Syntaxin Binding Protein, or STXBP1, is a human gene that is associated with neonatal and early onset epilepsy, as well as developmental delays. In this work, we replaced the C. elegans ortholog of STXBP1,
unc-18, with the human coding sequence for STXBP1. We then introduced 32 pathogenic and 25 benign variants into the sequence using CRISPR. The movement and morphology of the strains were analyzed using WormLab. We utilized the resultant 25 quantified features to train two parallel machine learning classifiers, Random Forest (RF) and Support Vector Machines (SVM). The classifiers were evaluating using the area under the Receiver Operating Characteristic curve (AUROC) and the precision-recall metric (PR). Both algorithms were found to perform well (AUROC = .94 and .85 for SVM and RF, respectively) and agreed on all variants except two 2 benign and 2 pathogenic. We next introduced 24 Variants of Uncertain Significance (VUS) into the human coding sequence for STXBP1 and phenotype the resultant strains. We found 6 variants that both classifiers classified as Pathogenic, 15 variants that both classifiers classified as Benign, and 3 for which the two classifiers disagreed. We conclude that protein humanization, variant introduction, and phenotypic characterization, and machine learning classification is a viable workflow for the functional analysis of variants in the STXBP1 gene and can help to resolve VUS that are clinically relevant.