Paper reading (四)：A primer on deep learning in genomics

论文题目：A primer on deep learning in genomics

scholar 引用：37

页数：7

发表时间：26 November 2018

发表刊物：Nature GeneTics

作者：James Zou 1,2,3*, Mikael Huss4,5, Abubakar Abid3, Pejman Mohammadi6,7, Ali Torkamani 6,7 and
Amalio Telenti 6,7*

摘要：Deep learning methods are a class of machine learning techniques capable of identifying highly complex patterns in large datasets. Here, we provide a perspective and primer on deep learning applications for genome analysis. We discuss successful applications in the fields of regulatory genomics, variant calling and pathogenicity scores. We include general guidance for how to effectively use deep learning methods as well as a practical guide to tools and resources. This primer is accompanied by an interactive online tutorial.

结论：

• best augment and complement human experience in making medical decisions
• how to avoid biases in training sets and how to interpret predictions
• there is a need for iterative experimentation

Introduction：无

正文组织架构：

1.Deep learning as a class of machine learning methods

• Deep learning in genomics: key elements and guidance

2.Setting up deep learning

• Vocabulary and concepts

3.How to use deep learning effectively

4.Interpreting deep learning models

5.Applications in genomics

6.Resources

• Deep learning resources
• Cloud platforms
• Software libraries

正文部分内容摘录：

• training, validation and test sets
• The key difference between deep learning and standard machine learning methods used in genomics
• the input and output of a neural network
• feed-forward, convolutional and recurrent
• Recurrent neural networks(RNN)
• unsupervised learning
• training a neural network, loss functions
• back-propagation, L2 regularization, dropout
• design training datases that are appropriately balanced
• genomics data are often highly imbalanced
• precision, recall
• understanding the assumption and limitation behind different architectures
• genomics applications, five layers are sufficient
• linear regression, least absolute shrinkage and selection operator(LASSO), gredient boosting or random forest
• the predictive model
• regulatory genomics
• DeepSEA