Using machine learning to explore the genetics of tobacco addiction: A study of an African American population
Year: 2024
Type: other
Abstract: The World Health Organization estimates that smoking-related lung cancer account for more than 7 million deaths worldwide each year. Cigarette smoking is also a risk factor for various health issues including respiratory as well as other infections, osteoporosis, reproductive disorders, adverse postoperative events and delayed wound healing, duodenal and gastric ulcers, and diabetes. Because of these dangers, many tobacco users wish to quit smoking. Therefore, understanding the genetic factors underlying tobacco addiction could be a key step in developing effective therapies to fight it. The challenge lies in accurately predicting the phenotype solely on individual genetic variants. Studies indicate that the heritability of smoking ranges from 21% to 84% suggesting that it is possible to identify genetic factors associated with tobacco addiction. Approaches such as genome-wide Linkage analysis, candidate gene-based association, and genome-wide association studies (GWAS). In our study, we investigated the application of machine learning (ML) in bioinformatics, focusing on developing classification models to distinguish smokers from non-smokers based on genetic profiles. We used a dataset from the GEO database, which included genetic information and smoking status of African-American participants. The data underwent pre-processing, including cleaning and imputing missing values. Feature selection techniques were applied to identify the most relevant genetic variants for inclusion in our ML algorithms. We employed various ML methods, such as support vector machines (SVM) and logistic regression, to train classification models on the given training dataset. Specifically, our study focused on single nucleotide polymorphism (SNP)-based phenotype profiling of individuals of African origin. Tools such as Elastic Net (EN) were used for feature selection, and Logistic Regression (LR) was applied to classify smoking status based on SNP data. Additionally, we performed functional annotation to identify the genes associated with highly ranked SNPs and assess their relevance to nicotine dependence. Our results demonstrated robust classification performance, especially in the logistic regression and SVM models, which effectively discriminated smokers from nonsmokers. Identified SNPs have the potential to significantly impact smoking prevention efforts.
Institution: Institut Pasteur de Tunis
Language: English
Subfield: Physiology
Field: Medicine
Domain: Health Sciences
Open Access status: gold
OpenAlex ID: W4416502535