Analysis of Bacterial Association Patterns that trigger Bacterial Vaginosis
Keywords:Bacterial Vaginosis, Machine learning, Association rules, Quality metrics, ARules package functions.
Background: Bacterial Vaginosis (BV) is a dysbiosis of the normal flora residing in the patient’s vaginal mucosa. Objective: Running the apriori algorithm to mine association rules in a dataset that holds records of patients diagnosed with BV+. Method: To select the rules created with statistical significance the functions is.redundant, is.significant, and is.maximal were used. Also eight quality metrics were used. Results: The best percentage of support to find the frequent itemsets was 7%. The confidence percentage to create the rules was 90%. The best metric was Fisher’s exact test. The algorithm reported 58 rules. After selection with the fucntions and metrics, 17 rules were reported. Biological validation reduced the rules to 5. Rules reported that Atopobium vaginae, Gardnerella vaginalis, Megasphaera phylotype 1, and Ureaplasma parvum interact with each other to develop BV+. Conclusion: Knowing the bacteria (patterns) involved in the development of BV supports in the diagnosis of BV+.