Predicting quorum-sensing properties by the properties of the amino acids in gram-positive bacteria

Quorum sensing properties in peptides are incredibly important for gram-positive bacteria, as it enables those bacteria to communicate with each other. Such communication happens by means of chemical signals. Group behavior coordination is common among bacteria that are able to effectively communicate and sense their surroundings. In this study, a dataset composed of 1229 peptides (1004 negative and 225 positive), was collated with the properties of certain amino acids. The purpose of this study was to determine whether there was any correlation between the presence of amino acids with certain properties and the occurrence of quorum sensing properties in bacteria. The models used were the “Logistic Regression”, “Support Vector Machine”, “Random Forests”, “Gradient Boosting”, and a “Decision Tree”. Although with good results in most areas, the one area in which all of the models had difficulty was the Area Under the ROC Curve (AUC), with scores ranging from 0.443 to 0.5. The most probable explanation to the myriad of answers, and low overall AUC, would be underfitting. Quorum sensing properties are complex, and thus there should be a workflow capable of understanding the patterns that show up when analyzing the dataset.