01Overview
Ten63 Therapeutics is building what may be the most ambitious computational drug discovery platform in existence — a Large Quantum Chemistry Model (LQCM) that fuses 20+ years of provable protein design algorithms with quantum-mechanics-level molecular simulations, running at speeds that make conventional computational chemistry look glacial.
Unlike pattern-matching AI approaches that learn from existing data, BEYOND simulates molecular interactions from first principles — quantum mechanics. This physics-grounded approach means it can work on data-poor targets where no existing drugs or crystal structures exist.
Built on 20+ years of the Donald Lab's OSPREY algorithm at Duke University — provably optimal protein design with continuous backbone flexibility, ensemble-based scoring, and ML-accelerated quantum chemistry energy functions. BEYOND adds ~1M lines of code to extend into all drug-like chemical space.
~80% of human proteins lack well-defined binding pockets for conventional drugs. BEYOND discovers cryptic (hidden) binding sites on transcription factors, flat protein-protein interfaces, and disordered regions that have defeated traditional screening for decades.
BEYOND = Binding Evolution Yields Original New Drugs. The platform generates and evaluates molecules through iterative cycles analogous to evolutionary selection — optimizing binding affinity, selectivity, and drug-like properties simultaneously.
"AI not only accelerates drug discovery; it enables us to pursue high-impact targets in ways that were previously impossible. By perfecting a simulation environment that enables AI to explore and learn from trillions of molecular possibilities, millions of times faster than physical methods, while retaining experimental-level accuracy, we are advancing towards superintelligence for drug discovery." — Marcel Frenkel, PhD, CEO of Ten63 Therapeutics (Feb 2026)
Founding Story
Ten63 was born from grief and determination. Co-founder Marcel Frenkel's mother was diagnosed with pancreatic cancer in 2014 — driven by mutations in proteins that, despite being widely studied, remained undruggable. At Duke's Donald Lab, Frenkel joined forces with Mark Hallen (CTO) and Bruce Donald (Chair), whose OSPREY algorithm had already pioneered provable protein design algorithms. After Frenkel's mother passed in 2015, the three channeled their research into a mission: make nothing undruggable.
The company — originally Gavilán Biodesign — launched in December 2018 as a Duke/TTIC spinout, then joined SOSV's IndieBio accelerator. Ten63 raised a $15.9M oversubscribed Series A in 2023 and closed strategic financing bringing total funding to $45M+ in February 2026, with new backing from Chugai Venture Fund and the Gates Foundation.
Funding Timeline
02BEYOND Platform Architecture
BEYOND is described as the world's first Large Quantum Chemistry Model (LQCM). It unifies OSPREY's provable combinatorial optimization with ML-accelerated quantum chemistry to explore chemical space at a scale and resolution impossible for physical screening.
Technical Foundation: OSPREY → BEYOND
Provable algorithms — Unlike heuristic docking tools, OSPREY provides mathematically provable bounds on the quality of solutions. The K* algorithm computes partition functions for binding affinity with guarantees.
Continuous flexibility — Models backbone, sidechain, and ligand flexibility simultaneously using EPIC (Energy as a Polynomial In Continuous rotamers) and LUTE (Local Unpruned Tuple Expansion), capturing induced-fit effects missed by rigid docking.
All drug-like chemical space — OSPREY was limited to proteins; BEYOND extends the provable algorithms to small molecules, enabling exploration of 100+ trillion drug-like compounds per target.
ML-accelerated QM — Machine learning surrogates trained on quantum chemistry calculations provide near-QM accuracy at a fraction of the compute cost, enabling the 19.3M compounds/sec throughput.
Generative chemistry — Iterative molecular evolution cycles generate novel molecules optimized for binding, not just screening existing libraries.
Speed Comparison
03Pipeline
Ten63 is advancing a preclinical pipeline of first- and best-in-class small molecules against high-impact oncology targets — all historically considered undruggable or underdrugged. The Gates Foundation grant adds an infectious disease program targeting HPV.
Myc drives ~70% of all cancers. Despite 40+ years of effort and hundreds of attempts, no approved drug directly inhibits Myc. Ten63 claims its BEYOND-designed inhibitors surpass all previous attempts at Myc inhibition — the platform discovers cryptic pockets on this intrinsically disordered protein that conventional methods cannot find.
While recent KRas inhibitors (sotorasib, adagrasib) have reached the clinic, resistance mutations rapidly emerge. Ten63's strategy: use OSPREY's resistance-prediction algorithms to design inhibitors that anticipate and block escape routes before they arise in patients.
DDR1 and DDR2 remodel the tumor microenvironment to exclude immune cells — reorganizing collagen networks and creating physical/biochemical barriers that shield tumors from immune attack, drive resistance, and promote metastasis. Ten63 is designing degraders (not inhibitors) that eliminate DDR1/DDR2 entirely. Target cancers: pancreatic, ovarian, lung.
