Speaker: Meir_Pinco, BeiGene

Dr. Meir Pinco brings a wealth of experience as an R&D informatics professional, backed by a Ph.D. in Neuroscience from The Hebrew University and further education in clinical research and epidemiology from the University of California, San Francisco. With over 25 years dedicated to the field, Meir has contributed significantly to solving complex scientific and life-science challenges, particularly in software development, data management, and data analytics. Currently serving as the Associate Director of Data Science and Digital Innovations at BeiGene, Meir plays a key role in enhancing the company's Oncology portfolio through innovative data analytics and AI solutions. His work is built on a foundation of roles at notable organizations like Gilead Sciences Inc. and Merck Research Laboratories, where he focused on improving drug safety information systems and research informatics. Meir's expertise is broad, encompassing data analytics, software development, AI/ML, and drug development, yet it is his collaborative approach and commitment to shared goals that stand out. His combination of scientific understanding and technical skill enables him to contribute meaningfully to the biotechnology and pharmaceutical informatics field. Meir is valued for his strategic thinking, ability to foster teamwork, and dedication to developing high-performing informatics and analytic solutions.

Title: Evaluation of AI/ML-based medical devices in FDA/CDRH – a statistician’s role and perspective

Speaker: Qin Li, CDRH/FDA

Dr. Qin Li joined CDRH/FDA in 2009 after she graduated from the University of Florida. She has intensive experience in statistical review and regulation of various diagnostic medical devices, and is now the team leader in the Division of Biostatistics under the Office of Clinical Evidence and Analysis in the Center for Devices and Radiological Health (CDRH) at FDA.

Abstract: Artificial intelligence enables extraordinary advances in medical products and has increasing influence in medicine and healthcare. There was a rapid growth in AI-based medical device development and regulatory submissions in FDA/CDRH, especially in the past 1 to 2 decades. From the most recent update, FDA has authorized 882 AI-enabled medical devices by May 13, 2024. These submissions cover a wide range of medical areas and diverse intended uses of the medical devices. Despite the excitement of this innovative technology, AI algorithms have been criticized for their potential biases and lack of transparency. And given its complexity and the nature of its data-driven training, it creates challenges in regulatory evaluation of such medical devices. This presentation will introduce the role of statisticians in evaluating AI/ML-enabled medical device in CDRH and share some considerations in their evaluation.

Title: Angle-Based Sparse Learning for Multicategory Classification

Speaker: Wenjie Wang, Lilly

Dr Wenjie Wang is currently a senior research scientist working at Eli Lilly and Company. His recent works involve innovative diabetes research and drug discovery by statistical learning. He earned his Ph.D. degree in statistics from the University of Connecticut in the Summer of 2019. His dissertation was on integrative survival analysis with application to suicide risk. He also holds an outstanding graduate award from Tongji University, Shanghai, China. An enthusiast of open-source software, he has been developing and maintaining several R packages for some research projects.

Abstract: Multicategory classification is an essential topic in statistical learning with broad applications in many areas, including biology and medicine. A commonly used method is to convert the multicategory classification into sequential binary classification problems using a one-versus-rest or one-versus-one strategy, which can be inefficient and suboptimal. To overcome the drawback of the sequential procedure, we propose regularized angle-based classifiers with groupwise penalties and show it have superior prediction performance and computational efficiency.

Speaker: Nicholas Michaud, Vertex

Dr. Nicholas Michaud is a Principal Data Scientist at Vertex Pharmaceuticals. He works on the Data Strategies and Solutions team and contributes to projects that span many areas within the organization, with a focus on using data and statistical techniques to accelerate clinical trial recruitment. He joined Vertex in 2018, and before that was a Postdoctoral Scholar at the University of California, Berkeley, where he contributed to the NIMBLE R package for flexible Bayesian modeling. He received his PhD in Statistics from Iowa State University in 2016.

Title: A Poisson-multinomial model to improve estimation of AAV genome integrity using duplex digital PCR

Speaker: Jake Gagnon, Biogen

Jacob Gagnon is an associate director of biostatistics at Biogen. He leads statistical efforts in the latest omics pipelines (ie spatial transcriptomics), performs preclinical neurological statistics research, is a core member of the text mining center of excellence, and leads a ML/DL focus group. His research interests include deep learning, machine learning, translational biology, omics analysis, and text mining. He obtained a PhD in statistics from UMASS Amherst and did postdoctoral work in biostatistics at WPI.

Abstract: Recombinant adeno-associated virus (rAAV) vectors have become a reliable strategy for delivering gene therapies. As rAAV capsid content is known to be heterogeneous, methods for rAAV characterization are critical for assessing the efficacy and safety of drug products. Multiplex digital PCR (dPCR) has emerged as a popular molecular approach for characterizing capsid content due to its high level of throughput, accuracy, and replicability. Despite growing popularity, tools to accurately analyze multiplexed data are scarce. Here, we introduce a novel statistical model to estimate genome integrity from duplex dPCR assays. This work demonstrates that use of a Poisson-multinomial mixture distribution significantly improves the accuracy and quantifiable range of duplex dPCR assays over currently available models.
Authors: Lauren Tereshko , Xiaohui Zhao , Jake Gagnon, Tinchi Lin, Trevor Ewald, Yu Wang, Marina Feschenko, Cullen Mason
Presenter: Jake Gagnon