Multi-Modal Foundation · mist-models
MIST 28M · HIV
A molecular transformer predicting HIV-1 replication inhibition in T-cell lines from SMILES strings.
What MIST 28M · HIV is
MIST 28M · HIV is a molecular property prediction model developed by the Electrochemical Energy Group at the University of Michigan, Ann Arbor. It adapts the MIST (Molecular Insight SMILES Transformers) foundation model to screen chemical structures for potential antiretroviral activity.
The model specifically addresses the classification of small molecules according to whether they inhibit human immunodeficiency virus type 1 (HIV-1) replication. It was fine-tuned on the MoleculeNet HIV benchmark dataset, which encompasses 41,127 compounds tested in T-cell lines.
By mapping chemical structures represented as SMILES strings to inhibition predictions, the model provides computational chemists and medicinal researchers with an automated method to evaluate candidate compounds.
How it works
The model uses a RoBERTa-PreLayerNorm architecture comprising 8 transformer layers, a hidden dimension of 512, an intermediate dimension of 2,048, and 8 attention heads, supporting sequence lengths up to 2,048 tokens. It was pre-trained on Kekulized SMILES strings from the Enamine REAL Space compound library using a masked language modelling objective with the FusedLAMB optimiser.
For HIV activity prediction, the base encoder is fitted with a task network consisting of a two-layer multi-layer perceptron (MLP) with Gaussian Error Linear Units (GELU) activations and dropout. Hidden states are pooled using the representation of the first sequence token, and the network is fine-tuned using binary cross-entropy loss and the AdamW optimiser on an 80/10/10 scaffold split.
Key facts
- Architecture
- RoBERTa-PreLayerNorm encoder with two-layer MLP task head
- Parameter count
- 28M
- Context length
- 2,048 tokens
- Pre-training dataset
- Enamine REAL Space
- Fine-tuning dataset
- MoleculeNet HIV (41,127 compounds)
- Input format
- Kekulized SMILES strings
What you can use it for
- HIV-1 Inhibition Screening
- Score batches of small-molecule SMILES strings to generate unnormalised logits for binary anti-HIV activity prediction.
- Chemical Feature Extraction
- Extract pooled vector representations from candidate molecules for clustering or training downstream predictive models.
- Lead Prioritisation
- Filter and rank candidate chemical libraries tested against HIV replication before carrying out laboratory assays.
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 · HIV
Strengths and limits
Good at
- Pre-trained on Enamine REAL Space, capturing synthetically accessible chemical regions.
- Accommodates molecular sequences up to 2,048 tokens.
- Validated on MoleculeNet using a scaffold split to ensure generalisation across structural chemotypes.
Watch out for
- Restricted to binary classification of HIV-1 replication inhibition in T-cell lines.
- Expects input molecules formatted specifically as Kekulized SMILES strings.
- The required SMIRK tokeniser necessitates an installed Rust environment.
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 · HIV 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 · HIV 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 MIST 28M · HIV accept?
The model accepts chemical structures represented as text strings, specifically Kekulized SMILES.
What does the model output during inference?
Calling property prediction returns unnormalised logits for binary classification of HIV inhibition activity, while pipeline feature-extraction outputs pooled embedding vectors.
What evaluation metric is used for this model?
The model's classification performance is measured using the Area Under the Receiver Operating Characteristic Curve (AUROC) on an 80/10/10 scaffold split.
What tokenizer does the model require?
It uses the SMIRK tokeniser, which requires a functioning Rust installation to build and run.
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., and Viswanathan, V. Foundation Models for Discovery and Exploration in Chemical Space. arXiv preprint arXiv:2510.18900 (2025). https://doi.org/10.48550/arXiv.2510.18900
Written from MIST 28M · HIV'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.