Machine learning extracts oncogenic‐specific γ‐H2AX foci formation pattern upon genotoxic stress

AbstractH2AX is a histone H2A variant that becomes phosphorylated upon genotoxic stress. The phosphorylated H2AX (γ‐H2AX) plays an antioncogenic role in the DNA damage response and its foci patterns are highly variable, in terms of intensities and sizes. However, whether characteristic γ‐H2AX foci patterns are associated with oncogenesis (oncogenic‐specific γ‐H2AX foci patterns) remains unknown. We previously reported that a defect in the acetyltransferase activity of TIP60 promotes cancer cell growth in human cell lines. In this study, we compared γ‐H2AX foci patterns between TIP60 wild‐type cells and TIP60 HAT mutant cells by using machine learning. When focused solely on the intensity and size of γ‐H2AX foci, we extracted the TIP60 HAT mutant‐like oncogenic‐specific γ‐H2AX foci pattern among all datasets of γ‐H2AX foci patterns. Furthermore, by using the dimensionality reduction method UMAP, we also observed TIP60 HAT mutant‐like oncogenic‐specific γ‐H2AX foci patterns in TIP60 wild‐type cells. In summary, we propose the existence of an oncogenic‐specific γ‐H2AX foci pattern and the importance of a machine learning approach to extract oncogenic signaling among the γ‐H2AX foci variations.