The stochastic route of haematopoiesis: modelling and inference methods in clonal tracking studies

Gene therapy is a novel clinical treatment for curing rare genetic diseases or moderating their effects. The success of this therapy depends on the ability of the modified stem cells to replace the faulty genes with functioning copies, called clones. This helps re-establishing the normal formation of blood cells, a process called haematopoiesis.

Mathematical models of clonal dynamics provide useful insights on haematopoiesis and can support the design of safer and effective gene therapy strategies. In this work we propose a novel data-driven stochastic framework to shed light on the dynamics of haematopoiesis after a gene therapy treatment. First, we use a stochastic state-space model to describe how the modified stem cells differentiate into specific cell types. Subsequently, we employ a mixed-effects stochastic formulation to investigate if the growth of the cellular offspring is mainly guided by a few clones, a potential adverse event known as clonal dominance. Finally, we show how shape-constrained splines can be used to remove the effects of technical artefacts when quantifying the level of clonal diversity.

Synthetic studies show that our methods outperform the state-of-the-art. The application of our framework on real clonal tracking datasets allowed to (i) infer the dynamics of cell differentiation in two preclinical studies and under three different genetic diseases, (ii) detect possible events of clonal dominance in two preclinical studies, and (iii) objectively compare clonal diversity under two different treatments in a preclinical safety study. Our proposed framework provides statistical support in gene therapy surveillance analyses.