Machine learning based bioinformatics analysis of intron usage alterations and metabolic regulation in adipose browning

Abstract

Adipose tissue is the major energy depot of the body and is considered an endocrine organ. Adipose tissue involves many different cell types, first and foremost, the adipocytes. White adipose cells that store fat and brown adipocytes that take part in lipid oxidation and heat generation are the most common cell types in adipose tissue. Even though brown adipocytes which have a high number of mitochondria and high fat-burning capacity are rare in adults, they are abundant in newborns and rodents. White adipocytes can gain a temporal brown-like character with a process called browning, which can be induced with cold exposure providing white adipocytes with increased fat-burning capacity. Adipose tissue is the main tissue associated with obesity; therefore, the browning process has the potential to be used in the treatment of obesity. Here, we made use of machine learning techniques to better understand the browning mechanisms. We applied a computational approach based on generalized linear models (GLM) and decision trees for the identification and classification of alternative splicing events, followed by downstream bioinformatics analysis for the detection of differential regulatory events in the transcriptome of the adipocyte browning. Our analyses identified possible extracellular alterations in response to changes in cellular shape via alternative splicing events and remarkably an intron retention event on the Upstream stimulatory factor2 (Usf2) gene that may alter the activity of the regulator and take part in the regulation of the browning process. Targeted therapies for induction of the browning process via regulation of Usf2 may prevent and treat obesity which is a widespread health condition. To the best of our knowledge, this is the first study that combines alternative splicing events with regulatory network inference to reveal the mechanism of the browning process. Our methodology has the potential to reveal many other disease-related mechanisms and lead to novel therapy strategies.