Multi-omic data integration improves the resolution of the molecular etiology of autism in a mouse model.

Identification of EIF4A1 as a candidate molecular indicator of autism spectrum disorder using integrative bioinformatic and biological methods.

Autism Spectrum Disorder (ASD) is a heterogeneous neurodevelopmental condition with complex genetic and molecular mechanism. Identifying reliable molecular biomarkers remains a critical challenge. In this study, we integrated mRNA expression profiles from five post-mortem brain tissue GEO datasets to identify ASD-associated genes. Following batch effect correction, differentially expressed genes (DEGs) were analysed and Weighted Gene Co-expression Network Analysis (WGCNA) was performed to screen genes correlated with ASD. Then, five machine learning algorithms - Random Forest, LASSO, Boruta, CatBoost, and LightGBM - were applied to screen hub genes. Lastly, alterations of the hub gene(s) were investigated with a maternal immune activation (MIA) rat model using poly I:C by measuring mRNA expression of the hub genes in the rat nucleus accumbens (NAc) and caudate putamen (CPu). A total of 30 DEGs and 54 WGCNA module genes were identified, yielding 29 key candidates by intersecting these two gene sets. EIF4A1 (Eukaryotic Translation Initiation Factor 4A1) was the sole gene consistently ranked among the top five by all five machine learning algorithms. Analysis of the integrated dataset confirmed that EIF4A1 mRNA expression was significantly elevated in ASD subjects. Finally, using the MIA rat model of ASD, we found that EIF4A1 mRNA expression was significantly down-regulated in the NAc and CPu, and this deficit was rescued by treatment with the antipsychotics olanzapine or risperidone. In conclusion, the present study positions EIF4A1 as a promising candidate molecular indicator with potential implications for understanding disease mechanisms and developing targeted interventions of ASD.

An integrated systems biology and machine learning framework for identifying potential biomarkers and pathways in autism spectrum disorder

Background Autism spectrum disorders (ASD) are a group of neurodevelopmental disorders whose underlying molecular mechanisms and biological processes remain incompletely understood. In this study, we used a multi-layered systems biology approach to prioritize candidate genes and regulatory factors associated with ASD. Method Gene expression data from peripheral blood samples were obtained from the Gene Expression Omnibus (GEO) database (GSE18123). Using analyses performed in R software, differentially expressed genes (DEGs) in patients with ASD were identified (p-value < 0.05 and |log2FC| > 0.5). These DEGs were used to perform weighted gene co-expression network analysis (WGCNA) and construct a protein–protein interaction (PPI) network. By integrating the results of these network analyses with feature selection techniques (LASSO and random forest feature importance), candidate genes associated with ASD were prioritized and evaluated using qRT-PCR in the valproic acid (VPA)-induced rat model of autism. Furthermore, a gene regulatory network (GRN) was constructed to identify the regulatory factors associated with DEGs. Result TLR8 and CASP4 were prioritized as candidate genes that may be associated with ASD, because they were located within the co-expression module that showed the strongest correlation with ASD, were identified as key nodes of the PPI network, and were selected by feature selection algorithms. Our experimental validation showed increased expression of TLR8 and CASP4 in the autism model compared with controls; TLR8 was upregulated in both the hippocampus and peripheral blood, whereas CASP4 was upregulated only in the hippocampus. Furthermore, GRN analysis identified miR-891b and miR-627-3p as potential regulators of TLR8, and miR-26b-5p as associated with CASP4. Conclusion These findings indicate that CASP4 and TLR8, together with their associated regulatory miRNAs, may represent promising biomarkers and potential therapeutic targets for future ASD research and contribute to a better understanding of the pathophysiological mechanisms underlying ASD.

Circulating neuronal proteins are elevated in severe autism and associated with patient-specific neuronal dysfunction and behavioral deficits.

Key links between peripheral protein dysregulation and neuronal function and behavior are revealed, offering new insights into systemic contributions to ASD pathophysiology and highlighting potential therapeutic targets for mitigating symptom severity.

Autism mutations rewire protein interaction networks to drive neurodevelopmental pathology.

A placental transcriptional signature for autism

Autism development involves multiple genetic and early-life environmental factors. Studying the placenta's gene expression profile may reveal key mechanistic pathways in autism development. Here, using a nested case-cohort design within an Australian population-derived prebirth cohort study (n=1074), we identified 1,644 differentially expressed genes (DEGs; FDR<0.05) in the placenta of children with autism diagnosis (n=43), compared to those without (n=120). The top enriched pathways related to mitochondrial translation, oxidative stress, RNA processing and transcription regulation. CYP1A1, the most important xenobiotic-metabolising enzyme of the placenta, was the top downregulated DEG in the placenta of children with autism, while immuno-regulatory human leukocyte antigen (HLA)-related genes were among the top upregulated DEGs. A machine learning-based approach predicted autism from the transcriptomic data with a median sensitivity of 0.57 (2.5th-97.5th centiles: 0.29, 0.76) and median specificity of 0.92 (2.5th-97.5th centiles: 0.78, 0.98). Weighted Gene Correlation Network Analysis identified eight affected placental gene modules, with the largest five modules being enriched primarily for mitochondrial bioenergetics, oxidative phosphorylation and RNA processing pathways. This placental transcriptomic signature of impaired mitochondrial function and gene transcription regulation among infants subsequently diagnosed with autism has profound implications for understanding both risk factors and prediction, suggesting the possibility of identifying modifiable prenatal pathways to improve autism outcomes.

Statistical integration of GWAS, eQTL and pQTL data for multi-omics analysis of Alzheimer's Disease

Alzheimer's Disease (AD) is a complex neurodegenerative disorder with a strong genetic architecture. Genome-Wide Association Studies (GWAS) have identified numerous susceptibility loci. However, the majority of associated variants reside in non-coding regions, making it difficult to resolve their functional consequences and identify causal genes. To address this limitation, integration of GWAS with expression Quantitative Trait Loci (eQTL) and protein Quantitative Trait Loci (pQTL) data has emerged as a key strategy for linking genetic variation to downstream molecular phenotypes. This review discusses statistical frameworks for multi-omics integration in AD research, with a focus on approaches that enable causal inference and gene prioritization. Major methods include colocalization analysis for detecting shared causal variants, Mendelian Randomization (MR) for assessing putative causal relationships, and Transcriptome-Wide Association Studies (TWAS) for linking genetically predicted gene expression to disease risk. Applications of these frameworks have facilitated the identification of candidate causal genes and proteins, thereby improving the mechanistic interpretation of AD-associated loci. However, challenges remain, including tissue specificity and cell-type specificity, limited ancestral diversity in available datasets, and constraints in causal inference. Emerging single-cell and spatial multi-omics approaches are expected to provide a more detailed characterization of AD-associated molecular mechanisms while supporting therapeutic target discovery.