HPV is the most common STI globally, causing 650,000+ cancers/year. Gates Foundation grant funds development of affordable small molecules targeting viral proteins that lead to cervical cancer — viral targets that have so far been completely undruggable. Aim: cost-effective treatments accessible worldwide.
Pipeline vs. Conventional Drug Discovery
The Undruggable Oncogene: Myc
The Myc family (c-Myc, N-Myc, L-Myc) are transcription factors that regulate cell growth, division, and death. When mutated or overexpressed, they become powerful oncogenic drivers. But Myc proteins are:
- Intrinsically disordered — no stable 3D structure for traditional drug binding
- Nuclear localized — hard for large molecules to reach
- Lacking classic binding pockets — flat, featureless protein-protein interaction surfaces
- Essential for normal cells too — toxicity is a constant concern
Ten63's BEYOND platform addresses this by: (1) simulating the conformational ensemble of Myc to discover transient cryptic pockets; (2) designing molecules that trap specific conformational states; (3) predicting selectivity to minimize off-target effects. In head-to-head comparisons, Ten63 claims BEYOND outperforms 40 years of academic and industry best-efforts against Myc.
04The Undruggable Frontier
Approximately 80% of the human proteome is considered "undruggable" — these proteins lack the well-defined binding pockets that conventional small-molecule drug discovery depends on. This is the frontier Ten63's LQCM was built to conquer.
Why Are Proteins "Undruggable"?
Many disease-driving proteins — especially transcription factors (Myc, p53, β-catenin) — have flat, featureless surfaces with no deep hydrophobic pocket for a small molecule to dock into. Traditional HTS screens fail because there's nowhere for compounds to bind with high affinity.
Some critical proteins don't adopt a single stable 3D structure — they exist as dynamic ensembles of conformations. Structure-based drug design requires a fixed target, so these proteins slip through the cracks of conventional approaches.
Many disease mechanisms involve protein-protein interactions (PPIs) mediated by large, flat, shallow contact surfaces — the opposite of the deep, well-defined pockets that drugs bind. PPI modulators remain one of the hardest challenges in medicinal chemistry.
AI approaches trained on existing drug-target data struggle with undruggable targets precisely because there are no existing drugs or screening hits to learn from. This is where physics-based methods like BEYOND have an edge — they don't need historical binding data.
BEYOND's Approach: Cryptic Pocket Discovery
BEYOND addresses the undruggable problem through several converging strategies:
- Conformational ensemble sampling — Instead of targeting a single static structure, BEYOND simulates the full conformational landscape of a protein, identifying transient pockets that only appear in certain states.
- Cryptic site detection — Computational sampling reveals hidden ("cryptic") binding sites that open and close dynamically. These sites are invisible to X-ray crystallography but druggable if you can catch them.
- Allosteric modulation — When the active site is undruggable, BEYOND searches the entire protein surface for allosteric sites where binding can alter function remotely.
- Generative molecular design — Rather than screening existing compound libraries (which were optimized for druggable targets), BEYOND generates entirely new molecular scaffolds purpose-built for unconventional binding modes.
Historical Success Rate: Undruggable Targets
05Method Arena
How does a Large Quantum Chemistry Model compare to existing computational drug discovery approaches? This arena benchmarks BEYOND against the dominant methods across key dimensions.
| Method | Accuracy | Speed | Undruggable? | Scalability | Data Needed | Key Players |
|---|---|---|---|---|---|---|
| BEYOND LQCM Ten63 |
~Lab level | 19.3M/sec | ✅ Yes — core mission | 10¹³+ molecules | Minimal — physics-based | Ten63 |
| FEP+ Gold standard |
~1 kcal/mol | Hours/compound | ❌ Needs known pocket | ~100s compounds | Crystal structure + hits | Schrödinger |
| DFT (ab initio) First principles |
High (exact) | Days/molecule | ⚠️ Theory yes, scale no | Single molecules | None — first principles | Academic |
| ML Force Fields GEMS, ANI |
Near-DFT | Minutes/traj | ⚠️ Needs training data | Protein-scale | Large QM datasets | Microsoft, Meta |
| Virtual Screening (Docking) Legacy |
Low-moderate | Fast (~1M/hr) | ❌ Rigid receptor | 10⁶–10⁹ | Crystal structure | AutoDock, GOLD |
| AlphaFold + Diffusion Structure AI |
Structure only | Seconds/struct | ⚠️ Structure ≠ drug | Proteome-scale | Sequence only | DeepMind, Boltz |
| Generative AI (SMILES) Gen-AI |
Varies widely | Fast generation | ⚠️ Trained on druggable | Unlimited generation | Large SMILES datasets | Insilico, Recursion |
| Qubit (Polarizable QM) QM/MM |
Very high | GPU-accelerated | ⚠️ Pocket-dependent | ~1000s compounds | Crystal structure | Qubit Pharma |
Accuracy vs. Throughput
Capability Radar
What Makes LQCM Different?
