count: false class: left, bottom # Journal club ## December 14, 2020   <img src="IEU-logo-colour.png" width="50%"> --- layout: true .logo[.mrcieu[ MRC Integrative Epidemiology Unit     ]] --- ## EWAS in infants and children .striped[ |variable |tissue |n |assocs |notes | pmid|journal | |:----------------------------------------------|:----------------------|:-----------------------|:-------------------------------------|:-----------------------------------------------------|--------:|:------------| |Maternal first-trimester and delivery lipodome |Cord blood |100 |45 and 18 with two lipids at delivery | | 33290095|Epigenomics | |ART |cord blood |87 cases vs 70 controls |19 |no difference between IVF and ICSI | 33227132|Hum Reprod | |lung function trajectories up to age 26 |heel prick blood |786 |8 | | 33214203|Eur Respir J | |BMI |whole blood (age 2-18) |4133 |3 |strength of adult BMI associations increased with age | 33239103|Genome Med | |Maternal care (sensitivity) |whole blood at age 6 |235 |4 DMRs | | 33267929|Psychol Med | ] --- ## EWAS of mental health .striped[ |variable |tissue |n |assocs |notes | pmid|journal | |:-------------------------------|:-----------------|:---------------------------------------------------------------|:---------------------------|:-------------------------------------------------------------|--------:|:----------------| |Alzheimer's disease Braak stage |prefrontal cortex |1000 |3751 | | 33257653|Nat Commun | |Alzheimer's disease |whole blood |424 at baseline age > 65 (87 AD, 175 mild cognitive impairment) |MCI hypomethylated at 1 DMR |Reverts to hypermethylation as AD progresses | 33298155|Clin Epigenetics | |First-episode schizophrenia |whole blood |469 cases vs 476 controls |4277 |not confounded by long-term medication or disease progression | 33279932|Mol Psychiatry | |PTSD |whole blood |1896 |4 sites in AHRR |associations stronger in non-smokers | 33235198|Nat Commun | |Panic disorder |whole blood |133 cases vs 118 controls |61 | | 33208194|Clin Epigenetics | ] --- ## EWAS of exposure .striped[ |variable |tissue |n |assocs |notes | pmid|journal | |:----------------|:---------------------------|:--------------------------------|:--------|:--------------------------------------------|--------:|:--------------------| |Alcohol intake |whole blood |1004 (avg age 23) |24 |4 replicated in discordant twin analyses | 33277923|Alcohol Clin Exp Res | |Childhood hunger |whole blood in older adults |2221 |0 |hunger associated with CVD and depression | 33215951|Epigenomics | |Vegans |blood (buffy coat) |62 vegans vs 142 non-vegetarians |18 genes |Average methylation calculated for each gene | 33266012|Nutrients | |HIV |whole blood |397 |49 | | 33232214|Epigenomics | ] --- ## EWAS of phenotypes .striped[ |variable |tissue |n |assocs |notes | pmid|journal | |:----------------------------------|:--------------------|:------------------------|:-----------------|:---------------------------------------------------------------------------------------|--------:|:-------------------| |Lactobacilli (L-type) microbiome |cervicovaginal smear |161 L-type vs 150 O-type |173245 |AUC = 0.84 in replication data | 33228781|Clin Epigenetics | |Metabolic syndrome and traits |whole blood |1187 |33 (12 with MetS) |5 replicated | 33239708|Sci Rep | |Clinical/complete asthma remission |whole blood |72 |7/129 |1 of 7 and 8 of 129 replicated in nasal epithelium; 2 replicated in independent cohorts | 33303027|Clin Transl Allergy | ] --- ## Predicting colorectal cancer survival > Gào X, Zhang Y, Boakye D, Li X, Chang-Claude J, Hoffmeister M, Brenner H. *Whole blood DNA methylation aging markers predict colorectal cancer survival: a prospective cohort study.* Clin Epigenetics. 2020 Nov 30;12(1):184. - n=2206 colorectal cancer patients - DNAm aging in blood - 6.2-year survival. -- .box[ After adjusting for age, sex and stage: | Method | Adjusted HR (95% CI) | |:---------------|:-----------------| |AgeAccelHorvath| 1.17 (0.99, 1.38)| |AgeAccelHannum| 1.12 (0.94, 1.33)| |DNAmMRscore| 1.66 (1.32, 2.09)| |AgeAccelPheno| 1.35 (1.14, 1.59)| |AgeAccelGrim| 1.65 (1.37, 2.00)| ] --- ## X chromosome dosage Human female cells typically have two copies of the X chromosome with one copy mostly silenced. As a result, X chromosome expression levels are quite similar between females and males, who only have one copy of X. An exception is the female primordial germ cells and preimplantation embryos where *both* copies of the X chromosome. **Question:** What kind of total gene expression levels on X do we observe in these cells? -- <hr> **Answer:** The expression is dampened on each copy to match single chromosome levels. > Chitiashvili, T., Dror, I., Kim, R. et al. *Female human primordial germ cells display X-chromosome dosage compensation despite the absence of X-inactivation.* Nat Cell Biol 22, 1436–1446 (2020). --- ## Rare genetic variation and transcription factor binding sites Rare genetic variants inside transcription factor binding sites: 1. Reduce transcription factor binding to the DNA 2. Have large effects on methylation of nearby CpG sites (often *oultiers*) Does the changed transcription factor binding cause the change in methylation _or_ is the opposite? -- <hr> .ii[ **Answer:** The reduced binding causes the change in methylation because variants at more conserved positions have larger effects on methylation. > Martin-Trujillo A, Patel N, Richter F, et al. *Rare genetic variation at transcription factor binding sites modulates local DNA methylation profiles*. PLoS Genet. 2020;16(11):e1009189. ] .ii[ <img src="pgen.1009189.g003.jpg" width="90%"> ] --- ## Prediction pitfall > Schreiber J, Singh R, Bilmes J, Noble WS. *A pitfall for machine learning methods aiming to predict across cell types.* Genome Biol. 2020;21(1):282. **Aim:** Use one omic to predict/impute another (e.g. DNA methylation to predict gene expression). **Question:** Why might you want to not only train and test on different data but also on different genomic regions? -- <img src="prediction.png" width="90%"> --- ## Probe reliability We typically remove/flag probes the cross-hybridize or coincide with comment genetic variants. But we don't do the same for unreliable probes. Why not? .box[ "**Probes that were reliably measured** were more heritable and showed consistent associations with environmental exposures, gene expression, and greater cross-tissue concordance." "**Unreliable probes** were less replicable and generated an unknown volume of false negatives." > Sugden et al. *Patterns of Reliability: Assessing the Reproducibility and Integrity of DNA Methylation Measurement*. Patterns (N Y). 2020 Apr 23;1(2):100014. ] Note: Naeem et al. (https://pubmed.ncbi.nlm.nih.gov/24447442/) do identify probes that do not match whole genome bisulfite methylation levels.