SolubleMPNN extends ProteinMPNN with solubility-aware training/objectives to bias designed sequences toward improved expression and solubility.
Type
Inverse folding + solubility bias
Goal
Higher expression/solubility
Input
Backbone structure (experimental or predicted)
Output
Designed sequences
Resources
- Reference impls: community forks built on ProteinMPNN
- Related: NetSolP (solubility prediction) — paper
If no public canonical repo is available, implement as a ProteinMPNN post-filter or multi-objective scoring with a solubility model.