Dataset Information

Deep learning imputes DNA methylation states in single cells and enhances the detection of epigenetic alterations in schizophrenia

ABSTRACT: DNA methylation (DNAm) is a crucial epigenetic mark and plays a key role in gene regulation, mammalian development, and various human diseases. Single-cell technologies enable profiling of DNAm states at cytosines within the DNA sequence of individual cells, but they often suffer from limited coverage of CpG sites. In this study, we introduce scMeFormer, a transformer-based deep learning model designed to impute DNAm states for each CpG site in single cells. Through comprehensive evaluations, we demonstrate the superior performance of scMeFormer compared to alternative models across five single-nucleus DNAm datasets from the human brain and mouse generated by distinct technologies. Remarkably, scMeFormer exhibits high-fidelity imputation, even when dealing with significantly reduced coverage, as low as 10% of the original CpG sites. We applied scMeFormer to a single-nucleus DNAm dataset generated from the prefrontal cortex of four schizophrenia patients and four neurotypical controls. This enabled the identification of thousands of differentially methylated regions associated with schizophrenia that would have remained undetectable without imputation and added granularity to our understanding of epigenetic alterations in schizophrenia within specific cell types. Our study highlights the power of deep learning in imputing DNAm states in single cells, advancing our understanding of the epigenetic bases of schizophrenia within specific brain cell types.

ORGANISM(S): Homo sapiens

PROVIDER: GSE285847 | GEO | 2025/01/08

REPOSITORIES: GEO

ACCESS DATA

Dataset's files

Source:

			Action	DRS
		Other

Items per page:

1 - 1 of 1

Similar Datasets

Project description:DNA methylation (DNAm) plays diverse roles in human biology, but this dynamic epigenetic mark remains far from fully characterized. Although earlier studies uncovered loci that undergo age-associated DNAm changes in adults, little is known about such changes during childhood. Despite profound DNAm plasticity during embryogenesis and early development, monozygotic twins show indistinguishable childhood methylation, suggesting DNAm is highly coordinated during the pediatric period. Here we examine the methylation of 27,578 CpG dinucleotides in peripheral blood DNA from 398 boys, aged 3 to 17 years, and find significant age-associated changes in DNAm at 2,078 loci. We report a deficit of such loci on the X chromosome, a preference for specific nucleotides immediately surrounding the interrogated CpG dinucleotide, and a primary association with developmental and immune ontological functions. These pediatric age-associated loci overlap significantly with those previously identified in adults (p < 0.001) but most of the pediatric loci are unique, suggesting many are childhood-specific. Meta-analysis (n = 1080) with two adult studies reveals that the methylation changes in 29.5% of the age-associated pediatric loci follow a linear pattern from childhood into adulthood; however, we also find a three-fold higher rate of change in children compared with adults and that a higher proportion of lifelong changes are more accurately modeled as a function of logarithmic age. We therefore conclude that DNAm changes occur more rapidly during childhood and are imperfectly accounted for by statistical corrections that are linear in age, further suggesting that future DNAm studies are matched closely for age. Age associated DNA methylation was evaluated at each CpG locus using a fixed-effects linear regression model adjusting for BeadChip as a covariate to account for batch effects. Analysis of variance (ANOVA) was conducted for each CpG locus to assess the age affect at each CpG locus which was subsequently corrected for multiple hypotheses using a Benjamini-Hochberg FDR. SUPPLEMENTARY FILES: * Matrix_AllSampleDetection.txt: Detection P-values for each sample and each loci * Matrix_AllSampleSignal.txt: Signal for Probe A and Probe B for each sample and each loci * Matrix_Non-Normalized_AllSampleBeta.txt: Beta values calculated using GenomeStudio v2010.3 Methylation Module 1.8.5 as Signal of Probe B / (Signal of Probe A + Signal of Probe B + 100) * Matrix_Normalized_AllSampleBetaPrime.txt: Beta Prime values used for analysis calculated from the Probe B and Probe A signals as Beta Prime = Signal of Probe B / (Signal of Probe A + Signal of Probe A)

Project description:Background: Measurement of genome-wide DNA methylation (DNAm) has become an important avenue for investigating potentially functional changes in various pathological conditions. Illumina Infinium is a relatively inexpensive and user-friendly DNAm microarray platform used by many researchers to measure DNAm on a large scale. However it has been suggested that a subset of probes may give rise to misleading results due to issues related to probe design. To facilitate biologically significant data interpretation, we set out to enhance probe annotation of the newest HumanMethylation450 BeadChip array (with >485,000 probes covering 99% of RefSeq genes). Results: Annotation that was added or expanded on includes 1) SNPs documented in the probe target, 2) probe binding specificity, 3) CpG classification of target sites and 4) gene feature classification of target sites. Probes with documented SNPs within 10bp of the target site and especially those with documented SNPs at the target CpG, were associated with increased within-tissue variation in DNAm. An example of a probe with a SNP at the target CpG was used to demonstrate how sample genotype can confound the measurement of DNAm. 8.6% of probes were identified as non-specific, in other words, these probes map to multiple locations in silico. DNAm measured from these non-specific probes likely represents a combination of DNAm from multiple genomic sites. The expanded biological annotation demonstrated that based on DNAm, grouping probes by alternative CpG classes rather than UCSC islands provides a more distinctive classification system of CpG sites. Finally variable enrichment for tissue-specific differentially methylated probes was noted across CpG classes and gene feature groups, depending on the tissues that were compared. Conclusion: Probes containing SNPs and non-specific probes may affect the assessment of DNAm using the 450k array. Additionally CpG enrichment classes and to a lesser extent gene feature groups were associated with distinct patterns of DNAm. Thus, we recommend that confounded probes be removed from analyses and that genomic trends be considered in analyses of the Illumina HumanMethylation450 BeadChip. DNAm arrays offer a powerful approach for which thoughtful use of probe content can be utilized to better understand the biological processes affected. Bisulfite converted DNA from 4 buccal, 4 blood and 4 chorionic villus samples was hybridized to the Illumina Infinium HumanMethylation450 BeadChip array.

Dataset Information

Deep learning imputes DNA methylation states in single cells and enhances the detection of epigenetic alterations in schizophrenia

Dataset's files

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets