Event

PSI Webinar: Open Source Software - is it really a free-for-all?

Add to:

Sponsored by logo







Date:
 Wednesday 17th April 2024
Time: 14:00-15:30 BST | 15:00-16:30 CEST
Location: Online via Zoom
Speakers: Matthew Lyon (AstraZeneca), Ari Siggaard Knoph (Novo Nordisk) and Daniel Sabanes-Bove (Roche). 

Who is this event intended for? Statisticians and programmers who are working or thinking of working in software beyond SAS.
What is the benefit of attending? 
Learning from the experiences of teams working with R and software beyond SAS in the pharma industry.

Cost

This webinar is free to both Members of PSI and Non-Members.

Registration

To register for this event, please click here

Overview

Talks from the speakers will cover the use of R in a programming community, submitted to regulators using R, and also programming beyond R in C++ and Julia.

Speaker details

 

Speaker

Biography

Abstract

matthewedit
Matthew Lyon

Matt Lyon studied his Zoology BSc at the University of Liverpool. His degree included quantitative biology modules which used several statistical data packages. He started his career at The Francis Crick Institute in London as an Animal Technician before progressing to CRUK Manchester and finally moved to AstraZeneca in 2020 as an In-Vivo Scientist.

Matt is currently the Global Head and his departments’ representative for Inclusion and Diversity (I&D).He also heads up a small, international cross functional team which focusses on creating and rolling out initiatives across the department. This also includes liaising with other areas of the business to promote AZ as a great place to work.

Matt has taken over the Lead of the Steering Committee of the community of R users at AstraZeneca- ‘R@AZ’ – which currently has around 1600 members. Building on his I&D and quantitative biology skills, he is looking at expanding this community within AZ and beyond.

Building a BiggeR Community of R Users at AstraZeneca

In the past years, the pharma industry has seen a true paradigm shift in its use of R. Up until recently, one had to choose between R and SAS. Today, most collaborators with a quantitative background are trained in at least 2 programming languages. With this in mind, at AstraZeneca we built on the growing interest for R, at any stage of the drug development but also company-wide.

Since April 2021 we have launched a variety of initiatives, initially starting with a modified public initiative #azTidyTuesday and Function of the month. As the community grew, we launched our first AZ Conference, #AZRC2022 and this is now run on an annual basis. Throughout 2022, we have added a variety of initiatives to bring people together including: Lunch & LeaRn, AZ R Hot Desk and in 2023 we formed AZ R-Ladies.

In early 2022, the R@AZ Leads were asked 3 questions:
1. What is the value of your community?
2. What is a Key Performance Indicator (KPI) of your community?
3. What is the Return On Investment (ROI) of your community?

Prior to 2023, there were no definitive answers but whilst building this community internally, we have reached out and collaborated with external partners across the data science industry. In response, they have kindly provided us with educational and informative speakers for internal community events and continue to do so as we plan for 2024.

The combination of all these initiatives, derived from the growth and passion of the steering committee, has led to an 800% increase in the internal social media community members; actively representing over 200 different departments with AstraZeneca. This is now highlighting the value, its KPI and ROI over 3 years in the making by building a network and a community.

Ariedit
Ari Siggaard Knoph

Ari is a Principal Statistical Programmer and International Lead Programmer at Novo Nordisk with a seat on the pharmaverse council. He is a driving force behind the adoption of R in clinical deliverables and submissions at Novo Nordisk. Ari is also the author and maintainer of multiple R packages used in the Novo Nordisk ecosystem.

Completing a submission and beyond in R

In this presentation we will pull out some touch points of the Q&A phase of our first R-based submission. Examples of FDA interactions will be discussed and our thoughts on how to continuously manage a possibly drifting R environment in a submission will be shared.

danieledit
Daniel Sabanes-Bove

Daniel Sabanes Bove studied Statistics in LMU Munich and obtained his PhD at the University of Zurich for his research work on Bayesian model selection. He started his career in 2013 at Roche as a biostatistician, then worked at Google as a data scientist from 2018 to 2020 before rejoining Roche. He currently leads the Statistical Engineering team in Roche Pharma Product Development that works on productionizing R packages, Shiny modules and how-to templates for data scientists. Daniel is co-author of multiple R packages published on CRAN and Bioconductor, as well as the book "Likelihood and Bayesian Inference: With Applications in Biology and Medicine", and is currently openstatsware.org (ASA BIOP working group on Software Engineering).

 

R(omeo) and Julia - A Love Story by openstatsware

For implementing statistical methods in software, we recently started trying out the Julia language. "Julia is a high-level, general-purpose dynamic programming language, most commonly used for numerical analysis and computational science" (Wikipedia) and as such well suited for statistical applications. I will introduce two successful Julia projects. The first project implements joint models for time-to-event and longitudinal outcomes (see e.g. Kerioui et al. 2022), and is available open source as JointModels.jl. It uses Turing.jl for MCMC based Bayesian inference, based on a new distribution class for time-to-event data specified via hazard functions. The second project implements Bayesian safety signal detection as described by Brock et al. (2022), under construction and open source as SafetySignalDetection.jl, again using Turing.jl and with a suitable extension of the expectation-maximization algorithm for fitting Beta mixtures. I will discuss the reasons why these projects were successful, and describe how we could easily embed the Julia algorithms into an R based overall workflow. Finally, I will introduce openstatsware.org where a growing community of statistical software engineers comes together to build software packages and develop and share best practices for such.

 


Upcoming Events