A machine learning case-control classifier for schizophrenia based on DNA methylation in blood

Area of Science:

• Neuroscience
• Genetics
• Epigenetics

Background:

• Epigenetic dysregulation, particularly DNA methylation, is implicated in schizophrenia (SZ) etiology.
• Cell type-specific methylation patterns present challenges for population-based epigenetic studies of SZ.
• Systemic interindividual epigenetic variation (CoRSIVs) offers a potential avenue for SZ epigenetic research.

Purpose of the Study:

• To develop and validate an SZ case-control classifier using DNA methylation data from blood.
• To investigate the utility of CoRSIVs for SZ risk prediction.
• To explore the relationship between genetic and epigenetic risk factors in schizophrenia.

Main Methods:

• Utilized Illumina Human Methylation 450K (HM450) DNA methylation data from 414 SZ cases and 433 controls for training.
• Applied sparse partial least squares discriminate analysis (SPLS-DA) to identify predictive methylation patterns.
• Calculated a "risk distance" based on SPLS-DA dimensions to assess SZ probability and validated on an independent cohort.

Main Results:

• The CoRSIV-based classifier achieved an 80% positive predictive value (PPV), outperforming polygenic risk score (PRS) models.
• Risk distance derived from methylation was independent of medication use, mitigating concerns of reverse causality.
• A significant positive correlation between risk distance and PRS (r=0.28, P<1.28x10^-12) was observed, with mediation analysis suggesting genetic influences on SZ are partly mediated by CoRSIV methylation.

Conclusions:

• Developed a novel, highly accurate SZ classifier based on systemic DNA methylation variants in blood.
• Identified two innate dimensions of SZ risk: genetic and systemic epigenetic.
• Findings support a model where genetic predisposition to SZ is partially mediated by epigenetic alterations at CoRSIVs.