Paper reading (五)：progress in machine intelligence for rational drug discovery

论文题目：From machine learning to deep learning: progress in machine intelligence for rational drug discovery

scholar 引用：87

页数：6

发表时间：2017.12

发表刊物：Drug Discovery Today

作者：Lu Zhang, Jianjun Tan, Dan Han and Hao Zhu

摘要：quantitative structure activity relationship(QSAR) modeling

We summarized the history of machine learning and provide insight into recently developed deep learning approaches and their applications in rational drug discovery. 

We suggest that this evolution of machine intellligence now provides a guide for early-stage drug design and discovery in the current big data era.

结论：combi-QSAR; hybird QSAR

deep learning methods and their related studies have some limitations,

1. the avaliability of a large amount of high-quality data will affect the performance and reliability of deep learning modeling. The massive amounts of biomedical data generated by pharmaceutical companies  are normally not avaliable to the public but are kept as expensive private commercial assets.
2. a lack of rational interpretations of associated biological mechanisms

Introduction：

• The QSAR modeling procedure has been standardized across rational drug discovery processes.
• deep learning approaches that explain the vanishing effects of gradients are more suitable to use with raw high-dimensional data.

正文组织架构：

1. QSAR approaches

2.Combinatorial QSAR and hybird QSAR

3.Advances in rational drug discovery resulting from deep learning

• New drug molecule identification
• protein engineering
• Gene expression data analysis
• Pharmacodynamics modeling

正文部分内容摘录：

• Linear discriminant analysis (LDA)
• Support vector machines (SVMs)
• Decision trees (DTs)
• Random forest （RF）
• The k nearest neighbor (kNN)
• Artificial neural networks (ANN)
• Althogh QSAR approaches have been developed for decades, common issues remain that have not been solved by using any existing approaches.
• recent modeling studies in drug discovery have focused on predictions based on a combination of various types of model.
• another common issues related to QSAR modeling is the existence of 'active cliffs'.