Machine Learning Classification of Cancer Cells Migration in 3D Multi-cue Microenvironments1

摘要

Metastasis is the leading cause of deaths among cancer patients. Metastatic dissemination, during which cancer cells from the primary tumor reach the secondary organ, consists of a cascade of events, starting with cancer cell invasion and migration through the surrounding tissue [1]-[3]. During cancer cells invasion, tumor cell movement is directed by multiple guiding cues present in the tissue at different concentrations [1, 4]. Guiding cues can be biophysical, such as aligned collagen fibers inducing contact guidance or biochemical, such as gradients of chemoattractants inducing chemotaxis. Following individual cues will lead to directed cell migration, where the cell velocity and cell persistence will be determined by the concentration of the cue and the presence of the cell receptors that are responsible for communication with the particular cue. However, in complex microenvironments, where multiple cues are present at different levels, distances and orientations, cues can compete with or synergize each other to guide the cell migration. In such conditions, analyzing the cell migration parameters (velocity, persistence, directionality etc.) is a challenging task.