Isomorphic Labs Unveils IsoDDE, AlphaFold 3 Successor
20 Jul 2026
Isomorphic Labs has announced the Isomorphic Labs Drug Design Engine (IsoDDE), a new computational drug-design system positioned as the successor to AlphaFold 3, the structure-prediction model released by Google DeepMind in 2024. According to the company, IsoDDE progresses beyond AlphaFold 3's predictive accuracy while adding capabilities intended to bridge protein structure prediction with real-world drug discovery workflows.
What Isomorphic Labs is claiming
Per the developer's own announcement, IsoDDE more than doubles AlphaFold 3's accuracy (2x) on the "Runs N Poses" benchmark for protein-ligand structure prediction. In what the company calls the "high-fidelity regime," IsoDDE reportedly outperforms AlphaFold 3 by 2.3x and Boltz-2 — a competing model — by 19.8x on antibody-antigen interface prediction.
Beyond raw accuracy, Isomorphic Labs says IsoDDE can identify novel, ligandable pockets even without a known ligand, a capability the company suggests could support the discovery of first-in-class drug targets. The system is also said to demonstrate improved prediction of CDR-H3 loops, a component relevant to de novo antibody design.
For context on AlphaFold 3's reach, the model has reportedly been used by more than 3 million researchers in over 190 countries since its 2024 release — a scale IsoDDE would need to match or exceed if it becomes similarly available.
What's not yet known
Several important details remain unspecified in the announcement:
- The exact release or availability date of IsoDDE has not been disclosed.
- No independent or third-party benchmark validation has been provided for the reported performance multiples — all figures originate solely from Isomorphic Labs' own announcement.
- It is unclear whether IsoDDE will be publicly accessible, offered in limited release, or kept for internal use only.
- No information has been given on cost, computational requirements, or licensing terms.
- The methodology behind the "Runs N Poses" benchmark and the "high-fidelity regime" metrics has not been detailed.
These gaps matter: the benchmarks used are internal to Isomorphic Labs and may not fully reflect real-world drug discovery outcomes, and comparisons to Boltz-2 and AlphaFold 3 may not generalize across all target classes or use cases.
Why founders should care
For founders in AI-driven biotech and drug discovery, this announcement is likely worth monitoring rather than acting on immediately. If the reported accuracy gains hold up under independent scrutiny, it may signal a meaningful shift toward AI-native tools for structure prediction — potentially creating integration or partnership opportunities for startups building on top of such models. The pocket-identification feature, if validated, could plausibly open new avenues for teams targeting biological structures without known ligands, and the antibody design capability may point to emerging opportunities in AI-driven biologics.
However, because all performance claims currently trace back to a single source with no independent verification, founders should treat these figures as preliminary. Before making product, partnership, or investment decisions based on IsoDDE's reported capabilities, it would be prudent to wait for or actively seek third-party benchmarking, clarity on availability and licensing, and more detail on the underlying evaluation methodology.
Bottom line
Isomorphic Labs is positioning IsoDDE as a significant step beyond AlphaFold 3, with bold internal benchmark numbers to back the claim. But with no release date, no independent validation, and no pricing or access details yet available, the practical implications for founders remain speculative until more information emerges.