class: title-slide .header[ &nbsp; &nbsp; <img src="bristol-logo.png" style="width: 200px"></img> <img src="ieu-logo.png" style="width: 200px"></img> ] # Epigenetic Epidemiology Update .large[**Matthew Suderman**] .large[**Oct 3, 2022**] --- layout: true .footer[MRC Integrative Epidemiology Unit] --- ## EWAS of disease .striped[ | pmid|journal |variable |tissue |population |result | |--------:|:-------------------|:---------------------------------------|:--------------------|:----------------------------------|:--------------------| | 36170668|Diabetes |type 2 diabetes; HbA1c; fasting glucose |blood |1070 twins (China) |67;17;16 sites | | 36142605|Int J Mol Sci |lung cancer |pre-diagnostic blood |66 case-control pairs from EPIC HD |1106 sites | | 36109771|Alzheimers Res Ther |Alzheimer's disease |blood |1284 adults |2 sites (in females) | ] --- ## EWAS of exposure .striped[ | pmid|journal |variable |tissue |population |result | |--------:|:-----------|:-----------------------|:-------------------------------------------|:------------------------|:-------------------------------------------| | 36148884|Epigenetics |prenatal lead |maternal toe nail; infant toenail; placenta |>115 mothers and infants |480;27;2 sites | | 36127421|Nat Commun |dexamethasone treatment |blood |135 glioma patients |20355 sites; 2621 by CellDMC in neutrophils | ] --- .running[enhancers] ## Redundancy to mutation is built into enhancers Lin X ... Qi LS. **Nested epistasis enhancer networks for robust genome regulation.** *Science* . doi: [10.1126/science.abk3512](http://doi.org/10.1126/science.abk3512) -- .pull-left-70[ <img src="enhancers-fig1a.jpg" style="width: 100%"></img> ] -- ### Background *MYC* is regulated by **seven enhancers** split into **two clusters** 1.5Mb apart. -- ### Question Which **pair** of enhancers has the biggest effect on *MYC* expression? -- ### Approach Use **multiplexed CRISPRi** to knock out enhancer pairs. --- .running[enhancers] ### Effects of enhancer pairs on *MYC* expression .pull-left-70[ <img src="enhancers-fig1b.jpg" style="width: 100%"></img> ] -- <br> <br> <br> <br> <br> Heatmap shows **epistasis interaction scores** for all enhancer pairs. -- **epistasis interaction score** ~ effect of mutations in both enhancers on *MYC* expression -- Notice how effect **increases** with pair distance. --- .running[enhancers] ### Two-layer model hypothesis 1. **Layer I** is additive, i.e. expression decreases proportionally to the number of mutated enhancers -- 2. **Layer II** is synergistic, i.e. mutations must occur in both enhancer clusters to 'knock out' *MYC* -- .center[ <img src="enhancers-fig1d.jpg" style="width: 80%"></img> ] --- .running[enhancers] ### Identifying a mechanism for two layers Answer using **machine learning** (penalised regression): -- - **features**: co-occurence of transcription factors, histone modifications and DNA-DNA spatial contacts (HiChIP) at enhancer pairs (from public datasets) -- - **outcome**: 'epistatis interaction score' for enhancer pairs -- .center[ <img src="enhancers-fig2.jpg" style="width: 100%"></img> ] --- .running[enhancers] ### Hypothesis based on machine learning model A possible explanation for **synergy** between distant enhancers: -- - one mutation reduces spatial contacts and *BRD4* colocalisation with *MYC* promoter -- - one mutation *per* cluster required to prevent contacts and *BRD4* -- .center[ <img src="enhancers-fig4.jpg" style="width: 100%"></img> ] --- .running[enhancers] ### Alternative hypothesis > Elkon R, Agami R. **Two-layer design protects genes from mutations in their enhancers.