The key innovation of the LQCM paradigm isn't any single technique — it's the fusion of several that are usually separate:
- Provable optimization (from OSPREY) — guarantees on solution quality, not just heuristic search
- Quantum chemistry accuracy (from ML-accelerated QM) — near-DFT energies at massive throughput
- Generative molecular design — creates entirely new molecules, not just screening existing libraries
- Conformational ensemble awareness — captures protein dynamics, not just static snapshots
- Resistance prediction (unique to OSPREY lineage) — anticipates how targets will mutate to escape drugs
Most competitors have 1–2 of these capabilities. BEYOND integrates all five in a single platform — which is why they can claim to tackle targets that no one else can.
06Investment & Investor Map
Ten63 has raised $45M+ across multiple rounds, assembling a syndicate that spans pharma CVC, foundations, deep tech VCs, and strategic investors. The investor composition signals a rare convergence of commercial, scientific, and global health interest.
Investor Syndicate
Investor Category Breakdown
Competitive Funding Context
Investment Thesis
- LQCM is a genuinely novel paradigm — physics-based AI, not pattern matching
- Addressing 80% of proteome that all other AI drug companies cannot reach
- Myc program alone has blockbuster potential if successful (70% of cancers)
- Gates Foundation grant validates global health applicability
- Chugai CVC = potential pharma partnership/acquisition pathway to Japan market
- Resistance prediction is unique competitive moat from OSPREY heritage
- Entirely preclinical — no clinical validation of platform yet
- "Undruggable" claims are common in biotech; execution risk is enormous
- $45M is modest vs. competitors (Schrödinger $1B+, Recursion $800M+)
- Black-box platform — limited peer-reviewed benchmarks for BEYOND specifically
- RTP location may limit talent recruitment vs. Boston/SF hubs
- Myc has defeated hundreds of previous attempts — what's different this time?
07References
Primary Sources
- Ten63 Therapeutics. "Ten63 Therapeutics Closes Strategic Financing, Bringing Total Funding to More Than $45M." PRNewswire, February 19, 2026. Link
- SOSV. "How Ten63 turned an academic project into an AI platform for drugging undruggable cancers." SOSV Blog, February 2025. Link
- AllSci. "Quantum-powered Ten63 Therapeutics closes latest financing round for USD 45m total." AllSci News, February 2026. Link
- BioBuzz. "RTP-Based Ten63 Therapeutics Secures Strategic Investment Totaling $45M." BioBuzz News, February 2026. Link
- Ten63 Therapeutics. "Series A: $15.9 Million Oversubscribed Financing." GlobeNewsWire, May 2, 2023. Link
- Duke Capital Partners. "How a Duke Professor and Two Duke Ph.D.s are Building an AI Drug Discovery Start-Up." June 2023. Link
Scientific Foundations
- Hallen, M. A., Martin, J. W., et al. "OSPREY 3.0: Open-source protein redesign for you, with powerful new features." J. Comput. Chem. 39(30), 2494–2507, 2018. PubMed
- Donald, B. R. "OSPREY: Protein Design with Ensembles, Flexibility, and Provable Algorithms." Methods in Enzymology, 523, 87–107, 2013. PubMed
- Kaserer, T. & Blagg, J. "OSPREY Predicts Resistance Mutations Using Positive and Negative Computational Protein Design." Methods in Molecular Biology, 1529, 291–314, 2017. PubMed
- Unke, O. T. et al. "Biomolecular dynamics with machine-learned quantum-mechanical force fields trained on diverse chemical fragments." Science Advances, 10(14), eadn4397, 2024. Link
- Danel, T. et al. "∇²DFT: A Universal Quantum Chemistry Dataset of Drug-Like Molecules and a Benchmark for Neural Network Potentials." arXiv, 2024. Link
- Merz, K. M. Jr. et al. "Quantum Machine Learning in Drug Discovery: Applications in Academia and Pharmaceutical Industries." Chemical Reviews, 2024. Link
Context & Analysis
- TipRanks. "Ten63 Therapeutics Secures New Capital and Gates Grant to Scale AI-Driven Drug Discovery and HPV Program." February 2026. Link
- NC Biotechnology Center. "Drug discovery startup Ten63 Therapeutics secures new financing." February 2026. Link
- Qubit Pharmaceuticals. "AI and Quantum Chemistry: Powering the Next Leap in Drug Design." Blog, May 2025. Link