Domingo Salazar (AstraZeneca), Carsten Henneges (Syneos Health)

In this session we will introduce this new PSI Data Science SIG, its goals and current activities. Three talks will cover the following: An example of recent blog posts developed within the group to elicit engagement and discussions within the community and to help data science partitioners to expand their tool kits and data modalities. A presentation introducing current methods for feature selection and feature construction with their application to clinical data. A published project on predicting immunochemotherapy tolerability is used as the guiding example. A talk focused on an application of a work flow that combines traditional survival analysis techniques like KM curves and Cox models with state-of-the-art machine learning methods and explainers. As with all the activities of our SIG, our goal is to generate interest and discussion, not to advocate for particular methods, and to be introductory and practical.

Introduction to the PSI Data Science SIG: In this talk we will introduce this new PSI Data Science SIG, its goals and current activities. As an example, we will cover to recent blog posts developed within the group: one of the creation of effective dashboards and the other on the application of the particularities of the analysis of omics datasets. In general, the goal of our blogs is to elicit engagement and discussions within the community and to help data science partitioners to expand their tool kits and data modalities.

Feature Selection:
The presentation will introduce current methods for feature selection and feature construction with their application to clinical data. A published project on predicting immunochemotherapy tolerability is used as the guiding example to share learnings, and critically review and discuss approaches. Since feature selection is a wide-spread task, the aim is to work out and highlight the specialties and needs related to its application to clinical data.

Machine Learning in Survival Analysis: This talk will focus on an application of a particular tried-and-tested work flow that combines traditional survival analysis techniques like KM curves and Cox models with state-of-the-art machine learning methods and explainers. As with all the other activities of our SIG, our goal is to generate interest and discussion, not to advocate for particular methods. Also, in line with our philosophy, the talk would be introductory and practical.