Increased expression of CD33 in the brain has been suggested to be associated with increased amyloid plaque burden, while the peripheral level of CD33 in Alzheimer’s disease (AD) patients and its role in AD remain unclear. The current study aimed to systematically explore the bidirectional relationship between peripheral CD33 and AD. Genome-wide association study (GWAS) datasets of AD (Ncases: 21982; Ncontrols: 41944), blood CD33 mRNA level, the plasma CD33 protein level, and CD33 expression on immune-cell subtypes were obtained from GWASs conducted in the European population. Eligible IVs were extracted from the GWASs. MR estimates were calculated by inverse-variance weighting (IVW) and other sensitivity analyses. The main statistical analyses were conducted using TwoSampleMR (v.0.5.5) in R package (V.4.1.2).In the forward MR analysis (CD33 as exposure and AD as outcome), the IVW results indicated that elevated blood CD33 mRNA level (OR [95% CI] = 1.156[1.080, 1.238], p = 3.25e-05), elevated serum CD33 protein level (OR [95% CI] = 1.08 [1.031, 1.139], p = 1.6e-03) and increased CD33's expression on immune cell subtypes (p < 0.05) were all leading to a higher risk of AD. And sensitivity analyses supported these findings. While the reverse MR analysis (AD as exposure and CD33 as outcome) indicated that AD was not leading to the elevation of CD33's protein level in the blood (p > 0.05). In conclusion, our results indicated that elevated peripheral expression of CD33 was causal to the development of AD. Future studies are needed to work on developing CD33 as a biomarker and therapeutic target in AD.

Background Schizophrenia (SCZ) is a chronic and severe mental illness with no cure so far. Mendelian randomization (MR) is a genetic method widely used to explore etiologies of complex traits. In the current study, we aimed to identify novel proteins underlying SCZ with a systematic analytical approach. Methods We integrated protein quantitative trait loci (pQTLs) of the brain, cerebrospinal fluid (CSF), and plasma with the latest and largest SCZ genome-wide association study (GWAS) via a systematic analytical framework, including two-sample MR analysis, Steiger filtering analysis, and Bayesian colocalization analysis. Results The genetically determined protein level of C4A/C4B (OR = 0.70, p = 1.66E−07) in the brain and ACP5 (OR = 0.42, p = 3.73E−05), CNTN2 (OR = 0.62, p = 2.57E−04), and PLA2G7 (OR = 0.71, p = 1.48E−04) in the CSF was associated with a lower risk of SCZ, while the genetically determined protein level of TIE1 (OR = 3.46, p = 4.76E−05), BCL6 (OR = 3.63, p = 1.59E−07), and MICB (OR = 4.49, p = 2.31E−11) in the CSF were associated with an increased risk for SCZ. Pathway enrichment analysis indicated that genetically determined proteins suggestively associated with SCZ were enriched in the biological process of the immune response. Conclusion In conclusion, we identified one protein in the brain and six proteins in the CSF that showed supporting evidence of being potentially associated with SCZ, which could provide insights into future mechanistic studies to find new treatments for the disease. Our results also supported the important role of neuroinflammation in the pathogenesis of SCZ.

BackgroundAlzheimer’s disease (AD) is the leading cause of dementia. Currently, there are no effective disease-modifying treatments for AD. Mendelian randomisation (MR) has been widely used to repurpose licensed drugs and discover novel therapeutic targets. Thus, we aimed to identify novel therapeutic targets for AD and analyse their pathophysiological mechanisms and potential side effects.MethodsA two-sample MR integrating the identified druggable genes was performed to estimate the causal effects of blood and brain druggable expression quantitative trait loci (eQTLs) on AD. A repeat study was conducted using different blood and brain eQTL data sources to validate the identified genes. Using AD markers with available genome-wide association studies data, we evaluated the causal relationship between established AD markers to explore possible mechanisms. Finally, the potential side effects of the druggable genes for AD treatment were assessed using a phenome-wide MR.ResultsOverall, 5883 unique druggable genes were aggregated; 33 unique potential druggable genes for AD were identified in at least one dataset (brain or blood), and 5 were validated in a different dataset. Among them, three prior druggable genes (epoxide hydrolase 2 (EPHX2),SERPINB1andSIGLEC11) reached significant levels in both blood and brain tissues. EPHX2 may mediate the pathogenesis of AD by affecting the entire hippocampal volume. Further phenome-wide MR analysis revealed no potential side effects of treatments targetingEPHX2,SERPINB1orSIGLEC11.ConclusionsThis study provides genetic evidence supporting the potential therapeutic benefits of targeting the three druggable genes for AD treatment, which will be useful for prioritising AD drug development.

Purpose To develop a model for predicting response to total neoadjuvant treatment (TNT) for patients with locally advanced rectal cancer (LARC) based on baseline magnetic resonance imaging (MRI) and clinical data using artificial intelligence methods. Methods Baseline MRI and clinical data were curated from patients with LARC and analyzed using logistic regression (LR) and deep learning (DL) methods to predict TNT response retrospectively. We defined two groups of response to TNT as pathological complete response (pCR) versus non-pCR (Group 1), and high sensitivity [tumor regression grade (TRG) 0 and TRG 1] versus moderate sensitivity (TRG 2 or patients with TRG 3 and a reduction in tumor volume of at least 20% compared to baseline) versus low sensitivity (TRG 3 and a reduction in tumor volume <20% compared to baseline) (Group 2). We extracted and selected clinical and radiomic features on baseline T2WI. Then we built LR models and DL models. Receiver operating characteristic (ROC) curves analysis was performed to assess predictive performance of models. Results Eighty-nine patients were assigned to the training cohort, and 29 patients were assigned to the testing cohort. The area under receiver operating characteristics curve (AUC) of LR models, which were predictive of high sensitivity and pCR, were 0.853 and 0.866, respectively. Whereas the AUCs of DL models were 0.829 and 0.838, respectively. After 10 rounds of cross validation, the accuracy of the models in Group 1 is higher than in Group 2. Conclusion There was no significant difference between LR model and DL model. Artificial Intelligence-based radiomics biomarkers may have potential clinical implications for adaptive and personalized therapy.

No abstract