Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding
Convolutional neural network (CNN) based methods have outperformed conventional machine learning methods in predicting the binding preference of DNA-protein binding. Although studies in the past have shown that more convolutional kernels help to achieve better performance, visualization of the model...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2019-11-01
|
| Series: | Frontiers in Genetics |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/article/10.3389/fgene.2019.01145/full |
| _version_ | 1819069950800691200 |
|---|---|
| author | Xiao Luo Weilai Chi Minghua Deng Minghua Deng |
| author_facet | Xiao Luo Weilai Chi Minghua Deng Minghua Deng |
| author_sort | Xiao Luo |
| collection | DOAJ |
| description | Convolutional neural network (CNN) based methods have outperformed conventional machine learning methods in predicting the binding preference of DNA-protein binding. Although studies in the past have shown that more convolutional kernels help to achieve better performance, visualization of the model can be obscured by the use of many kernels, resulting in overfitting and reduced interpretation because the number of motifs in true models is limited. Therefore, we aim to arrive at high performance, but with limited kernel numbers, in CNN-based models for motif inference. We herein present Deepprune, a novel deep learning framework, which prunes the weights in the dense layer and fine-tunes iteratively. These two steps enable the training of CNN-based models with limited kernel numbers, allowing easy interpretation of the learned model. We demonstrate that Deepprune significantly improves motif inference performance for the simulated datasets. Furthermore, we show that Deepprune outperforms the baseline with limited kernel numbers when inferring DNA-binding sites from ChIP-seq data. |
| first_indexed | 2024-12-21T16:58:11Z |
| format | Article |
| id | doaj.art-4ff95bb117cb4061a43c11803f22a67e |
| institution | Directory Open Access Journal |
| issn | 1664-8021 |
| language | English |
| last_indexed | 2024-12-21T16:58:11Z |
| publishDate | 2019-11-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Genetics |
| spelling | doaj.art-4ff95bb117cb4061a43c11803f22a67e2022-12-21T18:56:42ZengFrontiers Media S.A.Frontiers in Genetics1664-80212019-11-011010.3389/fgene.2019.01145491339Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein BindingXiao Luo0Weilai Chi1Minghua Deng2Minghua Deng3School of Mathematical Sciences, Peking University, Beijing, ChinaCenter for Quantitative Biology, Peking University, Beijing, ChinaSchool of Mathematical Sciences, Peking University, Beijing, ChinaCenter for Quantitative Biology, Peking University, Beijing, ChinaConvolutional neural network (CNN) based methods have outperformed conventional machine learning methods in predicting the binding preference of DNA-protein binding. Although studies in the past have shown that more convolutional kernels help to achieve better performance, visualization of the model can be obscured by the use of many kernels, resulting in overfitting and reduced interpretation because the number of motifs in true models is limited. Therefore, we aim to arrive at high performance, but with limited kernel numbers, in CNN-based models for motif inference. We herein present Deepprune, a novel deep learning framework, which prunes the weights in the dense layer and fine-tunes iteratively. These two steps enable the training of CNN-based models with limited kernel numbers, allowing easy interpretation of the learned model. We demonstrate that Deepprune significantly improves motif inference performance for the simulated datasets. Furthermore, we show that Deepprune outperforms the baseline with limited kernel numbers when inferring DNA-binding sites from ChIP-seq data.https://www.frontiersin.org/article/10.3389/fgene.2019.01145/fulldeep neural networksmotif inferencenetwork pruningconvolutional neural networksinterpretation |
| spellingShingle | Xiao Luo Weilai Chi Minghua Deng Minghua Deng Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding Frontiers in Genetics deep neural networks motif inference network pruning convolutional neural networks interpretation |
| title | Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding |
| title_full | Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding |
| title_fullStr | Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding |
| title_full_unstemmed | Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding |
| title_short | Deepprune: Learning Efficient and Interpretable Convolutional Networks Through Weight Pruning for Predicting DNA-Protein Binding |
| title_sort | deepprune learning efficient and interpretable convolutional networks through weight pruning for predicting dna protein binding |
| topic | deep neural networks motif inference network pruning convolutional neural networks interpretation |
| url | https://www.frontiersin.org/article/10.3389/fgene.2019.01145/full |
| work_keys_str_mv | AT xiaoluo deepprunelearningefficientandinterpretableconvolutionalnetworksthroughweightpruningforpredictingdnaproteinbinding AT weilaichi deepprunelearningefficientandinterpretableconvolutionalnetworksthroughweightpruningforpredictingdnaproteinbinding AT minghuadeng deepprunelearningefficientandinterpretableconvolutionalnetworksthroughweightpruningforpredictingdnaproteinbinding AT minghuadeng deepprunelearningefficientandinterpretableconvolutionalnetworksthroughweightpruningforpredictingdnaproteinbinding |