Optimizing study designs for iPSC-based disease modelling
In a study published in Molecular Psychiatry, Jessie Brunner, Hanna Lammertse and Annemiek van Berkel (FGA) teamed up with Sophie van der Sluis (CTG) and the proteomics facility (MCN) to optimize statistical power for iPSC-based study designs.
Induced pluripotent stem cells (iPSCs) greatly facilitate the investigation of human disease mechanisms, the characterization of patient-specific cellular phenotypes, and the development of new, personalized treatments. For brain disorders, iPSC-based disease modelling is particularly advantageous as access to primary tissue is highly restricted. However, optimal study designs with sufficient statistical power were poorly defined.
To address this problem, Jessie Brunner, Hanna Lammertse and Annemiek van Berkel (FGA) teamed up with Sophie van der Sluis (CTG) and the proteomics facility at MCN. Together they described commonly used study designs, identified the specific research questions that these designs address, and established the most appropriate statistical analyses. They generated immunocytochemical, electrophysiological, and proteomic data from iPSC-derived neurons of five healthy subjects, analysed variation in these data, and used this information to set-up realistic power simulations. These simulations demonstrate that published case-control iPSC studies are generally underpowered.
To reach higher statistical power, isogenic designs, where mutations are generated or corrected within the same genetic background, can be used. However, these designs are limited in their generalizability. Instead, studying multiple isogenic pairs in parallel increases absolute power up to 60% or requires up to 5-fold fewer lines, while allowing generalization of the findings to the larger patient population.
In addition to the power calculations based on their own data, the authors aimed to provide a framework for researchers to design optimal, statistically rigorous iPSC-based studies. To this end, they generated a free web tool that researchers can use to a priori explore the power of different study designs, using any (pilot) data. With this tool, the authors hope to inspire researchers to design rigorous iPSC studies, that optimize the possibilities of this versatile technique while yielding robust and replicable results.