Your data is talking. Are you listening?

Curated, license-clean antibody, biologics and RNA data, with the provenance to defend it.

LatentProt consolidates public and licensed sources into one catalog. Every record is tiered by the strength and independence of its evidence and screened for retractions and anomalous data before it reaches a model.

Request access View the source inventory

What LatentProt indexes, as of 29 October 2026

Aggregate figures from published source metadata. No sequences or data files are published on this page.

commercially licensed sources, with instant access

22,155 TB

of open data indexed across 176 sources¹

40,510

cross-source target matches, evidence-tiered

By domain

7.0B

DNA sequences indexed across 5 sources, incl. ENA, GenBank, SRA, DDBJ and Ensembl³

4.0B

antibody sequences and records indexed across 52 sources²

1.8B

protein and nucleotide sequences indexed across 29 sources²

1.1B

protein structures indexed across 13 sources²

126M

RNA sequences and records indexed across 10 sources²

62.8M

enzyme and kinetics records indexed across 25 sources²

1.5M

mass-spec peptides indexed across 6 immunopeptidomics and proteomics sources²

³ ENA, NCBI GenBank/SRA and DDBJ form the INSDC and mirror one another daily; they are counted once, using the largest single published figure (GenBank: 7.0B sequences; SRA: 21,832 TB open access).
¹ Sum of the lower bound of each published dataset size (87 sources publish one; INSDC counted once³), rounded down.
² Counts where published, summed from catalog metadata and rounded down; sources without a published count are listed but not counted.

Evidence tiers and a bad-actor screen

Most catalogs tell you what a record says. LatentProt also tells you how far to trust it.

Tier	Meaning	Matches
T5	Strongest, independently corroborated	5,459
T4	Strong	345
T3	Moderate	2,478
T2	Single-source or weak	31,030
T1	Lowest, including flagged records	1,198

Retraction screen

Every linked publication is checked against the open Retraction Watch database, including paper-mill and fabrication notices. Flagged records drop to T1.

Measurement quality

Structures are scored on method, resolution and validation. Low-quality records are capped at T3.

Anomaly tests

Statistical checks catch fabricated or duplicated values. Origin country is never used as a signal.

Engagement options

Pricing is scoped per engagement.

Data

LatentProt

Tiered, license-filtered dataset
Per-record provenance and quality scores
Refresh cadence by agreement

Data + Audit

LatentProt + Latch

Everything in Data
Latch audit trail: source, license and version for every record
Diligence-ready lineage reports

Full stack

LatentProt + Latch + CorTex

Everything in Data + Audit
CorTex managed compute for training and inference
Models trained only on cleared data

Contact

Tell us your use case and we will follow up. Please do not send confidential information by email.

Request access Source inventory