CRISPR-Cas9, a promising gene-editing tool, sheds light on gene therapy. The normal DNA cleavage of CRISPR-Cas9 is programmed by a guide RNA (gRNA) template. However, recent studies show that Cas9 cleavage occurs even without guidance of a gRNA in the presence of Mn2+ ions, implying the existence of off-target effects of Cas9 [1]. In this study, we revealed the possible mechanism of RI-cleavage and further engineered Cas9 to reduce RI-cleavage via the power of machine learning.