Nabla Bio
George Church-lab spin-out whose JAM platform designs antibodies computationally — backed by $550M+ and $1B pharma partnerships.
1. Core Product / Service
Nabla Bio (Cambridge, MA) builds JAM ("Joint Atomic Model"), a multimodal generative AI platform that designs antibodies from scratch. JAM models the target's amino-acid sequence and/or structure and "autocompletes" the antibody that will bind it; the newer JAM-2 jointly models the antibody sequence, the target's atomic structure, and their interaction [1]. A defining capability is "introspection" — test-time scaling via best-of-N rejection sampling, where the model generates candidate batches, keeps the best, and seeds the next round, so compute buys quality [1].
Nabla claims first-in-class computationally designed binders to hard targets including the GPCR CXCR7 (ACKR3) and Claudin-4, plus the first computationally designed antibody GPCR agonists [1].
2. Target Users & Pain Points
- Pharma partners (AstraZeneca, Bristol Myers Squibb, Takeda) who want de novo antibodies and multispecifics against difficult targets.
- Pain solved: hard-to-drug targets (GPCRs, ion channels) resist traditional discovery; Nabla's design-first approach opens them to biologic intervention, and its integrated wet lab (reported ~100,000 antibodies characterized in ~3 weeks) closes the design→validation loop in 3–6 weeks [1].
3. Competitive Landscape
| Player | Approach | Vs. Nabla |
|---|---|---|
| absci | Zero-shot antibody design + wet lab | Absci is public with its own clinical pipeline; Nabla is partnership-funded |
| chai-discovery | Chai-2 de novo antibody design | Chai is open-source-first; Nabla is a partnered biotech |
| isomorphic-labs | AlphaFold lineage + wet lab | Big-tech scale vs. a focused antibody-design startup |
| insilico-medicine | Pharma.AI generative suite | Broader small-molecule focus; Nabla is antibody-specialist |
4. Unique Observations
- Partnership-funded, capital-light: with only ~$37M raised but >$550M and then >$1B in pharma deal value, Nabla is the sector's clearest example of funding a platform through milestone-heavy biobucks rather than VC equity — low dilution, high dependency on partner programs advancing [1].
- "Design, not discovery" thesis: Nabla's founding bet — that next-gen antibodies will be designed by AI rather than discovered by screening natural repertoires — is a direct articulation of the shift this week's ai-for-science research tracks [1].
- JAM-2 as a named model in local research: this week's notes specifically flagged JAM-2 in the antibody-design company survey, alongside Absci and Chai, suggesting it is now a canonical reference point for AI antibody design [local].
5. Financials / Funding
- Seed: $11M (Dec 2021; co-led Khosla Ventures + Zetta Venture Partners) [1].
- Series A: $26M (May 2024; led Radical Ventures) — total VC ~$37M [1].
- Partnerships: AstraZeneca + BMS + Takeda (May 2024, >$550M aggregate upfront+milestones); Takeda expanded collaboration (Oct 2025, >$1B potential milestones) [1].
6. People & Relationships
- Co-founders: Surge Biswas (CEO), Frances Anastassacos; academic co-founder George Church (Harvard/Wyss).
- Partners: AstraZeneca, Bristol Myers Squibb, Takeda.
- Peers: absci, chai-discovery, isomorphic-labs, insilico-medicine.
Sources
- [1] BioPharmaTrend, "Takeda Expands AI-Driven Antibody Design Partnership with Nabla Bio in a Billion-Dollar Deal" (2026-08-24); AI Wiki — Nabla Bio profile (2026-08-24)