AI for Bio · Model

Borzoi

Predicts RNA-seq coverage from DNA and scores variant effects on transcription, splicing, polyadenylation

Genomics & RNABig-tech labUnited StatesMaintained only中文

Overview

NameBorzoi · aliases: Borzoi
Full nameNo separate expanded name
Ambiguous names (not auto-linked)Flashzoi, Borzoi Prime
FamilyNot a family entry. Single-model entry, not a family
OrganizationCalico Life Sciences
Organization typeBig-tech lab
CountryUnited States
Model classSequence-to-function model
DomainGenomics & RNA
First release · preprint
Basis: bioRxiv v1 posted (Crossref issued the same day) source
Other milestones
  • Preprint v1: · bioRxiv v1 source
  • Journal online: · Nature Genetics publication source
  • Code release: · GitHub v1.0.0 source
Latest update · journal
Basis: Nature Genetics publication
Published in Nature Genetics; later commits to 2026-09-30 are CI and dependency maintenance only source
The most recent sourced event, not a claim that nothing changed since.
StatusMaintained only · Only maintenance commits after publication
OpennessCode: Open · Apache-2.0
Weights: Open
Open source; code under Apache-2.0
Benchmarks reported2 counted (self 1, retrospective 1)
Wet-lab validationNone recorded
Unverified fieldsNone
As of2026-11-16

What it does

Predicts RNA-seq coverage directly from DNA sequence and uses it to score variant effects on transcription, splicing and polyadenylation; continues the Basenji/Enformer line and is in the same class as AlphaGenome.

Benchmarks reported

Every row is one pairing from the survey's relation file; a score belongs to the benchmark page's ranking, where it is shown with metric, eval set, source type and date. 2 of 2 pairings are counted.

BenchmarkRelationEval set / protocolSource typeNote and evidence
08 Basenji/Enformer test setDeveloper report, official protocol—Self-reportedThe paper uses its own variant tasks and also reports on this test set · evidence
37 TraitGymRetrospective dataset use—Third-partyEvaluated on this set by the benchmark's authors. †partly verified · evidence

Other evaluations (outside the 44)

None recorded.

Developers report: Source states no developer claim

Independent evidence: TraitGym authors find it stronger on complex non-disease traits but behind alignment-based models on Mendelian and disease traits

Training data

Group in Figure 7: Public data only.

ClassValueDetail
a. Public databasesYesENCODE RNA-seq, GTEx v8, FANTOM5 CAGE, DNase, ChIP-seq etc.; 7,611 human output tracks
b. Private or licensed dataNo
c. Own wet-lab data used for trainingNo
d. Wet lab used only for validationNo

Wet-lab validation

None. The developers report no wet-lab validation.

Background

  • Organization: Calico is an Alphabet aging-biology company announced 2013-09; Basenji and Borzoi come from its team
  • Funding: Not disclosed or not applicable
  • Products and partnerships: The AbbVie partnership in place since 2014 reportedly ended in 2025-11 (press report)
  • Related models: AlphaGenome, NTv3

Reading notes

  1. Enformer test-set comparisons were run by the developers
  2. TraitGym results show clear differences across trait classes

Caveats

  1. bioRxiv posting day 2023-08-31 vs 09-01 is uncertain; API record gives 09-01
  2. Borzoi Prime and Flashzoi not verified and not merged here

Sources

  1. Crossref record 10.1038/s41588-024-02053-6 (journal)
  2. bioRxiv preprint (preprint)
  3. GitHub repository calico/borzoi (repository)
  4. Journal article DOI 10.1038/s41588-024-02053-6 (journal)
  5. bioRxiv preprint (preprint)

Corrections

1 correction
FieldWasNowBasisFound
first_release2023-08-312023-09-01 (bioRxiv v1 and Crossref issued date)source2026-11-16

Part of the AI for Bio benchmarks survey · all 112 models · open in the explorer. Record as of 2026-11-16.