Cognitive signatures of depressive and anhedonic symptoms, and affective states, using computational modeling and neurocognitive testing.

BACKGROUND: Deeper phenotyping may improve our understanding of depression. Because depression is heterogeneous, extracting cognitive signatures associated with severity of depressive symptoms, anhedonia, and affective states is a promising approach.

METHODS: Sequential sampling models (SSMs) decomposed behavior from an adaptive approach-avoidance conflict (AAC) task into computational parameters quantifying latent cognitive signatures. Fifty unselected participants completed clinical scales and the AAC task by either approaching or avoiding trials offering monetary rewards and electric shocks.

RESULTS: Decision dynamics were best captured by an SSM with linear collapsing boundaries varying by net offer values, and with drift rates varying by trial-specific reward and aversion, reflecting net evidence accumulation towards approach or avoidance. Unlike conventional behavioral measures, these computational parameters revealed distinct associations with self-reported symptoms. Specifically, passive avoidance tendencies, indexed by starting point biases, were associated with greater severity of depressive symptoms (R=0.34, p=0.019) and anhedonia (R=0.49, p=0.001). Depressive symptoms were also associated with slower encoding and response execution, indexed by nondecision time (R=0.37, p=0.011). Higher reward sensitivity for offers with negative net values, indexed by drift rates, was linked to more sadness (R=0.29, p=0.042) and lower positive affect (R=-0.33, p=0.022). Conversely, higher aversion sensitivity was associated with more tension (R=0.33, p=0.025). Finally, less cautious response patterns, indexed by boundary separation, were linked to more negative affect (R=-0.40, p=0.005).

CONCLUSIONS: We demonstrate the utility of multidimensional computational phenotyping, which could be applied to clinical samples to improve characterization and treatment selection.