TITLE: Unleashing the Power of Machine Learning and Deep Learning to Accelerate Clinical Development
INSTRUCTOR: Li Wang, Yunzhao Xing, and Sheng Zhong, AbbVie
MODERATOR:Jingjing Ye
Abstract:
With the rapid advancement of machine learning (ML) and deep learning (DL) methodology in the last decade, the performances of prediction tasks in many computer science fields (e.g., imaging processing, natural language processing) have been greatly improved especially thanks to generative AI and transformers. There are many applications of ML and DL in drug discovery side. However, the impact of ML/DL in the field of clinical development has been relatively limited. Hence, we would like to propose a short course to educate, motivate and encourage the use of ML/DL especially in clinical development. Since we are statisticians by training, we would like to provide statistical perspectives on ML/DL as well.
The course starts with an overview of ML/DL methodology evolution over time and the related key concepts (e.g., back-propagation, hyperparameter tuning, train-test split, key performance metrics including AUC, precision, recall, F1 etc.). Then we will talk about the close relationship between ML and statistics. We will then introduce the latest developments in image processing and natural language processing, together with their novel applications in clinical development from our recent real projects and submitted papers to illustrate the vast potential of ML/DL in clinical development.
In terms of the course outline, the materials of the course are divided into three sections:
- General ML/DL methodology and relationship between ML and statistics
- Image processing and applications: deep convolutional neural networks (DCNN), object detection and segmentation, Region-based CNN (R-CNN), YOLO, and applications (e.g. psoriasis area and severity prediction)
- Natural language processing and applications: word embeddings (word2vec), recurrent neural networks and language models, self-attention and transformers, pre-train and fine-tune paradigm and applications (e.g., adverse drug event prediction)
Instructors’ Biography:
Li Wang, PhD, is currently Senior Director and Head of Statistical Innovation group in AbbVie. Li is leading Design Advisory which provides strategic and quantitative consulting as requested to all Development teams in all Therapeutic Areas to facilitate innovative thinking and complex innovative design evaluation. Li also leads Clinical Trial Innovation capability in AbbVie to drive Machine Learning and Advanced Analytics research and application in Development. Prior to this senior leadership role, he led Immunology and Solid Tumor statistical design and strategy discussions and multiple ML, RWE and Bayesian innovation projects from 2017 to 2019. From 2006 to 2017, he contributed to and subsequently led several NDAs and SNDAs including blockbusters Eliquis, Onglyza and Rinvoq. He is enthusiastic in teaching statistical courses to non-statisticians, and investigating/ promoting novel statistical and machine learning methodologies.
Yunzhao Xing is the Associate Director of Statistical Innovation at AbbVie, boasting a PhD in Material Science from the University of North Carolina at Chapel Hill and a background in Physics. Prior to AbbVie, he served as a senior scientist at Halliburton, focusing on sensor modeling and simulation. Since joining AbbVie in 2018, Yunzhao has led numerous successful projects in machine learning, deep learning, and image processing. His skill set encompasses web scraping, simulation modeling, and interactive web application development, making him a pivotal contributor to AbbVie’s Statistical Innovation Group. Yunzhao is recognized for his commitment to pushing the boundaries of statistical innovation.
Dr. Sheng Zhong is the Director of Statistics at AbbVie Inc. He received his Ph.D. in Statistics from the University of Chicago. At AbbVie, he led multiple innovative predictive modeling projects across different fields such as clinical trial enrollment duration forecasting, virtual controls based on targeted learning in single-arm trials, and predictive clinical safety monitoring based on structured and text data. His recent works have led to multiple publications and manuscripts under review. Before joining AbbVie in 2016, Dr. Zhong worked at a big data analytics start-up for heavy machine equipment maintenance, where his work led to 3 US patents.