Optima: a GUI-based toolkit for developing and validating interpretable machine learning based supervised...

Area of Science:

• Computational Chemistry
• cheminformatics
• Machine Learning in Drug Discovery

Background:

• Machine learning (ML) algorithms and large datasets are revolutionizing predictive modeling in scientific research.
• Developing interpretable ML models for QSAR/QSPR/QSTR remains a challenge for researchers with limited coding expertise.

Purpose of the Study:

• To introduce Optima, a Python-based GUI toolkit designed to streamline the development of interpretable ML-based classification QSAR/QSPR/QSTR models.
• To provide researchers with a user-friendly platform for optimizing, constructing, and interpreting ML classification models.

Main Methods:

• Optima offers an intuitive GUI with customizable settings for plot aesthetics.
• The toolkit facilitates rational dataset splitting, efficient feature selection, and the development of seven distinct ML classification QSAR models.
• It covers the entire workflow from model optimization to interpretation, ensuring transparency and reproducibility.

Main Results:

• Optima simplifies the process of building and interpreting ML-based classification QSAR models.
• The toolkit enhances model performance through robust optimization and an explainable approach.
• It empowers users with domain knowledge but limited coding skills to develop sophisticated predictive models.

Conclusions:

• Optima addresses a critical need for an accessible and efficient tool in QSAR/QSPR/QSTR modeling.
• The toolkit promotes transparency and reproducibility in ML-based predictive modeling.
• Optima is available for Windows and can be downloaded from GitHub.