AI/ML API¶

Prerequisites¶

Before proceeding with the installation or deployment of the AI/ML API, ensure you have the following prerequisites:

Python 3.9
Anaconda
PipelineAlternative_clinicaldata data
PBPK (Physiologically Based Pharmacokinetic) data
cddd (Compound Database for Drug Discovery) data

Installation¶

Clone the Repository¶

git clone git@github.com:ALTERNATIVE-EU/ai-ml-api.git

Prepare Data Directories¶

Place the required data directories in their respective locations:

PipelineAlternative_clinicaldata models directory
cddd directories inside PipelineAlternative_clinicaldata
PBPK directory inside models

Copy the contents of the patches directory into the models directory:

cp -r patches/* ./models/

Set up Virtual Environments¶

cddd Environment¶

Create and activate a virtual environment for cddd:

cd cddd
conda env create -f environment.yml
conda activate cddd
pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.10.0-cp36-cp36m-linux_x86_64.whl
pip install -e .
conda deactivate

Alternative Environment¶

Create and activate the alternative virtual environment:

conda create -n alternative python=3.9
conda activate alternative

Install dependencies from the requirements file:

pip install -f requirements.txt

Start the Server¶

Launch the API server by running:

python3 app.py

Installation with Docker¶

Build and run the AI/ML API container using Docker:

Build the image:
```
docker build -t ai-ml-api .
```
Run the container:
```
docker run -p 5000:5000 ai-ml-api
```

Deployment in Kubernetes¶

Update and apply the deployment files located in the deployment/kubernetes directory:

kubectl apply -f deployment/kubernetes

For authentication, use Istio with a custom Envoy filter. Follow the instructions provided in the envoy-filter documentation.

API Usage¶

Endpoints¶

ML Evaluation¶

curl -X POST -H "Content-Type: application/json" -d '{"smiles": "c1ccccc1O"}' https://ai-ml-api.platform.alternative-project.eu/clinicaldata/ml/evaluate -o results.csv

AI Evaluation¶

curl -X POST -H "Content-Type: application/json" -d '{"smiles": "c1ccccc1O"}' https://ai-ml-api.platform.alternative-project.eu/clinicaldata/ai/evaluate

AOP (Alternate Operating Procedure) Evaluation¶

curl -X POST -H "Content-Type: application/json" -d '{"smiles": "c1ccccc1O"}' https://ai-ml-api.platform.alternative-project.eu/clinicaldata/aop/evaluate

Doxorubicin PBPK Model¶

curl -X POST https://ai-ml-api.platform.alternative-project.eu/pbpk/doxorubicin      -H "Content-Type: application/json"      -d '{"dose_mg": 60, "age": 50, "weight": 70, "height": 190}'

HTTK (Human Toxicology Toolkit) Model¶

curl -X POST https://ai-ml-api.platform.alternative-project.eu/pbpk/httk -H "Content-Type: application/json"      -d '{"chem_name": "Bisphenol A", "species": "human", "daily_dose": 1, "doses_per_day": 1, "days": 15}'

IsAlive Check¶

Verify if the server is alive:

curl https://ai-ml-api.platform.alternative-project.eu/isalive

Testing¶

To run unit tests, use the following command:

python -m unittest app_test.py