Property prediction

ProteinMPNN-ddG

Estimate changes in protein stability upon point mutation

A modification of ProteinMPNN to use full sequence context. It introduces a decoding scheme to improve computational efficiency and enable saturation mutagenesis studies at scale.

Documentation ↗ Input parameters ↗ Examples ↗ Home page ↗ Paper ↗ Source code ↗ License ↗

No jobs before this; it will run immediately.

Not logged in. Register an account to keep track of your jobs and results, and add presets and bookmarks. Email results to this address when complete:

Ready

Example preset

Choose an example to fill in its settings and input structures.

Inputs

Official documentation ↗

Job name A label for this run and its results.

Protein structure PDB structure whose point mutations will be scored. In this form: Upload PDB text or choose the preset containing the official AlphaFold example structure.

Chains to load Comma-separated PDB chains used as structural and sequence context, e.g. A,B,C. By default, substitutions are predicted for the first chain.

Chain to predict Optional chain whose substitutions are predicted. It is moved to the front of the loaded chain list; blank uses the first chain above.

Model and averaging

Official documentation ↗

ProteinMPNN model 48-neighbor ProteinMPNN checkpoint trained with the indicated backbone noise.

Random seed Seed used to split the per-repeat JAX random keys.

Model repeats Run with this many keys split from the input seed and average the resulting predictions.

Disable the ddG correction Write raw logit differences without the paper's single-residue ddG correction. The correction is defined only for v_48_020, so this must be enabled with every other model.

These local CLI options are not applicable to Bio Web and are omitted from its inputs:

outpath: The service creates a separate output path for each job and archives the prediction CSV.

Ready