Poster 33

[P-33-01] Decoding Subconscious Emotional Regulation Pathways in PTSD: Integrative Deep Learning of rs-fMRI and Chronobiological Epigenetics

*Prihantini Prihantini¹, Rifaldy Fajar², Sahnaz Vivinda Putri³, Andi Nursanti Andi Ureng⁴, Asfirani Zahaz⁵ (1.Bandung Institute of Technology(Indonesia), 2.Yogyakarta State University(Indonesia), 3.International University Semen Indonesia(Indonesia), 4.Andini Persada College of Health Sciences(Indonesia), 5.Bonto-Bonto General Hospital(Indonesia))

Keywords:

PTSD,Subconscious Emotional Regulation,Resting-State fMRI (rs-fMRI),Chronobiological Epigenetics,Deep Learning Integration

Background/Aim: Post-Traumatic Stress Disorder (PTSD) significantly disrupts emotional regulation, yet the subconscious neural mechanisms remain underexplored. Most studies focus on conscious emotional responses, neglecting the interplay of neural connectivity and epigenetic chronobiology. This study aims to decode subconscious emotional regulation pathways in PTSD by integrating multi-scale resting-state functional MRI (rs-fMRI) connectivity and chronobiological epigenetic markers using deep learning. Methods: Data were derived from the UK Biobank (n=2,472; PTSD diagnoses: n=1,125) and GEO (DNA methylation data, n=1,048). Emotional regulation metrics were calculated using validated cognitive-emotional harmonization tasks. A convolutional neural network (CNN) was developed to extract rs-fMRI features, emphasizing connectivity in amygdala-prefrontal cortex (PFC) and hypothalamic pathways. Attention-based transformers analyzed clock gene methylation patterns (CLOCK, PER1, BMAL1) to detect PTSD-specific chronobiological disruptions. A variational autoencoder (VAE) fused neural and epigenetic features into a unified latent representation. Generalized additive models predicted emotional regulation outcomes, validated via 10-fold cross-validation. Key metrics included AUROC for accuracy, feature interpretability (SHAP values), and neurobiological clustering. Results: The model achieved an AUROC of 0.84 (95% CI: 0.81–0.87) in predicting PTSD-related emotional regulation deficits. Disrupted amygdala-hypothalamic connectivity (n=-0.41, p<0.001) strongly correlated with PTSD. Chronobiological disruptions in PER1 and CLOCK methylation patterns were linked to amygdala-PFC dysfunction (r=0.68, p<0.001). Patients with greater latent disruption scores exhibited significantly impaired emotional regulation metrics (effect size=1.42, p<0.0001). The fusion model improved predictive performance by 21% over single-modality models. Conclusions: This study identifies disrupted amygdala-hypothalamic connectivity and chronobiological epigenetic alterations as key factors underlying subconscious emotional regulation deficits in PTSD, offering a novel integrative framework for predictive modeling and therapeutic strategies.