Drug Evaluation Committee Machine learning has begun.
Data Science Subcommittee
May 2020
This report introduces planning, data processing, and execution of machine learning using actual data, assuming a situation where data scientists in a broad sense involved in pharmaceutical companies use machine learning and AI based on it for data analysis and development of services using familiar data including clinical data. In addition, methods to increase data to compensate for the lack of quantity and quality of training data, which is a major obstacle in the development of AI, and methods to use pre-trained models for deep learning will be introduced. Deep learning is not covered in this book due to the difficulty of securing data from the beginning.
Chapter 1, "Projects Involving Machine Learning," provides an overview of how using machine learning differs from normal system development. Chapter 2: Acquisition and Processing of Development Data (Annotation)" and "Chapter 3: Data Division for Learning and Learning Methods" explain how to acquire and increase data, how to process data, and how to divide data for machine learning, including the methods used in Chapters 4 and 5. In "Chapter 4, Implementation Examples of Using Machine Learning for Clinical Trial Data and RWD," the author introduces in detail the process of actually performing machine learning using data such as blood glucose levels of diabetic patients who use insulin, and using the data for services. Chapter 5, "Explanation of Using Learned Models and Introduction of Practical Examples," introduces the concept and practice of transfer learning and fine tuning, in which learned models are modified and reused, using examples of images and natural language. Although none of the methods or examples in this book are state-of-the-art, there are many open-source libraries and other useful information on the Web. Interested readers should start with something familiar and simple (e.g., performing logistic regression analysis with a single-layer neural network). We hope that this book will encourage users who intend to use machine learning methods in pharmaceutical companies, especially from clinical development to post-marketing situations.
Data Science Subcommittee, Committee on Drug Evaluation, Japan Pharmaceutical Manufacturers Association
Task Force 1 Machine Learning Team
We have started machine learning. (6.5MB)
Deliverables of the Drug Evaluation Committee Return to list of all deliverables