A Pretraining-Retraining Strategy of Deep Learning Improves Cell-Specific Enhancer Predictions
A Pretraining-Retraining Strategy of Deep Learning Improves Cell-Specific Enhancer Predictions
复制标题
DOI：
10.3389/fgene.2019.01305
复制
发表时间：
2020
影响因子：
3.7
通讯作者：
Xuehai Hu
中科院分区：
文献类型：
--
作者：
Xiaohui Niu;Kun Yang;Ge Zhang;Zhiquan Yang;Xuehai Hu
Deciphering the code of cis-regulatory element (CRE) is one of the core issues of today ’ s.biology. Enhancers are distal CREs and play signi fi cant roles in gene transcriptional.regulation. Although identi fi cations of enhancer locations across the whole genome.[discriminative enhancer predictions (DEP)] is necessary, it is more important to predict.in which speci fi c cell or tissue types, they will be activated and functional [tissue-speci fi c.enhancer predictions (TSEP)]. Although existing deep learning models achieved great.successes in DEP, they cannot be directly employed in TSEP because a speci fi c cell or.tissue type only has a limited number of available enhancer samples for training. Here, we.fi rst adopted a reported deep learning architecture and then developed a novel training.strategy named “ pretraining-retraining strategy ” (PRS) for TSEP by decomposing the.whole training process into two successive stages: a pretraining stage is designed to train.with the whole enhancer data for performing DEP, and a retraining strategy is then.designed to train with tissue-speci fi c enhancer samples based on the trained pretraining.model for making TSEP. As a result, PRS is found to be valid for DEP with an AUC of 0.922.and a GM (geometric mean) of 0.696, when testing on a larger-scale FANTOM5 enhancer.dataset via a fi ve-fold cross-validation. Interestingly, based on the trained pretraining.model, a new fi nding is that only additional twenty epochs are needed to complete the.retraining process on testing 23 speci fi c tissues or cell lines. For TSEP tasks, PRS.achieved a mean GM of 0.806 which is signi fi cantly higher than 0.528 of gkm-SVM, an.existing mainstream method for CRE predictions. Notably, PRS is further proven superior.to other two state-of-the-art methods: DEEP and BiRen. In summary, PRS has employed.useful ideas from the domain of transfer learning and is a reliable method for TSEPs.
