Wave-resonance virtual screening,
at production speed.

MolDyn — Wave-Resonance Virtual Screening

Screen 2.6M molecules in 16.77 ms. AUROC 0.782 on HIV-1 protease (DUD-E benchmark).

Virtual screening bottlenecks drug discovery.

Hit discovery campaigns stall at the library-screening step. Modern compound libraries have outgrown the tools built to search them, and every month lost in screening is a month off your program timeline.

1

Slow

Legacy screening infrastructure takes hours or days per target. Teams triage around the tool, not the biology.

2

Expensive

Commercial docking suites price per seat and per CPU-hour. A single virtual screen can cost tens of thousands of dollars in compute alone.

3

Linear scaling

Every new target, every library refresh, every mode explored — screen time grows with library size. There is no economy of scale.

MolDyn by CellsWave.

A production-grade wave-resonance screening platform built for multi-million compound libraries, deployed wherever your data lives.

16.77 ms
Per query

Sub-20-millisecond search across 2.6 M ChEMBL compounds on a single GPU. Interactive screening replaces overnight batch runs.

0.782
AUROC on HIV-1 protease

Validated against the DUD-E gold-standard benchmark. Enrichment at 1% of 12.66 × random — inside the published band of industry-standard methods.

BYOC
Bring your own cloud

Runs on your infrastructure. Your queries never leave your environment. Deploy on AWS, GCP, Azure, Hetzner, or bare-metal.

Validated against the industry standard.

Every claim on this page comes from the public DUD-E benchmark — the same dataset used by ECFP4 + Tanimoto, ROCS, AutoDock Vina, and Glide.

Target Class AUROC EF@1%
HIV-1 protease Aspartic protease 0.782 12.66×
Thrombin Serine protease 0.744 6.63×
Envelope note. MolDyn v1.0 is validated for protease targets. Kinase targets are outside the v1.0 deployment envelope — contact us to discuss roadmap and pilot scoping for other target classes.

AUROC = Area Under ROC Curve (1.0 = perfect, 0.5 = random). EF@1% = enrichment factor in top 1% of ranked library (1.0× = random). Benchmark protocol: 3-query consensus, leave-N-out. Full methodology and statistical tests available in the benchmark report.

Industry-standard dataset Validated on the public DUD-E benchmark — the same dataset used to publish results for ECFP4, ROCS, Vina, and Glide.
Published methodology Mysinger et al., J. Med. Chem. 2012. No private test sets, no cherry-picked targets.
Independent validation Full benchmark report and per-target breakdown available upon NDA. See the public report →

Three ways to run MolDyn.

Every deployment option delivers the same API surface and the same performance envelope. Pick the one that fits your security and compliance posture.

Cloud SaaS

Hosted by CellsWave

Get up and running in a day. HTTPS API with API-key authentication, SLA-backed uptime, usage-metered billing.

  • No infrastructure to manage
  • Automatic updates
  • Best for exploration and small teams

On-premise

Your hardware, your network

Airgapped installation for regulated or secrecy-sensitive environments. Runs on a single GPU server; no outbound connections required.

  • Zero network egress, fully airgapped
  • Offline license mode supported
  • Best for pharma and biotech compliance requirements

Priced to pay back within a single screen.

Start with a scoped pilot or move straight to a platform license. Enterprise engagements are scoped individually.

Pilot
$30,000
per target
  • 1–2 protease targets
  • 4-week engagement
  • Ranked hit list
  • Reproducibility report
Enterprise
Custom
talk to us
  • On-premise deployment
  • Private library indexing
  • Custom SLA
  • Milestone + royalty available
Contact Us

Frequently asked questions.

What GPU do I need?
NVIDIA GPU with 8 GB+ VRAM — RTX 3080, A100, H100, or equivalent. CUDA 12+ driver. MolDyn is GPU-only; a CPU fallback is not shipped in v1.0.
What molecule formats do you support?
SMILES strings and SDF files for ligands; standard PDB format for protein structures. PDB IDs are auto-downloaded from the RCSB if not in your local cache.
Which targets are supported in v1.0?
Protease targets: HIV-1 PR, HCV NS3, SARS-CoV-2 Mpro, thrombin, factor Xa, and antibacterial proteases. Kinase support is on the v1.1 roadmap.
How is my data protected?
In BYOC mode, all data stays on your infrastructure. We never see your compounds, your pockets, or your results. License activation is the only outbound call, and an airgapped offline mode is available for on-premise deployments.
What is the difference between trial and paid?
Trial gives 1,000 API calls for 30 days so you can reproduce the benchmark and screen your real target against the full library. Paid plans remove the call limit and add priority support.

Get in touch

Pilot scoping, technical review, enterprise deployment, or anything else — the fastest route is email. We reply within one business day.

Based in the EU · Available worldwide