Subnet 68: NOVA
NOVA is Bittensor Subnet 68, a drug-discovery subnet that turns early-stage compound search into a rewarded service. The project site is metanova-labs.ai, and its on-chain description is simply “Accelerating drug discovery.” The codebase for the subnet is metanova-labs/nova.
What NOVA Rewards
NOVA runs a continuous competition that asks miners to find candidate therapeutics for a rotating set of biological targets. The core idea is binding affinity: a drug can only act on a target protein if it physically interacts with it, so the subnet rewards submissions predicted to bind that target as strongly as possible.
According to the project’s miner documentation, the competition has two tracks. In the small-molecule track, miners search large chemical databases for drug-like molecules that are predicted to bind a given protein. In the biologics track, miners design nanobody-style protein sequences that are scored on how confidently they fold against the target, how strongly they interact with it, and how stable and non-toxic they are likely to be. The targets are refreshed on a weekly cycle, which keeps the search pointed at new problems rather than a fixed benchmark.
Miner and Validator Roles
Miners produce candidate molecules or sequences for the current target and submit them to the network. Submissions are encrypted at the time they are committed and only revealed for scoring after the round, so miners compete on the quality of their candidates rather than by copying each other.
Validators gather those submissions, decrypt them, and check that each one is well formed and obeys the competition’s chemical and biological rules. Valid candidates are then scored with machine-learning models that estimate how well a molecule or sequence binds the target. Validators rank the results, identify the strongest submissions for the round, and set weights on the network accordingly, which is how reward flows through Yuma Consensus. At the article level the split is straightforward: miners supply candidate compounds, while validators decide how good each one is and weight it.
Validity and Scoring Boundary
The NOVA miner documentation describes the competition as two related drug-discovery tracks: small molecules and biologics. Both tracks are tied to target proteins, but they ask miners for different kinds of candidates. A small-molecule submission is a candidate compound, while a biologics submission is a nanobody-like amino-acid sequence designed to interact with the target.
The same source separates validity from scoring. Before a candidate can win, it has to satisfy the track’s basic chemical or biological constraints. That distinction matters because a candidate can look promising in concept and still fail the subnet if it is malformed, duplicated, outside allowed composition limits, or otherwise unusable for the target round.
The NOVA validator documentation describes validators as gathering encrypted miner submissions, decrypting them, validating molecules and nanobody sequences, scoring valid submissions against target proteins, ranking results, and then setting weights. In article terms, NOVA’s reward path is a gate-and-score pipeline: invalid submissions stop at validation, while valid submissions move on to model-based ranking.
The shared pipeline also keeps the two tracks comparable without pretending they are the same scientific object. Molecules and nanobodies pass through different domain checks, but validators still treat both as target-binding candidates whose scores must be converted into subnet weights. That lets NOVA support chemically different search spaces inside one Bittensor incentive loop.
For readers, this keeps the subnet’s drug-discovery claim precise. NOVA is not rewarding arbitrary lists of molecules or protein strings. It rewards candidates that survive validity checks and then rank well against the current target. The useful output is therefore not just novelty, but target-specific candidate quality under a validator-controlled scoring process.
On-Chain Identity
Live SN68 subnet data is available on TaoStats, which identifies the subnet as NOVA. The live Finney identity for netuid 68 registers the subnet name as NOVA, with the description “Accelerating drug discovery.” The GitHub repository is metanova-labs/nova/.
Relationship to Yuma Consensus
Subnet 68 uses Yuma Consensus to convert the target-binding weight vectors that validators submit into the emission shares distributed to miners and validators within the subnet each tempo. The Yuma Consensus documentation describes how validator weight submissions are aggregated into consensus weights for each miner registered on the subnet.
In NOVA’s context, validators gather encrypted miner submissions, decrypt them after the round, validate each candidate against chemical and biological constraints, score valid submissions using machine-learning binding-affinity models, rank the results, and submit weight vectors reflecting those per-round scores. The Emission documentation describes how those consensus weights determine each participant’s share of the subnet’s accumulated emission each tempo.
Reader Boundary
Subnet 68 NOVA should not be read as generic Bittensor subnet documentation, a clinical diagnosis tool, or proof that any proposed molecule is safe or effective in humans. It describes an early-stage drug-discovery competition where miners submit candidate small molecules or nanobody-like sequences for rotating target proteins, and validators apply validity checks before scoring binding affinity with machine-learning models (NOVA miner documentation, NOVA validator documentation).
It is also not a free-form molecule list generator. Rewards are tied to target-specific candidates that survive the validator gate-and-score pipeline, with validator weights flowing through Yuma Consensus to determine emissions each tempo (Yuma Consensus, Emission).
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For Subnet 68, that sequence applies to the standard Bittensor lifecycle: localnet for isolated development, testnet for shared non-production testing, and mainnet for live operation with real emissions.
On mainnet, Subnet 68 is registered as the live production subnet at netuid 68. The Bittensor Networks reference separates mainnet, testnet, and localnet. Participation examples or emission outcomes from one environment should not be read as representing production subnet performance in another environment.
Reader Boundary
Subnet 68 NOVA should not be read as generic Bittensor subnet documentation, a pharmaceutical or medical service, or a claim that any proposed molecule is a usable drug. It names one subnet’s early-stage drug-discovery competition — miners propose molecules and protein sequences predicted to bind strongly to target proteins — on netuid 68 (NOVA repository, Understanding Subnets, Glossary: Netuid).
A miner’s reward reflects predicted binding against the target, so the article describes a computational search signal rather than experimental validation, safety, or efficacy (NOVA site — metanova-labs.ai).