HYBRID EVENT: You can participate in person at Baltimore, Maryland, USA or Virtually from your home or work.
Richmond Essieku, Speaker at Obesity Conferences
Texas Tech University, United States


Background - Finding pharmacological targets for the treatment and prevention of obesity has gained more attention as a result of the significant burden that obesity places on patients and the healthcare system. This involves understanding the biological mechanisms that contribute to obesity and developing drugs or other interventions that target these mechanisms. Predictive models that predict obesity by using genetic variations are, however, lacking.

Objectives - In this paper, we intended to identify dominant predisposing predictors to model the Hypothalamic Pituitary Adrenal (HPA) axis function for obesity. These will be used to develop a robust predictive model to predict the risk for obesity based on an individual’s gene profile by exploring statistical and data mining techniques.

Methods - In this regard, we incorporate two main techniques: Significance Analysis of Microarray (SAM) and Machine Learning (ML) based approaches. We employ recursive feature elimination cross-validation - support vector machine method for the feature engineering process. Additionally, these features were used to build six ML approaches namely logistic regression, k-NN, naive bayes, random forest, gradient boosting and multilayer perceptron neural network classifier.

Results - The Multilayer Perceptron Neural Network (MLP) Classifier yielded the highest median accuracy (83.13%) together with the following highest evaluation metrics; area under the receiver operating characteristics curve - AUC (median: 75%), recall (median: 90%), precision (median: 85%), and the fastest model’s execution time at 1.87 seconds. SAM and MLP analyses had identified 13 genes which were associated with obesity-related traits, and hence may be of highly potential biomarkers and could therefore become targets for the treatment or prevention of obesity.

Audience take-away:

  • The study discovered gene biomarkers which are essential resources and can provide crucial information to represent a logical adjunct to improve diagnosis and clinical trial of obesity treatment.
  • A robust predictive model that can determine if an individual has a risk factor for obesity.
  • The study’s findings could serve as pharmacological targets for developing medications for the treatment and prevention of obesity.
  • The project contributes to the healthcare field and benefit all stakeholders including students and researchers because of low frequency of research publications relating to this topic due to insufficient data on the part of Hypothalamic Pituitary Adrenal (HPA) Axis for obesity.
  • Learn more about the Significance Analysis of Microarray (SAM) method and how to use it together with some specific machine learning algorithms.


Richmond Essieku is a health data analyst who earned his master's degree in applied statistics and data science at the University of Texas Rio Grande Valley and is now pursuing a PhD at Texas Tech University. His research area focused on public health especially on obesity research, health data analytics, health economics and machine learning.