A Novel Cell-Type Deconvolution Algorithm Reveals Substantial Contamination by Immune Cells in Saliva, Buccal and Cervix

Shi-qing, Zheng; Webster, Amy P; Dong, Danyue; Feber, Andrew; Graham, David; Sullivan, Roisin; Jevons, Sarah; Lovat, Laurence; Beck, Stephan; Widschwendter, Martin; Teschendorff, Andrew E.

doi:10.2217/epi-2018-0037

Cited by 144 publications

(198 citation statements)

References 51 publications

Supporting

Mentioning

186

Contrasting

Order By: Relevance

“…hypomethylated in one and hypermethylated in another) ( Fig.1d ), and which may not be identifiable by current state-of-the-art DMC calling algorithms (see later). To estimate cell-type fractions, CellDMC applies our previously validated EpiDISH algorithm 16 in an iterative hierarchical procedure, called HEpiDISH 17 , which leads to improved cell-type fraction estimates in complex tissues by recognizing that cell-types are naturally arranged along a developmental tree ( Online Methods, SI fig.S2 ). In the context of epithelial tissues, HEpiDISH accomplishes this by using two distinct DNAm reference matrices, a primary reference matrix for the estimation of total epithelial, total fibroblast and total immune-cell (IC) fractions, and a separate, secondary, non-overlapping DNAm reference for the estimation of underlying IC cell subtype fractions.…”

Section: Resultsmentioning

confidence: 99%

Identification of differentially methylated cell-types in Epigenome-Wide Association Studies

Breeze

Beck

Teschendorff

2018

Preprint

Self Cite

View full text Add to dashboard Cite

An outstanding challenge of Epigenome-Wide Association Studies (EWAS) performed in complex tissues is the identification of the specific cell-type(s) responsible for the observed differential DNA methylation. Here, we present a novel statistical algorithm, called CellDMC, which is able to identify not only differentially methylated positions, but also the specific cell-type(s) driving the differential methylation. We provide extensive validation of CellDMC on in-silico mixtures of DNA methylation data generated with different technologies, as well as on real mixtures from epigenome-wide-association and cancer epigenome studies. We demonstrate how CellDMC can achieve over 90% sensitivity and specificity in scenarios where current state-of-the-art methods fail to identify differential methylation. By applying CellDMC to a smoking EWAS performed in buccal swabs, we identify differentially methylated positions occurring in the epithelial compartment, which we validate in smoking-related lung cancer. CellDMC may help towards the identification of causal DNA methylation alterations in disease.

show abstract

Section: Resultsmentioning

confidence: 99%

Identification of differentially methylated cell-types in Epigenome-Wide Association Studies

Breeze

Beck

Teschendorff

2018

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although highly consistent and expected given the analogous observations made at the bulk level (see e.g. 37, 51 ), it is not automatic that bulk and single-cell results would agree so well given the confounding effects of cell-type heterogeneity in bulk samples, specially in a tissue like lung where over 40% of cells are stromal cells 60 . Many of the lung-specific TFs, which SCIRA predicts to be inactivated in lung tumor epithelial cells (e.g.…”

Section: Discussionmentioning

confidence: 68%

“…Importantly, a second limma analysis is performed by comparing the tissue of interest to individual tissue types if these contain cells that are believed to significantly infiltrate and contaminate the tissue of interest. Thus, in the case of liver we perform two limma analyses: comparing liver to all other tissue-types, and separately, liver to only blood, since blood contains immune cells which are known to infiltrate liver tissue accounting for approximately 40% of all cells found in liver 60 . We require a liver-specific TF to be one with significantly higher expression in both comparisons.…”

Section: Methodsmentioning

confidence: 99%

Leveraging high-powered RNA-Seq datasets to improve inference of regulatory activity in single-cell RNA-Seq data

Wang¹,

Teschendorff

2019

Preprint

View full text Add to dashboard Cite

9Inferring the activity of transcription factors in single cells is a key task to improve our 2 0 4 1 relationships function at the single-cell level and therefore their implications can only be truly 4 2 understood at single-cell resolution 3-8 . 4 3 Recent single-cell RNA-sequencing (scRNA-Seq) studies have significantly advanced our 4 4 understanding of cellular development and disease 4, 7, 9-29 , yet how best to infer regulatory 4 5 activity at the single-cell level is still unclear. While a number of methods for inferring 4 6 regulatory activity have appeared 30-32 , all attempt to do so by first inferring the regulatory 4 7interactions from the noisy scRNA-Seq data itself, an approach recently assessed by others to 4 8 be highly suboptimal 33 , mainly because of the relatively high dropout rate of such data 5, 34 . 4 9 7 1 Motivation and rationale for the SCIRA algorithm 7 2

show abstract

“…The majority of our samples are from blood, and we observed a significant improvement in the prediction results for the samples from saliva when more blood samples were included in the training set (Figure 1, Supplementary Figure 1) . This increase is expected since samples from saliva were reported to exhibit more than 80% contamination by immune cells 27 . To quantify whether our predictor has a good performance in non-blood tissues, we downloaded 13 data sets (Supplementary Table 4) that contain samples from other tissues.…”

Section: Resultsmentioning

confidence: 94%

Improved prediction of chronological age from DNA methylation limits it as a biomarker of ageing

Zhang

et al. 2018

Preprint

View full text Add to dashboard Cite

DNA methylation is associated with age. The deviation of age predicted from DNA methylation from actual age has been proposed as a biomarker for ageing. However, a better prediction of chronological age implies less opportunity for biological age. Here we used 13,661 samples (from blood and saliva) in the age range of 2 to 104 years from 14 cohorts measured on Illumina HumanMethylation450/EPIC arrays to perform prediction analyses. We show that increasing the sample size achieves a smaller prediction error and higher correlations in test datasets. We demonstrate that smaller prediction errors provide a limit to how much variation in biological ageing can be captured by methylation and provide evidence that age predictors from small samples are prone to confounding by cell composition. Our predictor shows a similar or better performance in non-blood tissues including saliva, endometrium, breast, liver, adipose and muscle, compared with Horvath’s across-tissue age predictor.

show abstract

A Novel Cell-Type Deconvolution Algorithm Reveals Substantial Contamination by Immune Cells in Saliva, Buccal and Cervix

Abstract: The degree and variation of IC contamination in complex epithelial tissues is substantial. We provide a valuable resource and tool for assessing the epithelial purity and IC contamination of samples and for identifying differential methylation in such complex tissues.

Cited by 144 publications

References 51 publications

Identification of differentially methylated cell-types in Epigenome-Wide Association Studies

Identification of differentially methylated cell-types in Epigenome-Wide Association Studies

Leveraging high-powered RNA-Seq datasets to improve inference of regulatory activity in single-cell RNA-Seq data

Improved prediction of chronological age from DNA methylation limits it as a biomarker of ageing

Contact Info

Product

Resources

About