Multi-Modal Foundation · mist-models
MIST 28M · BACE
Predicts beta-secretase 1 inhibition from molecular SMILES strings for Alzheimer's disease drug discovery.
What MIST 28M · BACE is
MIST 28M · BACE is a molecular property prediction model designed to classify whether small chemical compounds inhibit beta-secretase 1 (BACE-1), an enzyme targeted in Alzheimer's disease therapeutics. Developed by the Electrochemical Energy Group at the University of Michigan, Ann Arbor, the model belongs to the Molecular Insight SMILES Transformers (MIST) family.
The model builds upon a 28-million-parameter encoder pre-trained on molecular SMILES from the Enamine REAL Space database. By fine-tuning this foundation encoder on the MoleculeNet BACE benchmark, the model assists researchers in categorising potential inhibitors and prioritising chemical candidates for experimental testing.
How it works
The underlying architecture is an 8-layer RoBERTa-PreLayerNorm encoder with 8 attention heads, a hidden size of 512, an intermediate size of 2048, and a maximum sequence length of 2048 tokens. It was pre-trained using a masked language modelling objective with cross-entropy loss and the FusedLAMB optimizer on SMILES strings from Enamine REAL Space.
For binary classification, the pre-trained encoder is coupled with a task network consisting of a two-layer multi-layer perceptron with Gaussian Error Linear Unit (GELU) activations and dropout. Sequence representations are pooled using the hidden state of the first token. The full architecture was fine-tuned with binary cross-entropy loss and the AdamW optimizer on 1,513 small molecules from the MoleculeNet BACE dataset.
Key facts
- Model architecture
- RoBERTa-PreLayerNorm encoder (8 layers, 8 heads) with a 2-layer MLP task network
- Parameter count
- 28M
- Context length
- 2048 tokens
- Pre-training data
- Enamine REAL Space
- Fine-tuning data
- MoleculeNet BACE benchmark (1,513 small molecules)
- Input format
- Kekulized SMILES strings
- Output
- Binary classification logits or pooled representation vectors
What you can use it for
- BACE-1 inhibitor screening
- Classify candidate small molecules to identify probable inhibitors of beta-secretase 1 for Alzheimer's disease drug discovery.
- Virtual library filtering
- Screen batches of Kekulized SMILES strings to prioritise active chemical series before wet-lab assay evaluation.
- Molecular embedding extraction
- Derive pooled sequence vector embeddings of input compounds for downstream analytical or clustering workflows.
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
MIST 28M · BACE
Strengths and limits
Good at
- Initialised from an encoder pre-trained on Enamine REAL Space, capturing synthetically accessible chemical space.
- Evaluated using an 80/10/10 scaffold split to test performance across distinct structural frameworks.
- Supports molecular sequence lengths up to 2048 tokens, accommodating larger chemical representations.
Watch out for
- Fine-tuning was conducted on a relatively small dataset of 1,513 molecules.
- Tokenizer implementation relies on the external SMIRK library, which requires a Rust environment to install.
Inputs and outputs
You provide
Sequence or text
A protein or nucleotide sequence, or free text — whichever this model was trained on.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · See model card
The See model card licence carries terms worth reading before you depend on it.
Questions
Can I use MIST 28M · BACE commercially?
Check the licence before commercial use. The See model card licence carries terms worth reading before you depend on it.
How do I run MIST 28M · BACE 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 CPU only, typically in 12s, and a run costs about 28 credits. Visitors without an account get 1,000 credits a month.
What inputs does the model expect?
The model expects molecular SMILES strings, specifically formatted in Kekulized SMILES, processed using the SMIRK tokenizer.
What biological target does the model address?
It targets beta-secretase 1 (BACE-1), an enzyme implicated in Alzheimer's disease pathology.
What is the primary evaluation metric used for this model?
Model performance is evaluated using the Area Under the Receiver Operating Characteristic curve (AUROC) on an 80/10/10 scaffold split of the BACE dataset.
What does the model output during inference?
The model outputs unnormalised logits for binary BACE-1 inhibition, or a mean-pooled embedding vector when utilised as a feature extractor.
How to cite it
Wadell, A., Bhutani, A., Azumah, V., Ellis-Mohr, A. R., Kelly, C., Zhao, H., Nayak, A. K., Hegazy, K., Brace, A., Lin, H., Emani, M., Vishwanath, V., Gering, K., Alkan, M., Gibbs, T., Wells, J., Varshney, L. R., Ramsundar, B., Duraisamy, K., Mahoney, M. W., Ramanathan, A., & Viswanathan, V. (2025). Foundation Models for Discovery and Exploration in Chemical Space. arXiv:2510.18900 [physics]. https://doi.org/10.48550/arXiv.2510.18900
Written from MIST 28M · BACE'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.