A Machine Learning Prediction Model for the Affinity Between Glucose and Binder

Abstract

The glucose is an important source of fuel for the body. The binding affinity is an essential indicator of the interaction of a glucose molecule with its binder. This paper proposes a novel machine learning model for predicting the binding affinity of a small glucose molecule with the binder. Seven regression algorithms were compared on a dataset is generated based on Molecular Mechanics-Generalized Born and Surface Area (MM-GBSA). Through the comparison, Random Forest and Decision Tree were selected for our model, in light of their robustness and accuracy. The established model predicts binding affinity from the interaction properties of compounds and glucose, which are obtained through GLIDE program from Schrödinger software suite 2018-4. Finally, the prediction accuracy of our model was confirmed through k-fold cross-validation. Our research provides an efficient and low-cost method for screening of molecules during the development of glucose binders.

Comparative Analysis of Naive Bayes and K-NN Approaches to Predict Timely Graduation using Academic History

“…The primary goal of data cleaning is to eliminate irrelevant or noisy data. This process includes tasks such as removing duplicate records and handling missing values [29]. Overall, data cleaning ensures that the dataset is accurate, consistent, and suitable for predictive analysis.…”