iPro-TCN: Prediction of DNA Promoters Recognition and Their Strength Using Temporal Convolutional Network

Abstract

Promoters are an important regulatory element in the genome that control gene expression, and their abnormalities have been linked to various diseases. Therefore, accurately promoter identification is essential for biological research as we as drug development. But the identification of the promoter using laboratory approaches is highly costly. In order to address this issue, we proposed a computational model called iPro-TCN to predict promoter and their strength using temporal convolutional network (TCN) with a word2vec feature representation. This model includes a feature descriptor known as Word2Vec and achieved high performance to predict promoters, including strong and weak promoters. The iPro-TCN model obtained accuracy of 91.86% to predict promoter in the first layer for, and an accuracy of 84.63% to predict strong and week promoter in the second layer using cross validation test. On benchmark datasets, the proposed iPro-TCN model produced better performance than previous computational models in term of all performance metrics.

Keywords

• Deep learning
• promoters
• temporal convolution network
• natural language processing
• DNA
• word2vec