Although lipid biology may play a key role in the pathophysiology of mental health disorders such as schizophrenia (SCZ) and bipolar disorder (BD), the nature of this interplay and how it could shape phenotypic presentation, including cognitive performance is still incompletely understood. To address this question, we analyzed the association of plasma level of different lipid species with cognitive performance in the transdiagnostic PsyCourse Study. Plasma lipidomic profiles of 623 individuals (188 SCZ, 243 BD, 192 healthy controls) belonging to the PsyCourse Study were assessed using liquid chromatography and untargeted mass spectrometry. The association between 364 annotated lipid species from 16 lipid classes and six cognitive tests was evaluated. Likewise, the association of polygenic risk scores (PRS) for SCZ, BD, executive function (EF), and educational attainment (EA) with lipid plasma levels were also investigated. In the regression analysis, three lipid species belonging to phosphatidylethanolamine plasmalogen and one belonging to ceramide class showed significant negative association with Digit-Symbol test scores. Lipid class-based enrichment analysis in LipidR replicated the significance of the phosphatidylethanolamines class for the Digit-Symbol test, which evaluates the processing speed in cognitive tasks. Polygenic load for SCZ, BD, EF, or EA was not associated with lipid levels. Our findings suggest a link between lipids and cognitive performance independent of mental health disorders. Still, independent replication is warranted to better understand if phosphatidylethanolamines could represent an actionable pharmacologic target to tackle cognitive dysfunction, an important unmet clinical need that affects long-term functional outcomes in individuals with severe mental health disorders.

Psychotic and affective disorders, including schizophrenia (SCZ) and depression (MDD), affect millions of people globally. The overlapping symptoms of these diseases and the lack of objective diagnostic tools could lead to misdiagnosis. Recent studies suggest that the analysis of plasma lipid levels may help to develop new diagnostic tools. In this study, we investigated the plasma lipidome of psychiatric patients and healthy controls to identify disease-specific lipid species. Using untargeted mass spectrometry, we profiled blood plasma lipids from 416 patients with common psychotic and affective disorders and 272 healthy individuals from two different cohorts. We observed lipidome alterations in SCZ and MDD consistent with earlier findings. In total, 144 lipids showed significant changes, with 107 of them being concordant across both disorders, and 37 being discordant. Lipids that differentiated SCZ from MDD were mainly triacylglycerols with polyunsaturated fatty acid residues decreased in MDD. In an additional group of 111 patients with bipolar, schizotypal, and schizoaffective disorders, these lipid markers suggested a trend toward separating psychotic and affective disorders. Furthermore, a logistic regression model trained on lipid data distinguished SCZ from MDD with an ROC AUC of 0.83. Taken together, these results suggest that blood lipid profiling may aid in the objective differentiation of psychotic and affective disorders.

Schizophrenia (SCZ) and bipolar disorder (BD) are characterised by pervasive clinical heterogeneity, compromising biological discovery and individualized treatment. Data-driven clustering has identified a transdiagnostic severe psychosis subtype marked by high psychotic symptom load, cognitive impairment, and poor functioning, yet its biological underpinnings remain to be understood. To address this, we examined associations among 14 PRS, plasma lipidomics (361 species across 16 classes), and a machine-learning-derived severe psychosis probability score in a sample spanning the psychosis spectrum (PRS n=1,190; lipid subset n=428) in the PsyCourse study. Genetic risk and lipid profiles were both associated with illness severity but appeared to act independently: SCZ-PRS was positively associated with severe psychosis subtype probability (q=0.045), while BD-PRS (q=0.02) and educational attainment PRS (q=0.045) showed negative associations. Eight lipid classes were enriched with severe subtype probability, including increased sphingolipids (dSM, dCer), glycerophospholipids (PC, PE-P, PE), TAG, and CE, alongside decreased PC-P. Enrichment patterns were robust to adjustment for diagnosis, medication, and symptom severity. PRS and lipid profiles each explained 7–8% of variance, jointly ~16%. No significant mediation or shared genetic-lipidomic covariation was detected. Together, these findings suggest that polygenic liability and lipidomic dysregulation contribute to disease severity through partly independent pathways, and their convergence on a symptom-derived severity score supports the premise that multidimensional phenotyping can reveal biological structure that diagnostic categories alone obscure. The lipidomic signal points toward pharmacologically tractable pathways, particularly ether-lipid remodelling and sphingolipid dysregulation, suggesting that lipidomics and polygenic risk are complementary dimensions for understanding and ultimately treating severe psychosis.

Aging drives tissue-specific lipidome remodeling, but comprehensive multi-tissue lipidomic profiles in rodents remain scarce. Here, we conducted untargeted LC–MS lipidomics on frontal cortex (FC), cerebellum (CB), liver (LV), muscle (MS), and plasma (PL) from young and old male Wistar rats, quantifying 884 lipids across 36 classes. Muscle showed the largest number of lipidomic changes, with PG and PE O- enrichment indicating mitochondrial dysfunction. Plasma showed alterations in phosphocholine LPC and PC O-, while liver displayed an unexpected decline in triglycerides, particularly shorter- and odd-chain species, consistent with reduced gut microbiota–derived fatty acids. Brain regions showed myelin-related changes: decreased MGDG O- in cerebellum and mixed SHexCer alterations in the frontal cortex. Our cross-tissue analyses revealed that aging trajectories clustered by tissue function rather than baseline composition, with muscle more closely resembling brain, and liver mirroring plasma alterations. While no individual lipid showed ubiquitous age-related changes across all tissues, we found consistent patterns at the lipid class level, particularly among solid tissues. Notably, PG showed the strongest coordinated age-related changes, highlighting mitochondrial lipid remodeling as a common aging feature. Our data supports tissue-specific lipid aging processes, as well as mitochondrial dysfunction patterns shared across multiple tissues.

In this paper, we tackle a problem of predicting phenotypes from structural connectomes. We propose that normalized Laplacian spectra can capture structural properties of brain networks, and hence graph spectral distributions are useful for a task of connectome-based classification. We introduce a kernel that is based on earth mover's distance (EMD) between spectral distributions of brain networks. We access performance of an SVM classifier with the proposed kernel for a task of classification of autism spectrum disorder versus typical development based on a publicly available dataset. Classification quality (area under the ROC-curve) obtained with the EMD-based kernel on spectral distributions is 0.71, which is higher than that based on simpler graph embedding methods.