After a long period of development and testing, the QsarDB repository has added support for the Open Neural Network Exchange (ONNX) format. Please note that the preferred and core model representation format for QsarDB is still Predictive Model Markup Language (PMML). PMML works well for represent...
Each model in QsarDB can be used for prediction if the user calculates or determines experimentally the descriptors in the model and enters them in the “QDB Predictor” application [1] or via the REST service [2]. For some models, the prediction service also works directly from the structure. In this case, the QsarDB repository calculates the required molecular descriptors on the fly. Currently, QsarDB supports this functionality for a limited number of models, and they use open source software (Chemistry Development Kit, PaDEL-Descriptor and XlogP3) to calculate molecular descriptors.
Lisa Bang has developed a simple QsarDB archive parser for the Python programming language. This parser can be useful to someone interested in reproducing the QSAR provided in the QsarDB archive format.