Prediction of protein–protein interactions between anti-CRISPR and CRISPR-Cas using machine learning technique

CRISPR-Cas system, responsible for bacterial adaptive immune response, has evolved as the game-changer in the field of genome editing and has revolutionized both animal and plant research owing to its efficiency and feasibility. CRSIPR- associated (Cas) protein, an integral component of the CRSIPR-Cas toolkit, cut the target genetic material for making the desirable edits. However, unchecked nuclease activity of Cas protein may lead to unforeseen off-target effects. Anti-CRISPR (Acr), small proteins usually found in phages and other mobile genetic elements, are the natural inhibitors of the Cas proteins that help phages to escape the immune system of the host. Acr proteins regulate the activity of the Cas nuclease by interacting with its different domains which results in the blockage of CRISPR activity. Thus, it is essential to understand the interactions between these two rival proteins in order to switch off the cutting machinery when needed. Experimental methods to identify protein–protein interaction, are often costly, time-consuming, and labor-intensive. Computational strategies, such as data- driven predictive models, can complement experimental studies by providing fast, efficient, reliable, and cheaper alternatives to predict protein interactions. Herein, we report the first machine learning-based predictive model to identify novel interactions between Acr and Cas proteins using an ensemble strategy. The accuracy of our proposed ensemble model was more than 95%, indicating its high predictive power. The developed model can contribute to automate the process of discovering the natural inhibitors of Cas protein for controlling the off- target cleavage and improving the efficiency of CRISPR-Cas technology. To extend the support for diverse levels of end-users, a web application named AcrCasPPI was developed which is available at http://login1.cabgrid.res.in:5020/ . Alternatively, a python package named acrcasppi-ml, is also available at https://pypi.org/project/acrcasppi-ml/ .