Publicly Available Models§

Below an overview of our available open-source models. Find them on GitHub here.
Go to Deploying Models for more detailed deployment options beyond the quick-start.

SMI-TED / Property Prediction on SMILES Input§

openad-service-smi-ted compose.yml Instructions

This OpenAD service provides access to SMILES-based Transformer Encoder-Decoder (SMI-TED), a foundation model for materials science and chemistry. SMI-TED is an encoder-decoder model pre-trained on a curated dataset of 91 million SMILES samples sourced from PubChem, equivalent to 4 billion molecular tokens. SMI-TED offers several predictive models, including quantum property prediction, with two main variants ( 289 M and 8 × 289 M ).

More information:
github.com/IBM/materials
huggingface.co/ibm/materials.smi-ted
arxiv.org/abs/2407.20267

Support for:
✅ Docker / Podman Compose
✅ Docker / Podman
✅ Google Cloud Run
✅ Apple Silicon - more info

Quick start with Docker Compose:

Text Only

curl -O https://raw.githubusercontent.com/acceleratedscience/openad-service-smi-ted/main/compose.yaml

Text Only

docker compose create

Text Only

docker compose start

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as smi_ted

BMFM-SM / Property Prediction on SMILES Input§

openad-service-bmfm-sm compose.yml Instructions

This OpenAD service provides access to BioMedical Foundation Models: Small Molecules (BMFM-SM), namely the Biomed-multi-view foundation model. BMFM-SM has models for predicting many properties from the well-known MoleculeNet benchmarks:

BACE	BBBP	CLINTOX	ESOL	FREESOLV	HIV

LIPOPHILICITY	MUV	QM7	SIDER	TOX21	TOXCAST

More information:
github.com/BiomedSciAI/biomed-multi-view
arxiv.org/abs/2410.19704

Support for:
✅ Docker / Podman Compose
✅ Docker / Podman
☹️ Google Cloud Run
✅ Apple Silicon - more info

Quick start with Docker Compose:

Text Only

curl -O https://raw.githubusercontent.com/acceleratedscience/openad-service-bmfm-sm/main/compose.yaml

Text Only

docker compose create

Text Only

docker compose start

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as bmfm_sm

BMFM-PM / Property Prediction on FASTA Input using MAMMAL§

openad-service-bmfm-pm compose.yml Instructions

This OpenAD service provides access to the Biomed-multi-alignment foundation model, with two models for protein property prediction that take FASTA string input: protein solubility (Sol) and drug-target interaction (DTI), which takes SMILES for the drug input and FASTA for the target input.

Sol task is from benchmark data defined here: https://academic.oup.com/bioinformatics/article/34/15/2605/4938490
DTI task is from benchmark data from TD Commons: https://tdcommons.ai/multi_pred_tasks/dti/

More information:
github.com/BiomedSciAI/biomed-multi-alignment

Support for:
✅ Docker / Podman Compose
✅ Docker / Podman
☹️ Google Cloud Run
☹️ Apple Silicon - more info

Quick start with Docker Compose:

Text Only

curl -O https://raw.githubusercontent.com/acceleratedscience/openad-service-bmfm-pm/main/compose.yaml

Text Only

docker compose create

Text Only

docker compose start

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as bmfm_pm

REINVENT 4 / Generative Models with SMILES Output§

openad-service-reinvent4 compose.yml Instructions

This OpenAD service provides access to the REINVENT 4 molecular design tool, which is used for de novo design, scaffold hopping, R-group replacement, linker design, molecule optimization, and other small molecule design tasks. REINVENT uses a Reinforcement Learning (RL) algorithm to generate optimized molecules compliant with a user-defined property profile defined as a multi-component score. Transfer Learning (TL) can be used to create or pre-train a model that generates molecules closer to a set of input molecules.

More information:
github.com/MolecularAI/REINVENT4
link.springer.com/article/10.1186/s13321-024-00812-5

Support for:
✅ Docker / Podman Compose
✅ Docker / Podman
✅ Google Cloud Run
☹️ Apple Silicon - more info

Quick start with Docker Compose:

Text Only

curl -O https://raw.githubusercontent.com/acceleratedscience/openad-service-reinvent4/main/compose.yaml

Text Only

docker compose create

Text Only

docker compose start

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as reinvent4

Generation / Generative Models with SMILES or SELFIES Output§

openad-service-gen Instructions

This OpenAD service provides access to generative algorithms that output SMILES or SELFIES.

Regression Transformer, 2023. Uses transformers for both regression and generation. Generates SELFIES based on desired properties.
PaccMann, 2020. Uses autoencoders to generate molecules to target cancer based on omics profiles.
TorchDrug, 2021. Offers two kinds of graph-based networks to generate SMILES: GCPN and GraphAF.
MOSES, 2020. GuacaMol, 2019. And more.

These generative algorithms were previously offered in the open source library, GT4SD.

Support for:
☹️ Docker / Podman Compose
✅ Docker / Podman
✅ Google Cloud Run
☹️ Apple Silicon - more info

Quick start with Docker:

Text Only

git clone https://github.com/acceleratedscience/openad-service-gen

Text Only

cd openad-service-gen

Text Only

docker build -t gen .

Text Only

docker run -p 8080:8080 gen

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as gen

Properties / Property Prediction on SMILES, FASTA or CIF Input§

openad-service-prop Instructions

This OpenAD service provides access to property predictive models in computational chemistry (some in the area of small molecules, others in proteins) and in materials science. For small molecules, input is in SMILES format. For proteins, input is in FASTA format. For materials science, input is in CIF file format. These models were previously offered in the open-source library, GT4SD.

Support for:
☹️ Docker / Podman Compose
✅ Docker / Podman
☹️ Google Cloud Run
☹️ Apple Silicon - more info

Quick start with Docker:

Text Only

git clone https://github.com/acceleratedscience/openad-service-prop

Text Only

cd openad-service-prop

Text Only

docker build -t prop .

Text Only

docker run -p 8080:8080 prop

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as prop

MoLeR / SMILES Generation on SMILES Scaffolds§

openad-service-moler Instructions

This OpenAD service provides access to MoLeR, a model for conditional, small-molecule (SMILES) generation based on molecular scaffold input (also SMILES). MoLeR uses TensorFlow as its deep learning platform.

More information:
github.com/microsoft/molecule-generation
arxiv.org/abs/2103.03864

Support for:
☹️ Docker / Podman Compose
✅ Docker / Podman
☹️ Google Cloud Run
☹️ Apple Silicon - more info

Quick start with Docker:

Text Only

git clone https://github.com/acceleratedscience/openad-service-moler

Text Only

cd openad-service-moler

Text Only

docker build -t moler .

Text Only

docker run -p 8080:8080 moler

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as moler

MoLFormer / Property Prediction Highlights on SMILES Input§

openad-service-molf Instructions

MoLFormer / Three finetuned property prediction models on MoleculeNet tasks: BACE classification, ClinTox multiclass, and QM9 alpha regression.

Support for:
☹️ Docker / Podman Compose
✅ Docker / Podman
☹️ Google Cloud Run
☹️ Apple Silicon - more info

Quick start with Docker:

Text Only

git clone https://github.com/acceleratedscience/openad-service-molf

Text Only

cd openad-service-molf

Text Only

docker build -t molf .

Text Only

docker run -p 8080:8080 molf

Text Only

openad

Text Only

catalog model service from remote 'http://127.0.0.1:8080' as molf