SolubleMPNN Protein Design FAQs

Question 1

Can SolubleMPNN handle protein-ligand complexes?

Accepted Answer

SolubleMPNN is optimized for protein-only design. If your target includes a small molecule or ligand, you should use the LigandMPNN skill instead.

Question 2

Is this tool suitable for mammalian expression systems?

Accepted Answer

While it can be used, SolubleMPNN is specifically tuned for E. coli expression. For mammalian systems, post-translational modifications and standard ProteinMPNN models are often more relevant.

Question 3

How does SolubleMPNN differ from standard ProteinMPNN?

Accepted Answer

While standard ProteinMPNN focuses on structural recovery, SolubleMPNN uses weights specifically trained to prioritize soluble expression and reduce hydrophobic surface patches that cause aggregation.

Question 4

Which SolubleMPNN model should I use?

Accepted Answer

The v_48_020 model is recommended for standard E. coli expression optimization, while the v_48_030 variant is designed for particularly difficult targets requiring maximum solubility bias.

Question 5

What are the hardware requirements for running SolubleMPNN?

Accepted Answer

It requires a GPU with at least 8GB of VRAM (16GB recommended). Alternatively, it can be run via Modal to utilize cloud-based T4 GPUs without local setup.

SolubleMPNN Protein Design

Key Features

Use Cases

SolubleMPNN Protein Design

Key Features

Use Cases