The ability to verify the results of data that has been collected or generated during the course of any drug trial or simulation is paramount to the validity of the study. Often, as biostatistics and pharmaceutical teams become even more reliant on digital tools to record and draw insights from these studies, creating a “digital footprint” of the steps and materials that went into each one becomes more and more important. This is where JuliaHub comes in. The JuliaHub platform was specifically designed for scientific research teams to get access to high performance computing power and provide pharmacology teams a verifiable “single source of truth” for all of the digital activities. That is why our reproducibility feature, called “Time Capsule” has become a cornerstone to how pharmaceutical development research teams reproduce and verify work with their compliance departments.
Time Capsule has two primary parts. The first allows teams to re-run any job now or in the future. The second piece lets team admins save groups of these jobs as “capsules” to organize them by date or by study. The main things that time feature allows pharmacometrics teams to do is:
Can you show me how Time Capsule makes batch jobs reproducible?
As long as you have the Time Capsule feature active on your organization’s subscription of JuliaHub - the process is extremely easy. Anytime you have a batch job that you have already run in JuliaHub, those jobs are saved automatically and any of those batch jobs in JuliaHub can be re-run. JuliaHub saves all the inputs, outputs, and environment variables for each batch job so you can run those jobs again now or anytime in the future. You can see how that works in this video:
What are the challenges of reproducibility in pharmaceutical research?
Reproducibility in pharmaceutical research provides teams a systematic approach to duplicating the results of a study. When reproducibility is needed, the entire software lifecycle needs to be available in an explicit way. All the inputs, outputs, environmental variables, configuration options, job metadata, and package versions need to remain the same or trying to reproduce the same result may not be possible - especially if time has passed and package versions or application configuration options have changed. This kind of reproducibility ensures safety and efficacy allowing the relevant compliance bodies to make sure research is sound before drug trials continue and drugs reach the public.
Normally, this reproducibility has been difficult because of the changes that happen rapidly in your software environment, but with Time Capsule, we’ve saved you all the headache of keeping the environment stable. We can reproduce the results of the experiment even years later.
What is a batch job in JuliaHub?
In JuliaHub, what we refer to as a batch job is generally a job that is meant to execute a script rather than to launch an interactive application. For example, if you need to compile some kind of output or code by using an input of a combination of extracting data and combining multiple large files using an algorithm - the job that would run that algorithm would be called a “batch job”. What makes JuliaHub perfect for these kinds of runs is that Julia as a language is extremely fast at computational tasks as well as the fact that JuliaHub provides the ability to scale up your computational power seamlessly from a 2vCPU instance to a 32vCPU instance. JuliaHub even provides access to GPUs such as the nVidia v100 chip.
What information does it save specifically?
JuliaHub saves the following information for each batch job and allows time capsule to reproduce your jobs using this information.
In fact, for those of you looking to use this information in a wider Software Bill of Materials (SBOM) report, our team can provide this information to you in a digestible format - including and in addition to each package’s SBOM info via the Julia package manifest.toml.
What is the “archiving and group use case”:
The newest addition to Time Capsule is UI that allows admins to organize batch jobs into archived groups. You can save a number of batch jobs and then add a title and description to those groups of jobs. This comes in handy in case you want to label a certain group of jobs by date or by study. That looks like this:
Conclusion
Features that provide reproducibility are very important for quantitative research teams that need to be able to produce regulatory compliant data and analysis for drug trials. JuliaHub was built to support regulatory frameworks and includes a number of trust and security features. Furthermore, JuliaHub is continuing to advance in bringing end-to-end traceability around every workstream that scientists and developers do on the platform. Combined with the high performance computing environment available and the speed of the Julia language, the platform is perfect for executing pharmacologic analysis quickly and with an audit trail and version history to provide quality control. Whether you are using JuliaHub as a single user or needing to share your data with multiple teams during multiple studies across, the platform was built to provide Pharmacometrics a single source of truth.
In fact, you can see how companies like Pfizer and Astra Zeneca have already used Julia in their preclinical drug trial research.
Learn more about our pharmaceutical use cases here: https://juliahub.com/industries/pharma and watch our Youtube channel for more information.