Live · 14,328 simulations underway
[00]Decentralized oncology researchEST. 2024

An autonomous pipeline of agents that design, refine and simulate protein candidates against tumor antigens — funded transparently on-chain and validated by a global mesh of researchers.
See it runbacked by 1,400+ researchers
400300200100000
0.00
EQ · 0°+ N– S
A1ANTIGENP01116
MUTG12D
S3SIMULATOR
S2CRITIC
S1DESIGNER
N7NODE7A · 99.94%
τRUNTIME38:14
helix · design
candidate · HX-9221.b3refining
FRAME0832 / 4096
204 ms / step
Target
UniProt P01116
Mut G12D
Constraints
druggable
no off-target HK
MW < 18 kDa
Suggestions
swap.residue(F47, Y)
rotate.helix(α2, +12°)
add.disulfide(C18,C32)
console.helix.bio
uptime 99.94%
powered by
AlphaFold-3ESM-2RFdiffusionOpenFoldBoltz-1GROMACSRosettaSolana RPCFilecoinArweaveVitaDAOBIO Protocol

Cancerresearchisbroken.Tofixit,werebuiltthesubstratefromfirstprinciples.

[01]

Twenty million new diagnoses a year. The science needed to help them already exists — locked behind paper walls and fragmented compute.

[01]ONE STACK

Composable, not closed

Every agent is open weight, replaceable, and verifiable. Fork the Critic, retrain the Designer, ship a custom Sentinel — all on the same substrate.

Read the spec
[02]OPEN COMPUTE

Decentralized, not gate-kept

A global mesh of researcher-run nodes settles every simulation. Reputation tokens flow back to whoever moved the field — not the journal.

Read the spec
[03]REAL OUTCOMES

Patient-first, not paper-first

Clinical co-design from day zero. Every candidate ships with provenance, drug-likeness, and a path to wet-lab validation.

Read the spec
IDPBATA · LAB 03 · 2026
liveSF · 2026
BATA
case study · 2026
view ↗

We replaced six months of wet-lab triage with a 38-minute Helix run. The candidates that came out the other side were already de-risked.

Dr. Anya Voloshina· PI · Bay Area Tumor Atlas
[01]
20+
hours of triage replaced
per candidate, vs. legacy pipeline
[02]
62% → 4%
false-positive rate
after Critic stage in production
[03]
Open
weights · data · provenance
every artifact, every run, every node
[02] · who is on Helix


Researchers across 26 institutions and 9 DAOs run candidates through Helix every day — not because they have to, but because the alternative is a notebook and a wait list.

BATA
Bay Area Tumor Atlas
VITA·DAO
Longevity collective
ICR
Institute of Cancer Research
PARK·LAB
Parkinson Bio
MARIGOLD
Pediatric oncology
DESCI·EU
European DeSci alliance
HELIX·NET
research mesh
ANVIL
Open compute fund
26 institutions · 9 DAOs · 14k researcher-nodes onlineSee network map ↗
[03] · the pipeline

Three agents, one closed loop. From a target sequence to a vetted candidate in under 38 minutes — every step verifiable on-chain.

01· Target ingest

Physical proteomics

Tumor antigens, mutational landscapes, and patient-specific surfacing — pulled from PDB, AlphaFoldDB, and clinical caches as a typed graph.

read the agent spec
01Target ingest
stage_01.execlive
> query.target('KRAS-G12D')
> load.context(N=128k)
> graph.attach(antigens=4,212)
> ready · 14ms
02· Business data

Generative design

RFdiffusion + ESM-2 produce 10⁴ candidate scaffolds per hour. Each carries druggability, half-life, and synthesis cost annotations from birth.

read the agent spec
02Business data
stage_02.execlive
scaffolds
0
unique
0
passing
0
ATTENTION HEAT
03· Finance & ops

Simulation & dock

100 ns molecular dynamics, full free-energy landscape, off-target screen, and on-chain provenance — all before a single drop hits a flask.

read the agent spec
03Finance & ops
stage_03.execlive
FREE ENERGY · kcal/molΔG = -9.2
RMSD
1.4 Å
POSE
stable
OFF·T
02 / 9821

Aggregate, design, and simulate in one platform.

CUSTOMIZE EVERY LAYER
Legacy research stacks force-fit your hypothesis into rigid templates. Helix was built from the ground up for autonomous agents to compose, fork, and recombine — at the speed of a research question.
[04] · the products
01 · Generative protein binders

Manage candidate generation

Designer is the adaptive design core that combines diffusion-based scaffolding with a no-code agent interface — let ESM-2, RFdiffusion, and ProteinMPNN compose against any tumor antigen.

Explore Designer
A1ANTIGEN
S1SCAFFOLD
M1MUTATIONS
C1CONSTRAINTS
L1LIBRARY
P1SYNTH PATH
DSGN · running10,240 scaffolds / hr
[05] · take the loop


Fast to deploy. Easy to fork. Built to scale to a million simulations a day. Helix is open today — your candidate could be on-chain by tonight.

Read whitepaper
14k researcher-nodes online · 99.94% SLA
[06] · indications


Composable by design, Helix adapts to the unique antigen landscape of each cancer family. One platform, every modality, every patient cohort.

HELIXA1DESIGNA2REFINEA3VERIFY[IN] ANTIGEN[OUT] CANDINDICATION · MAPlung · breast · GI24 active runs · live000050100
01 · lung · breast · GI
Solid tumors
Active targets
KRAS-G12DEGFR-T790MHER2PD-L1VEGFR2
sims this month
8,214
Explore


Fast to deploy. Easy to fork. Built to scale.