ABSTRACT There is presently an urgent need to develop methodical approaches to evaluate the function of genomic vari- ants. Clinical genomic testing is growing rapidly, and with it a larger-than-ever number of genetic variants that cannot be defined as either disease-causing or benign. The problem affects clinicians and their patients, who struggle to understand and interpret molecular diagnostic reports, the implications of the results, and how to manage their patients in the absence of definitive information. Systems that involve single cells to generate high-content, high-resolution functional data are of paramount importance to solving the problem posed by var- iants of uncertain significance. Here, we propose to utilize a novel cell-based platform that uses machine learn- ing to determine the combination of morphological phenotypes that define pathogenicity. We will apply the technology to the comprehensive functional assessment of variants in IDS, the gene responsible for Hunter syndrome. The core hypothesis outlined in this proposal is that experimental data measuring the direct func- tional effects of variants will inform accurate disease risk prediction. In addition, we hypothesize that an in vitro, cell-based assay based on morphological features will more accurately detect disease compared to existing biochemical testing using artificial substrates. We will perform a functional assay in a variant library using the RaftSeq pipeline in the Buchser laboratory, where a cellular phenotype will be established and then tested us- ing a second set of variants combined with rescue experiments. The results will inform variant classification in IDS molecular testing and improve diagnosis of individuals including those identified by low iduronate-2-sulfa- tase activity on newborn screening.