Exploring proteasome inhibition using atomic weighted vector indices and machine learning approaches.

Area of Science:

• Biochemistry
• Cheminformatics
• Computational Biology

Background:

• The ubiquitin-proteasome system (UPS) regulates crucial cellular processes, including gene transcription and cell cycle progression.
• Dysregulation of the UPS is implicated in various diseases, making proteasome inhibition a therapeutic target.
• Cheminformatics and AI are increasingly used to predict proteasome ubiquitination pathway (UPP) inhibitors.

Purpose of the Study:

• To develop and validate a novel set of molecular descriptors, atomic weighted vectors (AWV), for modeling proteasome inhibition.
• To assess the efficacy of AWV-based datasets in predicting proteasome inhibitor activity (EC50).
• To compare the performance of various machine learning algorithms using these new descriptors.

Main Methods:

• Utilized a new cheminformatics tool to generate atomic weighted vectors (AWV) as molecular descriptors.
• Developed datasets based on AWV for training and evaluating machine learning models.
• Applied diverse machine learning techniques including linear regression, multiple linear regression (MLR), random forest (RF), K-nearest neighbors (IBK), multi-layer perceptron, best-first search, and genetic algorithm.

Main Results:

• The study successfully generated a novel set of atomic descriptors (AWV).
• AWV-based datasets demonstrated adequate modeling capabilities for proteasome inhibitors.
• Various artificial intelligence techniques effectively utilized AWV for predicting inhibitory activity.

Conclusions:

• Atomic weighted vectors (AWV) provide a promising approach for developing predictive models of proteasome inhibitors.
• The integration of AWV with machine learning offers an efficient strategy for drug discovery targeting the UPS.
• This method facilitates the development of novel therapeutic agents through accurate prediction of inhibitory activity.