SUBSEQ.BIO
DOCS-GENIE 3

Genie 3

Generate de novo protein backbones, scaffold motifs, or design binders with Genie-compatible inputs.

Overview

  • Unconditional backbone generation from an exact length.
  • Scaffolding benchmark or custom motifs.
  • Binder design against bundled benchmark targets or custom target structures.

Modes

ModeInput shapeWhen to use it
unconditional_generation
Unconditional Generation Default
No uploaded input is required by the mode itself. Generate de novo protein backbones from an exact length.
motif_scaffolding
Motif Scaffolding
Uses one selected file/source when file parameters are present. Scaffold a bundled benchmark motif or a Genie-compatible custom motif.
binder_design
Binder Design
Uses one selected file/source when file parameters are present. Design binders for a bundled benchmark target or a custom target structure.

Canonical Job Configuration

These are the fields exposed by the default job configuration for genie3. They are also returned by GET /api/v1/program/params?program=genie3 and submitted as the params JSON object to POST /api/v1/job/submit.

ParameterTypeModesWhat it does
samples
Samples
Integer All modes How many sequences or designs to sample.
Default: 1; Range: 1-32
length
Length
Integer Unconditional Generation Exact sequence length for unconditional generation.
Default: 50; Range: 50-300
motif_source
Motif Source
Text Motif Scaffolding Use the bundled benchmark or provide a Genie-compatible motif PDB plus segment layout.
Default: bundled; Options: bundled, custom
motif_selection
Motif Benchmark
Text Motif Scaffolding Bundled benchmark case for motif scaffolding.
Default: 22_1BCF; Options: 22_1BCF; Shown when motif_source is bundled
motif_structure
Motif PDB
Structure file Motif Scaffolding PDB containing REMARK 999 NAME and REMARK 999 INPUT segment lines, as expected by Genie 3.
Required; Files: .pdb; Shown when motif_source is custom
motif_layout
Segment Layout
Text Motif Scaffolding Comma-separated scaffold ranges and motif segment IDs from the motif PDB REMARK lines.
Required; Shown when motif_source is custom
binder_source
Target Source
Text Binder Design Use the bundled benchmark or provide a custom target PDB, sequence, and interface residues.
Default: bundled; Options: bundled, custom
binder_selection
Binder Benchmark
Text Binder Design Bundled benchmark case for binder design.
Default: 01_bhrf1; Options: 01_bhrf1; Shown when binder_source is bundled
target_structure
Target PDB
Structure file Binder Design Target structure for Genie 3 binder conditioning and template-mode evaluation.
Required; Files: .pdb; Shown when binder_source is custom
target_sequence
Target Sequence
Sequence Binder Design Target amino-acid sequence. Use colon-separated chains for multi-chain targets.
Required; Shown when binder_source is custom
target_segments
Target Segments
Residue selection Binder Design Target residue spans included in the binder-design context.
Required; Shown when binder_source is custom
hotspots
Hotspots
Residue selection Binder Design Target residues to condition the binder interface on.
Required; Shown when binder_source is custom
binder_length
Binder Length
Length or range Binder Design Length or length range for generated binder chains.
Required; Range: 3-500; Shown when binder_source is custom
folding_mode
Folding Mode
Text Binder Design Template mode uses the target PDB as the ColabFold template. MSA mode also requires a target MSA file.
Default: template; Options: template, msa; Shown when binder_source is custom
target_msa
Target MSA
MSA file Binder Design Target MSA used by Genie 3/ColabFold when folding mode is msa.
Required; Files: .a3m, .afa, .sto, .stk, .fa, .fasta; Shown when folding_mode is msa
Advanced configuration fields
ParameterTypeModesWhat it does
total_length
Total Length
Length or range Motif Scaffolding Optional total scaffolded protein length or range. Leave blank to sample from the segment layout ranges.
Range: 1-2000; Shown when motif_source is custom
extended_interface
Extended Interface
Residue selection Binder Design Optional broader interface residue set. Genie 3 defaults generation conditioning to this key.
Shown when binder_source is custom
problem_name
Problem Name
Text Motif Scaffolding, Binder Design Optional short key for generated custom Genie 3 problem files.
inverse_sequences
Inverse-Fold Sequences
Integer Binder Design ProteinMPNN sequences sampled per generated binder backbone during evaluation.
Default: 1; Range: 1-8
direction_scale
Direction Scale
Number All modes Optional sampler guidance strength. Leave blank for the workflow default.
Range: 0-2
experiment_name
Experiment Name
Text All modes Optional short name written into the generated Genie 3 config.

Outputs And Metrics

  • Generated backbone structures and associated metadata.
  • Result summaries for sampled designs when produced by the wrapper.
  • Use generated structures as candidates for sequence design, prediction, filtering, and manual structural inspection.

Common Examples

  • Unconditional generation: length 80, samples 8.
  • Motif scaffolding: bundled motif benchmark with optional total length range.
  • Binder design: custom target PDB, target sequence, interface residues, and sample count.

Example API params

{
  "mode": "unconditional_generation",
  "length": 80,
  "samples": 8
}

Caveats

  • Custom motif PDBs must use Genie-compatible REMARK lines and segment layout.
  • Binder interface residue lists must match the submitted target sequence/structure.
  • Backbones are proposals and need sequence design and independent validation.

Advanced Submit

Advanced submit is still available for direct program arguments through POST /api/v1/job/submit-advanced. Prefer canonical configuration unless you need exact low-level arguments or are reproducing a known command line.

  • Advanced submit can pass direct Genie configuration for custom campaigns.
  • Canonical configuration covers the common unconditional, motif, and binder modes.
curl -X POST https://subseq.bio/api/v1/job/submit \
  -H "Authorization: Bearer <api_key>" \
  -F program=genie3 \
  -F 'params={"mode":"unconditional_generation","length":80,"samples":8}'

Further Reading