NOVA

Bittensor Subnet 68

NOVA merges decentralized computing, ML-driven active learning, and domain-specific heuristics to tackle drug discovery at scale.

Our approach transforms the quest for new therapeutics into a hyper-competitive race. NOVA functions as a back to back hackathon, running 24/7. In this adversarial environment, NOVA allows us to rapidly identify and improve upon areas of low confidence in SOTA models unlike any private company or academic institution. This data can be used to sharpen our predictions, build more reliable chemical activity maps, and fine-tune robust models to support our own r&d and that of strategic partners.

pharma crisis

on average it takes

$1 B+

USD

10+

YEARS

to develop 1 drug with

10%

CHANCE

90% reused

CHEMICAL SCAFFOLD

ai can make drug discovery

better

faster

cheaper

novel

but decentralized merit-based competition is necessary to deliver the promise of ai

Centralized AI drug‑discovery platforms reward secrecy over scientific progress—models and data are locked behind NDAs to protect valuations, teams duplicate each other’s efforts, and exploration is confined to well‑trodden chemical scaffolds. This siloed approach breeds redundancy, stifles collaboration, and makes it very difficult to scale to truly novel regions of the chemical universe. Without open data‑sharing or adversarial stress‑testing, AI models overfit to narrow datasets and falter when faced with unprecedented targets.

NOVA flips the script. By harnessing a decentralized, tokenized incentive system, we align rewards with transparency and merit. Miners compete to submit diverse, high‑affinity molecules and then stress‑test open‑source oracles across billions of compounds. Dynamic target–antitarget challenges and a Shannon‑diversity bonus ensure broad chemical mapping and early toxicity filtering. In doing so, NOVA exposes high‑variance blind spots, iteratively refines predictive models and accelerates the discovery of breakthrough small molecules in regions no closed lab can reach.

what makes NOVA different

massive libraries

1 |

Miners tap into an ultra-large virtual database of 1+ billion synthesizable compounds.

This ensures that every “hit” is ready for further development.

low barrier of entry

2 |

Our subnet design encourages high miner registration given low hardware requirements and that there is no need for domain-specific expertise to participate.

Challenge is reduced to a search problem that rewards heuristic innovation.

heuristic innovation

3 |

Miners iteratively refine their search strategies, from brute force to advanced ML-based active learning. The duration of the challenges encourage miners to develop innovative heuristics to select top compounds at top speed.

high value outputs

4 |

All outputs are monetized. We are tokenizing new chemical IP, target chemical libraries, and wet lab datasets.

We combine the recurring revenue of a subnet with the possibility of outsize payments from licensing .

hyper competitive

5 |

“Winner-takes-all” challenges incentivize continuous improvement and reveal areas of low confidence in SOTA models at lightning speed

multiple parameters

6 |

As more miners join, the collective knowledge and power expand. We will move beyond activity predictions to include other key parameters for drug development.

Our plan is to integrate ADME/TOX metrics and wet lab validations to create first-in-class libraries.

how does it work?

NOVA is the starting point of a flywheel of digital and real world value creation

virtual drug screening at lighting speed

NOVA pushes virtual drug screening beyond state-of-the-art models and transforms it into a high‑speed race for breakthroughs. We are mapping the chemical universe one region at a time. Interested in custom chemical maps, novel libraries of top compounds, or specialized fine-tuned models?

calling all potential partners

FAQs

  • NOVA is a decentralized drug discovery subnet on the bittensor network, designed to rapidly screen billions of synthesizable compounds for potential therapeutic applications. Traditional pipelines often take over a decade and cost billions per drug—our approach leverages global compute power, cutting-edge ai, and a dynamic incentive model to reduce both time and expense, accelerating the path to new treatments.

  • Most AI platforms for drug discovery are centralized, limited by their own compute capacity and proprietary data. NOVA, on the other hand, distributes the workload across a global network of participants—miners, validators, and data scientists—who compete in real time. This collective intelligence, combined with on-chain transparency, transforms drug discovery into an open, merit-based process rather than a black-box service.

  • Each “challenge” is tied to a particular protein target or family. Miners submit candidate molecules from massive databases of synthesizable molecules, with only one active submission allowed at a time. Validator score these submission with our deterministic oracle. The highest-scoring miner receives alpha tao rewards. This winner-takes-all approach rewards genuine innovation and continually improves the quality of submissions.

  • Yes. NOVA focuses on the savi 2020 database, which exclusively contains synthesizable compounds—meaning each predicted molecule comes with a viable route to lab production. This direct link between virtual hits and real-world validation greatly accelerates the journey from in silico discovery to experimental testing, enhancing both scientific impact and commercial potential.