Network and machine learning analyses approach towards understanding the effects of bystander infection on both myeloid-related normal and clonal hematopoietic systems

Normal hematopoiesis (NH) refers to the physiological process that drives blood cell formation, whereas, clonal hematopoiesis (CH) refers to mutated clones conferring a selective growth advantage over normal cells. Additionally, the hematopoietic system gives rise to varying mature blood cells found in both NH and CH with the hematopoietic stem cells serving as the main source throughout the differentiation hierarchy. Bystander infection on the host has been found to influence hematopoiesis, however, the interrelationship between bystander infection and hematopoietic competence is not fully elucidated. In this study, we performed network design of myeloid-related NH and CH systems with and without bystander infection and, subsequently, we executed both centrality measurement and machine learning (ML) analyses (supervised and unsupervised). Additionally, robustness and sensitivity analyses were performed to evaluate the stability of the network-derived findings under perturbation and to reduce the likelihood that the conclusions were artefacts of network topology. From our designed myeloid-related NH and CH models without bystander infection, we identified the significant components and pathways within the network models using both centrality measurements and ML analyses. Interestingly, when we added a bystander infection perturbation to our myeloid-related NH and CH models, we noticed a shift among the highlighted components and pathways. In particular, we found that bystander infection may influence both the myeloid-related NH and CH systems by shifting toward myeloid-related components and pathways within the hematopoietic system, consistent with immune activation while also potentially influencing blood cell maturation. We emphasize that the results of both network and machine learning analyses only serve as exploratory and hypothesis-generating data and are not intended to serve as predictive nor mechanistic evidence.