Data-Driven Strategies for Optimization of Human Megakaryocyte Differentiation

WU, JIA JIE

doi:https://doi.org/10.21985/n2-4qnw-ey13

Work

Data-Driven Strategies for Optimization of Human Megakaryocyte Differentiation

Public

Download PDF

Download All Files (.zip)

Platelets are circulating anucleate discs derived from megakaryocytes, and play major roles in hemostasis, inflammation, thrombosis, and vascular biology. Multi-phase culture systems for inducing in vitro platelet production from mature megakaryocytes have been explored to allow progenitor expansion, megakaryocyte maturation, and promotion of platelet formation and shedding. In this thesis, I describe the development of several methods for identifying influential factors for multi-phase megakaryocyte differentiation. These methods combine both computational and experimental techniques, and build upon existing approaches. After initial experiments in cell-lines, I constructed a method to develop time-course networks for early, middle, and late megakaryopoiesis from transcription factor array data. Validation with prior knowledge and experimental approaches revealed several false positives and false negatives, which led to the development of a windowed Granger causal inference strategy for network discovery. To identify influential culture factors for megakaryocyte differentiation, we screened several strategies and small molecules for improved ex vivo production. I adapted and applied one of the machine learning frameworks embedded in SWING to characterize donor heterogeneity within individual megakaryocyte culture conditions to improve production and build a predictive framework. Finally, I demonstrate a platform for generation of megakaryocytes from valproic acid expanded cells, as well as a computational method to predict culture performance based on observed donor heterogeneity, which provides potential for identification and intervention in in vitro megakaryocyte production processes.

Creator