Pseudotemporal ordering of breast scRNA-seq data

SENRA, DANIELA; GUISONI, NARA; DIAMBRA, LUIS

Congreso; XI Argentine Congress of Bioinformatics and Computational Biology (XI CAB2C); 2021

The human breast is an organ composed mainly of glandular, adipose and connective tissue. The basic structure of the mammary gland consists of lobular units that produce breast milk interconnected by an intricate system of ducts. The vast majority of breast tumors arise from the epithelial cells lining the terminal ductal lobular units. For this reason, the characterization of the healthy mammary epithelium is an important aspect to comprehend the origins of breast cancer. Therefore, it is of particular interest to understand the differentiation pathway of the mammary epithelium. As single cell RNA sequencing use has widely increased, many trajectory inference techniques that use this data type have been developed in recent years. Most of the trajectory inference algorithms require the selection of a start cell to infer the path of differentiation, which is usually set by previously known stemness markers and may not be adequate due to dropouts in scRNA-seq. We propose a Protein-Protein Interaction Network (PPIN) approach to identify the breast stem cells for later trajectory inference. We implement the method to calculate the activity of the PPIN for each cell in R. As the goal is to find the stem cells, the activity of the protein network associated to cellular differentiation process is a parameter that indicates the differentiation activity. In this way, the cells with the highest differentiation activity are determined and selected as the root for the trajectory.