TRION Cloud

The no-install drug discovery AI workflow

Run AI-driven drug discovery workflows in the cloud, as much as you need — with no install and no infrastructure to set up.

Start with no install

Run straight from your browser — no software to install, no servers to set up.

Compute on demand

Run compute at the scale your research needs and pay only for what you use.

Expanding workflows

Target-specific activity prediction (TSAP), screening and cofolding today, with more workflows on the way.

Connected to analysis

Integrated all the way through selecting, comparing and reviewing your results.

Why TRION Cloud

Faster compute, seamless analysis, lower cost

Run AI-based drug discovery computations without standing up complex infrastructure, and carry every result through analysis and structure review in a single research flow.

Start with zero setup

No infrastructure to build and nothing to install — start computing right in your browser.

~1/10

compute cost vs. traditional HPC

~21 min

to dock 10,000 compounds

One flow into analysis

Filtering, clustering and structure review all continue without a break.

Built on proven AI models

Reproducible results on published, peer-reviewed docking and folding engines.

11 target models

Pretrained activity models for CBLB, EGFR, MAP4K1 and more, each with its validation results.

From first screen to lead series

Start on a new target or refine an existing series with the same set of tools.

Hit discovery

Virtual-screen a large library against a new target to rank your first actives.

Binding-mode check

Cofold a hit to see the predicted complex and confirm the binding mode in 3D.

Potency ranking

Rank a series by predicted activity with a target-specific model in TSAP (Boltz-2) or TSAP (AI-QSAR).

Workflow

From compute to analysis to structure review, in one flow

TRION Cloud connects AI-driven compute, result analysis and structure review in one cloud environment. Each workflow comes with a full guide on trion.cloud.

Core technology

TSAP (AI-QSAR)

Predict potency and see which interactions drive it

Each compound is docked, its protein–ligand interactions become features, and a target-specific QSAR model turns them into predicted activity.

  • QSAR on docking interaction features
  • Pretrained target models, ready to run
  • Train on your own activity data
  • Predicted activity for every compound
  • Feature contributions behind each prediction
Read the TSAP (AI-QSAR) guide →
TSAP (AI-QSAR) feature contributions

TSAP (Boltz-2)

Predict activity with a model tuned for your target

TSAP fine-tunes Boltz-2 on one target's activity data, so its predictions follow that target. Start from a pretrained target model or train one on your own measurements.

  • Boltz-2 fine-tuned per target
  • Pretrained target models, ready to run
  • Train on your own activity data
  • Predicted activity for every compound
  • Accuracy checked on held-out compounds
Read the TSAP (Boltz-2) guide →
TSAP (Boltz-2) prediction summary

Virtual Screening

Explore large candidate sets faster and more systematically

GNINA-based AI docking predicts the binding potential of thousands to tens of thousands of compounds, then selects promising candidates through docking-score and property-based filtering and structure-similarity clustering.

  • GNINA-based AI docking
  • Automated protein and ligand preparation
  • Docking-score & property-based filtering
  • Structure-similarity clustering
  • Result table and 3D pose review
  • Export to CSV, SDF, PDB
Read the virtual screening guide →
Virtual screening results dashboard
AI Cofolding result

AI Cofolding

Predict the binding structures of diverse biomolecules with AI

Boltz-2-based AI Co-folding predicts the binding structures of diverse biomolecular complexes — proteins, ligands, DNA/RNA and more. Check a molecule's binding structure on its own, or review Virtual Screening candidates at the structural level.

  • Boltz-2-based binding structure prediction
  • Protein–ligand complex prediction
  • Protein–protein interaction prediction
  • Protein–DNA/RNA structure prediction
  • 3D review of predicted structures
  • Save result structure files
Read the AI cofolding guide →

Less time on the tool, more on the science

PDB ID1HSGFetch
3D protein structure preview

Preview instantly

Type a PDB ID and the structure opens in the viewer. Check your target before you run.

✓

Target Protein

1HSG · 2 chains · Auto Box

2

Ligand Library

Diversity Library 10K · 10,238

3

Project Settings

gnina · exhaustiveness 4

Upload to export in one place

Every step, from upload to export, stays in one place.

194 / 9,812 resultsCNN 0.52–1 ✕
2DLigand ⇅TNCO ⇅ΔG ⇅
⬡Z2164932780.996-1.93
⬡Z7692956870.986-1.69
Clustering10 clusters
★Z220079260112
★Z285193238928

Sort and cluster results

Sort by score and properties, group similar compounds, and pick promising hits faster.

How it works

From AI compute to result review, in one flow

Once your input data is ready, TRION Cloud takes you from AI-based computation through candidate filtering, structure review and file export — all in one place.

1

Prepare inputs

Upload your input data, such as target proteins or compound structures.

2

Run compute

Run AI-based computation based on the workflow you choose.

3

Filtering

Narrow down promising candidates by score and physicochemical properties.

4

Clustering

Group and compare candidates by structural similarity.

5

Structure review

Inspect result poses in 3D, and predict the bound structure with AI Co-folding when needed.

6

Export

Save results in the formats your research needs — CSV, SDF, PDB and more.

Pricing

Pay only for the compute you need, as you use it.

TRION Cloud is credit-based, so you run only the compute you need. Scale flexibly from a quick test to large-scale screening, right-sized to your research.

Virtual Screening

Screen large compound libraries against your target. You pay per ligand, with no setup fee.

GNINA-based AI docking
Results dashboard
Filtering
Clustering
Export to CSV, SDF, PDB

$0.01

per ligand screened

1 credit each, 50 credits minimum per job

AI Cofolding

Predict the bound protein–ligand complex and view it in 3D. You pay for the GPU time the job needs.

Complex structure prediction
3D structure review
Batch execution
Export to PDB

From $1.38

per job, billed by GPU time

138 credits minimum

TSAP (Boltz-2)

Predict compound activity with a Boltz-2 model fine-tuned for one target. You pay for the GPU time the job needs.

Pretrained target models
Train on your own activity data
Predicted activity per compound
Holdout validation results

From $2.19

per job, billed by GPU time

219 credits minimum

TSAP (AI-QSAR)

Dock each compound and predict its activity with a target-specific QSAR model. You pay for the GPU time the job needs.

Pretrained target models
Train on your own activity data
Docking-based QSAR
Feature contributions per prediction

From $2.70

per job, billed by GPU time

270 credits minimum

Every account can use every workflow and the 3D viewer. There is no subscription; you pay with credits.

See full pricing →

Start your cloud drug discovery AI workflow with free credits.

Get 500 credits with no card required, and try TRION Cloud's drug discovery AI workflows right away.