** *Nature* . doi: [10.1038/d41586-022-02341-3](http://doi.org/10.1038/d41586-022-02341-3) -- The *MYC* promoter _switches_ between activating clusters. -- .center[ <img src="enhancers-alt-fig1.png" style="width: 80%"></img> ] --- .running[enhancers] .pull-left-50[ ### Question Are enhancer pairs with high 'epistatis interaction scores' hotspots for disease-associated genetic variants? ] -- .pull-left-50[ ### Method Machine learning model applied to GM12878 data to predict 'epistatis interaction scores' (SRE) for enhancer pairs. ] -- .pull-left-30[ <br><br><br> Pairs with high prediction scores more likely to identify genetic variants associated with ALL relapse risk. ] .pull-right-70[ <img src="enhancers-fig6.jpg" style="width: 90%"></img> ] --- .running[enhancers] ### Implications for omic studies? -- - **Enrichment** Associations should be enriched in enhancers with high 'epistatis interaction scores' -- - **Imputation** If trait association observed in one enhancer, then enhancers that interact should also have associations -- - **Cell-type specificity** Cell-type-specific interaction scores can be calculated from cell type reference data (e.g. Blueprint). Trait association patterns could be tested for enrichment relative to cell-type-specific scores to hypothesize most relevant cell type. --- .running[review: big data] Jiang P ... Ruppin E. **Big data in basic and translational cancer research.** *Nat Rev Cancer* . doi: [10.1038/s41568-022-00502-0](http://doi.org/10.1038/s41568-022-00502-0) -- * Common data types, e.g. gene expression, DNA methylation -- * Data repositories and platforms, e.g. TCGA, GEO -- * Integrative analysis - across modalities, e.g. gene expression with DNA methylation - across cohorts - across resolutions, e.g. single-cell analysis uses bulk for 'zeros' -- * Data for translational solutions - biomarkers in use, e.g. Oncotype DX for breast cancer - clinical studies using 'big data' - drug repurposing -- * Challenges - data availability - data-scale, e.g. ethnicity, rare diseases --- .running[review: psychiatric genetics] Derks EM, Thorp JG, Gerring ZF. **Ten challenges for clinical translation in psychiatric genetics.** *Nat Genet* . doi: [10.1038/s41588-022-01174-0](http://doi.org/10.1038/s41588-022-01174-0) <!-- Ten challenges for clinical translation in psychiatric genetics.--> -- <br> <br> <img src="psychiatric-fig1.png" style="width: 100%"></img> Possibly missing reducing diagnosis noise? --- .running[review: lung cancer prediction] Li P ... Wang C. **Liquid biopsies based on DNA methylation as biomarkers for the detection and prognosis of lung cancer.** *Clin Epigenetics* . doi: [10.1186/s13148-022-01337-0](http://doi.org/10.1186/s13148-022-01337-0) -- <br> <br> Review of -- * various liquid biopsies available -- * how DNA methylation is/could be measured -- * what is known relative to lung cancer detection and prognosis -- *A bit boring but good place to look up "what's known".* --- .running[dexamethasone exposure score] ## DNAm score for glucocorticoid exposure Wiencke JK ... Kelsey KT. **DNA methylation as a pharmacodynamic marker of glucocorticoid response and glioma survival.** *Nat Commun* . doi: [10.1038/s41467-022-33215-x](http://doi.org/10.1038/s41467-022-33215-x) .pull-left-60[ <div style="padding: 5px"> <img src="gluco-fig1.png" style="width:100%"></img> </div> ] -- <br> **Dexamethasone** is a glucocorticoid medication for a wide variety of diseases and health conditions. It has **anti-inflammatory** and **immunosuppressant** effects. -- <span style="color:#dfb286; font-weight: bold"> CellDMC identifies 2621 neutrophil-specific associations, only one in other cell types. </span> -- <span style="color:#b8d786; font-weight: bold"> Elastic net was applied to neutrophil-specific sites to construct the neutrophil dexamethasone methylation index (NDMI). </span> --- .running[dexamethasone exposure score] ### NDMI performance in independent data .center[ <img src="gluco-fig5.png" style="width:100%"></img> ] --- .running[dexamethasone exposure score] ### NDMI performance purified cell types **Note:** data generated for subset of training population .center[ <img src="gluco-fig6.png" style="width:60%"></img> ] --- .running[dexamethasone exposure score] ### NDMI and glioma survival .center[ <img src="gluco-fig7.png" style="width:100%"></img> ] **Note:** associations remain after adjusting for dexamethasone exposure, cell counts, age, etc. --- ## Announcements * [Epigenetic Epidemiology short course](https://www.bristol.ac.uk/medical-school/study/short-courses/2021-22-courses/epigenetic-epidemiology/) 8 - 10 May 2023 * [Advanced Epigenetic Epidemiology short course](https://www.bristol.ac.uk/medical-school/study/short-courses/2021-22-courses/epigenetic-epidemiology/) 18 - 19 May 2023