Exploration of Sleep-Centered Clinical Phenotypes to Characterize Anxious and Non-Anxious Major Depression

Background Anxiety depressive disorder frequently co-occurs with anxiety symptoms and sleep disturbances, contributing to substantial heterogeneity in clinical presentation and suicide risk. Conversely, the interrelationships among depressive symptoms, anxiety, sleep dysfunction, and daytime functioning remain incompletely characterized. Here, we investigated clinical phenotypes of anxious and non-anxious depression using multidimensional questionnaire data and unsupervised machine learning. Methods The retrospective study included patients with major depressive disorder treated between January and December 2024 at the The Second People's Hospital of Hunan Province.Six validated instruments assessing affective symptoms, suicidal ideation, sleep, and circadian preference were administered at baseline. Patients were categorized into anxious and non-anxious depression groups, and unsupervised machine learning using the Apriori algorithm was employed to identify associations among symptom dimensions and functional outcomes. Results Patients with anxious depression tended to be older than those with non-anxious depression (mean age: 34.0 vs. 27.5 years), although this difference did not reach statistical significance (p = 0.069). Sex distribution was comparable between groups, with women comprising approximately two-thirds of participants in both cohorts (66.7% vs. 62.2%, p = 0.525). Ethnic composition was similarly distributed, with over 97% of participants identifying as Han ethnicity. Body mass index did not differ significantly between groups (22.06 vs. 21.00 kg/m², p = 0.319).Moreover, patients with anxious depression exhibited substantially greater anxiety severity, as indicated by significantly higher HAMA scores. HAMD(p = 0.079),HAMA(p < 0.05),SIOSS(p < 0.05),ISI(p < 0.05),PSQI(p < 0.05) have correlation with anxiety depression.HAMA is closely associated with PSQI, HAMD can also be associated with PSQI.Conversely, through a rule-learning approach based on unsupervised machine learning, we found that anxiety was most severe in married women with severe HAMD scores, followed by very poor sleep quality and the condition of patients with severe HAMD scores. Amony sleep phenotype of clinical syndrome,sleep apnea syndrome(β = 0.1204527,p < 0.05).In molecular,through analysing 169SNPs, mostly exert Intronic/Regulatory function, PDE4B got the highest score in anxiety groups.BDNF exists mutiple highest SNPs. 5' UTR has mean score 0.355. ADORA2A as only directly effect gene in anxiety, cause hope therapy in brain regional.RAD51,SUOX,ZFHX3 and SDK1 are likely to cause pathogenic mutations.MIR144,MIR29B1,MIR29B2,MIR30A got highest score(score = 0.328976) in anxiety population. Conclusions These findings indicate that anxious depression constitutes a clinically distinct phenotype characterized by tightly coupled affective and sleep-related dysfunction. Integrating multidimensional clinical assessments with data-driven analytical approaches may improve phenotypic stratification and inform targeted interventions aimed at mitigating functional impairment and suicide risk in anxiety depression.Furthermore, collecting GWAS and MDDOmics database ,we found BDNF-COMT-ADORA2A network help alleviate anxiety patients.