Deconvolution of Heterogeneous Tissue Samples into Relative Presence of Macrophage Phenotype Based on Gene Expression

Macrophages, as a primary cell of the innate immune system, have a variety of phenotypes that correspond to various functions. The dysregulation of the appearance of these phenotypes can lead to symptoms seen in many diseases. Specifically, macrophage phenotype has been implicated as a potential source of sustained inflammation that prevents healing in chronic wounds. In order to design effective treatments, an understanding of the relative presence of macrophage phenotypes in tissue is necessary. Inferring the relative phenotype composition is currently challenging due to the heterogeneous nature, not only of the macrophages themselves, but also of tissue samples. They contain many different cell types, which express many of the same genes. We present here a proposed method to deconvolute those heterogeneous tissue samples into the composition of two main macrophage phenotypes. Our final model uses gene expression from gene signatures for each phenotype as input to a predictive model that infers sample composition with an average error of 14.6%, and generates predictions that strongly correlate with known compositions (r=0.905). Finally, we apply this model to understand macrophage behavior in wound tissues, using publicly available datasets to obtain expression input. The model was able to demonstrate changes in macrophage phenotype composition in the wound over time.M.S., Biomedical Engineering -- Drexel University, 201