Submit Manuscript  

Article Details


Exploiting XGBoost For Predicting Enhancer-Promoter Interactions.

Author(s):

Xiaojuan Yu, Jianguo Zhou, Mingming Zhao , Chao Yi, Qing Duan, Wei Zhou* and Jin Li  

Abstract:


It is well-known that gene expression and disease control are co-regulated by the interaction between the distal enhancer and the proximal promoter, and the study of enhancer promoter interactions (EPIs) can help us to gain insight into the genetic basis of diseases. Although the recent emergence of some high-throughput sequencing methods have given us a deeper understanding of EPIs, accurate prediction of EPIs still have some limitations. In this paper, we trained a XGBoost based model and introduced two sets of features (i.e. epigenomic and sequence feature) to predict the interactions between the enhancer and the promoter in different cell lines. We compared XGBoost with the other four methods. Extensive experimental results have shown that XGBoost based method is effective in predicting EPIs across three cell lines. Especially epigenomic and sequence features can boost prediction.

Keywords:

Enhancer-promoter interactions, Supervised learning, Machine learning, Gene expression, Feature extraction, XGBoost

Affiliation:

Software School of Yunnan University, Kunming, Software School of Yunnan University, Kunming, Software School of Yunnan University, Kunming, Software School of Yunnan University, Kunming, Software School of Yunnan University, Kunming, Software School of Yunnan University, Kunming, Software School of Yunnan University, Kunming



Full Text Inquiry