Obsessive-compulsive disorder (OCD) is heterogeneous in the way its symptoms present, and some individuals with OCD experience sensory-based urges that drive compulsive behaviors (e.g., uncomfortable physical sensations that drive a need to repeat a behavior to achieve a ‘just right’ feeling. The current study uses machine learning (ML)-based latent profile analysis to categorize subgroups of OCD patients based on ability to suppress eye-blinks to better understand different presentations of OCD. Participants were recruited from two New York-based medical centers; the final sample contained 82 OCD patients and 38 controls. Clinical interviews and self-report measures determined levels of OCD, anxiety, and depression symptoms. Eyeblink suppression, which asks participants to avoid blinking their eyes for prolonged periods of time, was chosen as a task to serve as a model for sensory-based urges in OCD. While participants engaged with the task, fMRI data was gathered to determine brain region activation, and eyeblinks were measured using Eyelink 1000-Plus eye tracking technology. Latent profile analysis to determine subgroups of patients was completed using Mplus software. To determine the best model of number of profiles, the Bayesian Information Criterion (BIC) was used. A four-latent group model of three patient groups and one control group best fit the study data. The first subgroup (OCD_Lo) had the fewest erroneous blinks (eyeblinks occurring during periods of suppression), the second (OCD_Mod) had a moderate number of erroneous blinks, and the third (OCD_Hi) had the highest number of erroneous blinks. When evaluating differences in clinical manifestations, the subgroups did not differ in overall OCD severity, however the OCD_Hi subgroup had higher interoceptive sensitivity; more OCD symptoms related to symmetry, completeness, and “not just-right experiences”; and more severe sensory symptoms. According to the neuroimaging results, early in the suppression period the OCD_Hi and OCD_Mod subgroups showed greater activation than the OCD subgroups in brain areas that have been associated with urges including the occipital region, insula, and parahippocampal gyrus. The OCD_Mod group also showed greater activation than the OCD_Lo subgroup in multiple frontal regions. Late in the suppression period, the OCD_Hi group also showed reduced activation in the middle frontal gyrus compared to the OCD_Mod subgroup. This work contrasts with previous work that has viewed OCD patients as a homogeneous group compared to those without OCD, as it uses an urge-suppression task and ML analysis methods to identify subgroups of patients with varied degrees of sensory-related symptoms. These findings suggest the potential of new treatments that target improving urge suppression and interoceptive sensitivity for subgroups of patients with OCD. These findings are supported by differences in brain region activation between the subgroups identified. One significant limitation of this study is that the urge to blink may differ qualitatively from the urge to perform compulsions among those with OCD.