Single-nucleus RNA sequencing and machine learning identify CACNA1A as a myocyte-specific biomarker for sudden unexplained death in schizophrenia

Abstract

Sudden unexplained death (SUD) in schizophrenia (SUD-SCZ) represents a distinct subtype of SUD and its diagnosis is particularly challenging in forensic pathology. In this study, we designed a three-phase analytical strategy to identify single-cell level biomarker(s) for the diagnosis of SUD-SCZ. In the discovery phase, we performed single nucleus RNA sequencing analysis of four cardiac tissues from SUD-SCZ cases. We found that cardiomyocytes showed the most prominent transcriptional changes, with significant upregulation of calcium channel-related genes such as CACNA1A, and marked downregulation of key contractile genes including ANKRD1 and PLN. In the validation phase, immunohistochemistry (IHC) staining was conducted on cardiac tissues from an independent cohort encompassing 29 SUD-SCZ cases and 54 age- and gender-matched control cases. CACNA1A protein expression was significantly elevated in the SUD-SCZ samples, with strong cell-type specificity. In the application phase, based on IHC image features extracted from the CACNA1A staining slides-including color, intensity, and texture etc-we developed a machine learning classifier and constructed a user-friendly diagnostic software (https://github.com/cxlcoc/SUD-SCZ-Diagnosis_v1.0). The final model achieved 99 % diagnostic accuracy in distinguishing SUD-SCZ from control samples in the internal case set, and accurately diagnosed all the five cases in an external test set. This study identifies CACNA1A as a novel biomarker for SUD-SCZ and establishes a paradigm that integrates IHC-based images and artificial intelligence, which may also benefit the forensic identification of SUD in a more generalized population.