QUEEN (QSAR Under Effective and Efficient Neural-Networks) is a product designed to automate construction, training, and deployment of predictive models for QSAR/QSPR (Quantitative Structure–Activity/Property Relationship) applications.
Although several statistical fitting methods are commonly used to build QSAR models, artificial neural networks are by far the most powerful and versatiles algorithms to handle this kind of task. The complexity of these models can be increased or reduced as needed and, with a proper parametrization, they are able to accurately learn almost anything.
QUEEN gathers any phase of the modelling process from the data import to the model deployment with a user-friendly interface.
The minimum information required to effectively use this software is a dataset including the following components: Target variable and Matrix of molecular descriptors.
Anytime QUEEN is executed in line with FAST, all the previous information is automatically taken from the FAST output.
- a simple and intuitive import method for training data
- a fully automated hyperparameters optimization algorithm based on Bayesian method
- a model training section
- the possibility to merge hundreds of individual models into a single response
This approach guarantees robust and performing results in any conditions while keeping a simple and almost parameterless setup.
The tool is completed by an export function that allows the final model saving. The exported file can be then transferred, reloaded, and used to predict the target property on any new sample. Each prediction comes with an applicability domain score that indicates the reliability of the model outcome for the given sample.
The exported model can be reimported and run both on QUEEN and QUEEN Runner, the free licence version of QUEEN made for running existing models only.
Finally, this software is part of the suite ALChemy (Automated Learning for Chemistry), a unified framework for a variety of QSAR modeling tasks