Our primary hypothesis is that disease outcome for FIP can be predicted by the probability of furin-mediated cleavage of the FCoV spike protein at the S1/S2 site (within spike domain D). Our prediction is that viruses with cleaved spikes readily infect epithelial cells (e.g. enterocytes), transmit well (as expected for FECV), but are relatively unstable structurally; within-host selection of viral variants with point mutations in key residues of the FCS leads to a reduction in cleavage that stabilizes the spike, allows for infection of macrophages and drives the progression of FIP. In this scenario, there is range of FCS mutations that corresponds to a range of distinct FIPVs, which we believe can largely be predicted by scoring the FCS following integration of structural information and computational modeling driven by machine learning (at the deep learning level). In this project, we will use wet lab validation, combined with computational tools and machine-learning approaches to determine a probability score for FIP, based on the sequences generated from clinical samples. This score may then be used in isolation or potentially be integrated with data on other genomic hotspots and information of viral load and associated cytological and clinical parameters, to build a new predictive test for FIP. Such a test is urgently needed in the context of widespread availability of effective antiviral drugs, to allow diagnosis and identification of cats earlier in the course of disease, and to prevent misuse of drugs in cats that are not at risk of FIP.