The mental health of Chinese graduate students presents a growing concern, with subthreshold depression—a clinically significant precursor to major depressive disorder—exhibiting particularly high prevalence rates. Despite its public health implications, research specifically examining subthreshold depression in this population remains limited within the Chinese context. This study addresses this critical gap by systematically investigating the phenotypic manifestations and environmental determinants of subthreshold depression among Chinese graduate students, with the ultimate goal of informing early intervention strategies. Grounded in ecological systems theory, our mixed-methods approach synergistically integrates computational linguistics techniques (including large-scale text mining of Bilibili and Zhihu platforms), machine learning analytics, and psychometric scale development. Through rigorous grounded theory analysis of acquired textual data, we established three validated lexical databases capturing: (1) symptom manifestations, (2) environmental risk factors (12 subcategories), and (3) environmental protective factors (10 subcategories), all organized within four macro-domains (familial, academic, interpersonal, and societal contexts). Our machine learning validation phase yielded robust classification accuracy for environmental factors (77.39% for risks; 86.21% for protective factors) using BERT-based models. The developed Environmental Determinants Questionnaire for Subthreshold Depression (EDQ-SD), refined through expert review and pilot testing (N=517), demonstrated excellent psychometric properties (α=0.899 for risk factors; α=0.956 for protective factors). Key findings revealed distinct symptom profiles, with behavioral manifestations (particularly relational avoidance) being most prevalent, followed by cognitive features (self-deprecation, rumination). Ecological analysis identified microsystem factors (family, academic, and interpersonal environments) as exerting the strongest influence, with protective factors serving as significant negative predictors of depression severity (β=-0.42, p<.001), though demonstrating nonsignificant moderating effects. Notably, temporal analysis highlighted the disproportionate impact of current-phase stressors compared to past or anticipated future stressors. These findings substantially advance the operationalization of subthreshold depression in academic populations while providing empirically-derived assessment tools and targeted prevention frameworks. We recommend institutional mental health initiatives prioritize microsystem-level interventions while incorporating temporal stressor management competencies to effectively address this growing public health challenge.

Subthreshold Depression, Graduate Students, Ecological Systems Theory, Risk/Protective Factors, Computational Psychopathology