The use and accessibility of Bayesian Inference in the Life Sciences

This dissertation explores the application and accessibility of Bayesian statistical methods for analyzing complex biological data. Conventional methodologies often rely on frequentist approaches such as t-tests, ANOVA, etc., but these methods can be inappropriate, particularly when dealing with complexities such as hierarchical structures or small sample sizes. In contrast, Bayesian statistics offer greater flexibility, as they allow for more adaptable models and the incorporation of prior knowledge. This is particularly relevant in the life sciences, where more accurate statistical approaches can lead to a reduction in the required sample size -- an outcome that is especially important in the context of animal research. The primary aim of this work is to highlight the application and benefits of Bayesian statistics in biomedical research. Another key objective is to reduce the barriers that prevent biologists from applying Bayesian methods themselves. To this end, the dissertation includes four publications, preceded by a brief introduction to a selected biological research field -- cell differentiation during bone formation -- and an overview of statistical methods, with a focus on Bayesian approaches. The first contributed paper underscores the conceptual benefits of Bayesian modeling and compares the results with a commonly used frequentist method. The comparison is based on real data from a well-known animal behavioral experiment. This paper emphasizes the use of more accurate statistical methods to obtain the same amount of information from fewer animals, as compared to conventional approaches. The second paper investigates the regulatory role of the transcription factors Gli3 and Trps1 in controlling Wnt5a expression during the early stages of chondrocyte hypertrophy. Experimental data and Bayesian modeling were used to quantify gene expression dynamics and to infer regulatory interactions. The third paper examines the effect of chromatin states on gene expression and their role in the differentiation of chondrogenic cells. A Bayesian model is used to capture the complexity and hierarchical structure of chromatin state dynamics associated with stages of differentiation, helping to clarify the role of specific histone modifications. The final paper introduces a new web-based tool designed to make Bayesian analysis more accessible to life scientists. It focuses on planning experiments based on sample size determination, data analysis, and generating reports to ensure transparent and reproducible results. By emphasizing explicit model building, Bayesian methods can yield more accurate results -- particularly with small sample sizes -- as demonstrated by comparing a well-structured model to one that relies on the statistical assumptions commonly implied by frequentist methods. Applying Bayesian models to additional biological data has contributed to a deeper understanding of cell differentiation in chondrocytes and the role of specific histone modifications in chromatin states. Reducing the barriers associated with less widely known statistical methods is important for enabling more accurate analyses -- and for reducing the use of animals in research. The current version of BAYAS can help overcome this barrier, but further development is needed to support additional use cases and to improve usability.