Background: Depressive symptoms are the strongest predictor for suicidal ideation (SI) (Kandel, Raveis, & Davies, 1991), along with non-suicidal self-harm (Burke et al., 2018). Feelings of hopelessness and self-directed hostility are higher in individuals who engage in repeated, deliberate self-harm (Brittlebank et al., 1990). The present study examines whether the GBI 7-Down (7D) items (excluding the suicidal ideation/better off dead item) (Youngstrom et al., 2013) or the Negative Affect (NA) total from the Differential Emotion Scale-IV (DES-IV) (Izard et al., 1994) better predicted SI. We hypothesized that the 7D would outperform NA at detecting SI because it specifically measures depressive symptoms, while the DES-IV measures all negative emotions. Methods: N=665 students in North Carolina and South Korea completed an anonymous survey including the 7U7D and the DES-IV during Psychology classes. Six 7D items served as one predictor set and the other was seven DES-IV negative affect subscores (self-hostility, anger, disgust, contempt, sadness, guilt, and shame). We compared predictor sets’ classification ability for SI presence/absence. Logistic regressions tested each predictor set with 1000 training and testing datasets bootstrapped from our original dataset. Parameters built in the training datasets were applied in testing datasets to generate the predictor sets’ predicted probabilities. We ran these predicted probabilities through ROC to generate bootstrapped AUCs to quantify classification performance. Delong’s test of paired AUCs evaluated if differences in each predictor set’s classification ability were statistically significant. The ROC curve selected optimal thresholds to maximize combined specificity and sensitivity. Results: The 7D items hopelessness (p=.0002), painful sadness (p=.011), and self-hate (p=.047) made statistically significant contributions. In the NA predictor set, self-hostility (p=4.11 x 10^-6) and sadness (p=.0019) made significant incremental predictions. The 7D predictor set produced a bootstrapped AUC=.84 (SE=.015), versus AUC=.80 (SE=.018) for the NA predictor set. The two sets did not differ in accuracy according to Delong’s test (D=-1.24, p=.293). At 7D predictor set’s optimal threshold, sensitivity=.82 and specificity=.74. At NA predictor set’s optimal threshold, sensitivity=.73 and specificity=.71. Discussion: Both depressive symptoms and negative affect significantly identified SI in an age-group at high risk for both ideation and behavior. The 7D and DES-IV NAs both detected SI with similar accuracy. On average, an individual with SI will have a higher 7D score than 84% of the controls (those without SI); someone with SI will have a higher NA score than 80% of the controls. Higher reports of 7D’s hopelessness and DES-IV’s self-hostility in the 7D measure incrementally predicted SI, consistent with prior work. To our knowledge, this is the first study to test and support that DES-IV NA scores can be used to identify SI. As a result, the DES-IV could be helpful in assessing suicidal risk indirectly. Future research should investigate the classification accuracy of SI when these predictor sets are combined into one set and perform regularization to determine the most significant predictors for SI.