Antibody & Biologics · IBM Research
MAMMAL solubility
Binary protein solubility classification directly from amino-acid sequences.
What MAMMAL solubility is
MAMMAL solubility is a machine learning model developed by IBM Research to predict protein solubility. Protein solubility is a critical attribute during both drug discovery and biopharmaceutical manufacturing, as poor solubility can undermine a protein's function, stability, and production yield.
The model operates as a binary classifier that takes an amino-acid sequence as its sole input and predicts whether the protein is soluble. It serves as an in silico screening tool to evaluate candidate molecules early in the development pipeline.
The model is a fine-tuned checkpoint of IBM's MAMMAL foundation model architecture (specifically biomed.omics.bl.sm.ma-ted-458m), adapted explicitly for solubility prediction benchmarks.
How it works
The model is built on IBM's MAMMAL (Molecular Aligned Multi-Modal Architecture and Language) framework, fine-tuning the 458-million-parameter base model biomed.omics.bl.sm.ma-ted-458m for sequence-based binary classification.
It processes raw amino-acid sequences using a modular tokenizer and generates prediction tokens and classification scores through an autoregressive generation routine. The model was trained using a benchmark dataset retrieved from Zenodo (record 1162886) based on the standard bioinformatics solubility benchmark.
Key facts
- Base architecture
- MAMMAL (biomed.omics.bl.sm.ma-ted-458m)
- Input modality
- Protein amino-acid sequence
- Output modality
- Binary classification table and prediction scores
- Memory requirement
- 24 GB GPU memory
- Release date
- 28 October 2024
- Model version
- 1.0
What you can use it for
- In silico solubility filtering
- Filter large libraries of designed or engineered protein sequences to eliminate candidates with poor predicted solubility prior to laboratory expression.
- Antibody developability screening
- Assess whether therapeutic antibody candidates meet early developability criteria regarding protein solubility.
- Protein expression triage
- Identify potentially insoluble proteins from primary sequence data to select suitable production hosts or expression conditions.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on On-demand GPU
Runs on our servers with on-demand compute. A first run needs time to load the model; active capacity can be reused and scales down when idle.Run this model
MAMMAL solubility
Strengths and limits
Good at
- Requires only the primary amino-acid sequence as input, eliminating the need for experimental or predicted 3D structures.
- Leverages pre-trained representations from the 458-million-parameter MAMMAL biomolecular foundation model.
- Provides output prediction scores alongside binary classifications to gauge prediction confidence.
Watch out for
- Restricted to binary classification rather than predicting continuous solubility values, aggregation rates, or concentration thresholds.
- Requires substantial computational resources, demanding at least 24 GB of GPU memory for inference.
- Relies on a single standard benchmark dataset without incorporating environmental parameters such as pH, temperature, or buffer composition.
Inputs and outputs
You provide
Protein sequence
Single-letter amino acid code. Whitespace and FASTA headers are stripped automatically.
You get back
Binary solubility classification from the MAMMAL protein_solubility checkpoint.
Licence
Commercial use permitted · Apache-2.0
The Apache-2.0 licence allows use in commercial research and products.
Questions
Can I use MAMMAL solubility commercially?
Commercial use permitted. The Apache-2.0 licence allows use in commercial research and products.
How do I run MAMMAL solubility without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on 1× A100 40GB, typically in 6.0s, and a run costs about 112 credits. Visitors without an account get 1,000 credits a month.
What inputs does the model require?
The model requires only the standard amino-acid sequence of the protein (such as a FASTA string).
What is the output format of the prediction?
The model outputs a binary classification indicating whether the protein belongs to the soluble or insoluble class, accompanied by generation scores.
What hardware is necessary to run inference with this model?
A GPU with at least 24 GB of memory is required to execute the model.
How to cite it
Yoel Shoshan, Moshiko Raboh, Michal Ozery-Flato, Vadim Ratner, Alex Golts, Jeffrey K. Weber, Ella Barkan, Simona Rabinovici-Cohen, Sagi Polaczek, Ido Amos, Ben Shapira, Liam Hazan, Matan Ninio, Sivan Ravid, Michael M. Danziger, Joseph A. Morrone, Parthasarathy Suryanarayanan, Michal Rosen-Zvi, Efrat Hexter. MAMMAL -- Molecular Aligned Multi-Modal Architecture and Language. arXiv:2410.22367 [q-bio.QM], 2024.
Written from MAMMAL solubility's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.