Customer Interview: Jamie D. McNicol, Analytical Project Manager
Q: Can you please tell us about your work on a high level?
Jamie: I am working as Data Scientist in the Human Immune Testing Suite (HITS) of the Immunology Department of McMaster University. We help our clients with immune monitoring for clinical trials. I am taking care of the overall
analysis pipeline for our running studies
Q: What is a longitudinal study for you?
Jamie: To me, a longitudinal study involves multiple time points and several patients observed over time to see the impact of a treatment. It has variables (e.g., it involves different time points or hundreds of samples) that prevents you from acquiring all the samples in one day. Often longitudinal studies are quite long and may be conducted across different institutions. This means that you cannot do the entire study with one set of antibodies or the same cytometer calibration. Detailed planning of data acquisition, management and analyzing is required.
Q: What kind of data for each patient do you need to store in this type of
study?
Jamie: We’re very blinded so we only store sample ID, treatment day and the collected FCS files. When we process blood for studies, there is a lot of paperwork involved to ensure the samples are processed similarly, and the resulting tubes are labelled accordingly prior to cryo-preservation. We have laboratory management software to hold the information. FCS files permit us to store a wealth of metadata, including antibody information, cytometer calibration settings, and other user-input data. I highly recommend taking advantage of these built-in ways of carrying data into downstream analyses.
Q: When planning a longitudinal study what are the most important points that you need to consider?
Jamie: I think it’s critical at the beginning of a study to set up a template for the analysis. So, each time an operator runs an experiment the template is used, and all required information needs to be filled in. The file
naming convention is critical. But obviously you also need to ensure that your antibody panel works: the cocktail is different in every project and over time there might be better reagents to be used in the study. [...] We have moved from manual
gating to an automated approach and now we see great results with having less bias involved. Furthermore, we do our quality control using clustering algorithms, so we don’t have to manually look at the data to make sure everything is fine.
Q: What do you find to be the biggest challenges in analyzing data from longitudinal
studies?
Jamie: In the end of a study, you need to be able to compare the data reliably. That is the most important thing for me. FCS files from different patients being acquired on different instruments or even on different days can
look very different and might not compare very well. An ideal raw data repository for a longitudinal study would be secure, continuously backed-up, searchable, and available to software analysis programs directly.
Q: What about file size?
Jamie: I think that the size of FCS files is not necessarily a limiting factor when it comes to analysis of longitudinal studies. Processing time is more of a concern, particularly when applying unbiased machine learning algorithms in your pipeline.
