The first comprehensive benchmark for de novo antibody design that measures what actually matters: binding, structure, developability, and immunogenicity.
25 pages • 5 platforms • Full methodology
Current benchmarks optimize for binding alone, creating a dangerous blind spot. A 1 pM binder that aggregates at 37°C or triggers an immune response is clinically worthless.
Existing benchmarks measure affinity but ignore the 85% of candidates that fail downstream gates.
Aggregation, poor solubility, and low thermostability kill programs after millions in investment.
25% of late-stage clinical failures are due to anti-drug antibodies—yet no benchmark measures this.
4 of 5 major platforms are closed-source, preventing fair comparison and reproducibility.
DADB-v1.0 introduces a composite scoring system that treats antibody design as a multi-event competition—because real drug discovery is never single-metric.
DADB Score = (0.40 × Binding) + (0.25 × Structure) + (0.20 × Developability) + (0.15 × Immunogenicity)
Affinity, kinetics, specificity
RMSD, CDR-H3 accuracy
GATEKEEPER
Tm, aggregation, solubility
T-cell assays, cytokine release
Developability acts as a binary gate: fail any threshold (Tm < 60°C, aggregation > 5%, solubility < 10 mg/mL, hydrophobicity flag) and receive zero for that component—regardless of picomolar affinity. No more pyrrhic victories.
DADB-v1.0 enables fair comparison across both open-source and proprietary platforms through standardized APIs and containerized evaluation.
Strength: Highest consistent hit rates (39% VHH-Fc)
Strength: Most extensive cryo-EM validation
Strength: Zero-prior epitope targeting
Strength: Only fully open-source platform
Strength: Highest monomeric affinity + immunogenicity data
The Immunogenicity Gap: Despite causing 25% of late-stage failures, Latent-X2 is the only platform with published human immunogenicity data. DADB-v1.0 makes this a scored component for the first time.
From public validation to clinical correlation—a phased approach to benchmark maturity.
5-platform comparison, immunogenicity track
Bispecifics, Fc engineering
ADC track, linker-payload optimization
Cell therapy, TCR design
In vivo prediction, clinical correlation
DADB's success depends on community participation. Whether you're an academic lab, AI platform, or biopharma partner, there's a role for you.
Private test set targets are confidential and shared only under NDA
Download the complete 25-page scientific paper with detailed methodology, platform comparisons, scoring formulas, and benchmark specifications.
Version 6.0 • Security-reviewed and scientifically corrected • January 2026