Enhancing the Open Source SciML Stack for Clinical Trial Simulations

Given the highly-publicised rapid adoption of SciML tooling across the major pharmaceutical companies, including both Pfizer and Moderna, questions of model verification and validation have ballooned as clinical trials accelerated during the pandemic. It became clear that without automated tooling, the cost of clinical analysis expands as researchers must painstakingly identify non-identifiability in a case-by-case manner. Automating parameter identifiability studies for the SciML-based pharmacometrics groups would have a massive impact on industry efficiency, along with promoting and accelerating good open science in other fields. This proposal focuses on a multi-pronged approach: the development of tooling for determining identifiability, accelerating the underlying numerical tools for stochastic simulation required to make automated identifiability analysis practical, and training and awareness campaigns through new community engagement projects and a convention.