healthy sequence, so a guide designed against the variant matches its own target site at one mismatch rather than zero. A pipeline reporting zero mismatches against the unmodified reference has a coordinate error, and that check is worth running before any sequencing money is spent.

Licence

CC BY 4.0. Attribution to Ethan Bradley, ORCID 0009-0008-8925-7975, is sufficient.