Validating KNIME Workflows

This workflow demonstrates a technique to deploy a model prediction workflow as a web service and to validate that it is doing what it's supposed to do. The key is the use of a reference set of input data and the Table Difference Checker in the "Run Tests if Necessary" component at the end of the workflow. The model can be validated in KNIME Analytics Platform using the KNIME Testing Framework and, once it has been deployed as a web service using KNIME Server, can be validated by calling the service without any rows.

The workflow itself generates bioactivity predictions for the input molecules using a set of pre-calculated and saved XGBoost models. It is configured so that it can be easily deployed as a REST web service using KNIME server or used as a predictor directly in KNIME Analytics Platform

Extensions required;
KNIME Testing Framework
RDKit
XGBoost