Fig. 5

Tree-like Bayesian networks learned from transcriptional data of a malaria challenge experiment in Macaca mulatta. Networks were learned using blood informative transcripts [21] to focus on potential Axes of variation in the transcriptional data. a Using the ten blood informative transcripts as originally published, two branches emerge that best describe the root (selected automatically and which is from Axis 3), consisting of other genes from Axis 3 and a combination of multiple genes from Axes 2, 4, and 7. b Using the top 25 genes from each Axis to build a network based on the same root, the relationship between the Axes becomes even more evident, as both Axis 2 and Axis 7 contribute the dominant genes in parallel branches of the tree, suggesting significant but distinct mutual information with their parent and ultimately with the root. These relationships were not evident using standard multivariate and clustering analyses, and were not expected a priori based on previous descriptions of the axes of variation and the fact that the gene lists were derived from whole blood, not bone marrow aspirate, transcriptional profiling analyses