The Aging Proteome

# The Aging Proteome
## Is Aging Programmed in our Genes?<br><br><br>
### Petr Simecek
### 2017/02/08

---

background-image: url("https://pbs.twimg.com/media/BelWk8HIQAAbd3P.jpg:large")
background-position: 50% 50%

---
class: inverse, center, middle

# Kidney & Heart DO Datasets

### [simecek.github.io/TheAgingProteome](https://simecek.github.io/TheAgingProteome)

---

# Overview

- Male and females, 6m / 12m / 18m old (32 mice per group)

- 187 / 188 animals, mRNA and protein expression

- DOQTL analyses and R/Shiny viewers:

- [heart](http://churchilldev.jax.org:49194/heart/) 
  - [kidney](http://churchilldev.jax.org:49194/kidney/) 
  - [others](http://churchill-lab.jax.org/website/shinytools)

- qtl2 analyses and Flask/plumber/D3 viewers:

- [kidney](http://churchill-lab.jax.org/qtl/kidney/DO188/)

- Results stored in [my Dropbox](https://www.dropbox.com/sh/m9tomk47gmeeu6w/AAB24166oefnqsoIVhW57TiGa?dl=0)

- Github Page (workflowr): [simecek.github.io/TheAgingProteome](https://simecek.github.io/TheAgingProteome)

---

# ANOVA Tests

For each gene/protein, we have fitted a linear model

$$ y ∼ Age+Sex+Generation $$

and then tested the significance of `$Age$` / `$Sex$` / `$Age:Sex$` interaction.

---

# Sex Effect is Weaker for Heart

![whatever](http://simecek.github.io/TheAgingProteome/figure/pval-histograms.Rmd/mrna_sex-1.png)
class: center, middle

---

# Sex Effect is Weaker for Heart (cont'd)

![whatever](https://simecek.github.io/TheAgingProteome/figure/pval-histograms.Rmd/protein_sex-1.png)
class: center, middle

---

# Age Effect is Weaker for mRNA

![whatever](https://simecek.github.io/TheAgingProteome/figure/pval-histograms.Rmd/mrna_age-1.png)
class: center, middle

---

# Age Effect is Stronger for Proteins

![whatever](https://simecek.github.io/TheAgingProteome/figure/pval-histograms.Rmd/protein_age-1.png)
class: center, middle

---

# PCA plots: Kidney / mRNA

---

# PCA plots: Kidney / Protein

---

# PCA plots: Heart / mRNA

---

# PCA plots: Heart / Protein

---

# Mediation Plots - Theory

For each protein, we have fitted ANOVA model

$$ Protein \sim Age + Sex+Generation$$

and a model with a corresponding RNA expression included in covariates

$$ Protein \sim Age + Sex+Generation+mRNA$$

Let us assume Age/Sex effect is highly significant in the first model (p-value close to zero). If the effect is mediated by mRNA expression (Age/Sex -> mRNA -> Protein) then the dependence in the second model should be much weaker (i.e. p-value higher).

If Age/Sex effect skips mRNA-level and influence protein expression directly, i.e. the effect is not mediated by mRNA, then we expect p-value not to change (or even decrease if strong mRNA-Protein relationship explains part of noise).

---

# Mediation Plots - Kidney

---

# Mediation Plots - Heart

---

# Example - Fmo5

<center>
flavin containing monooxygenase 5

---

# Example - Ace

<center>
angiotensin I converting enzyme (peptidyl-dipeptidase A) 1

---

# Example - Acta2

<center>
actin, alpha 2, smooth muscle, aorta

---
class: inverse, center, middle

# Workflowr package

### http://github.com/jdblischak/workflowr

---
background-image: url("http://tub.tubgit.com/images251/uv02t5vgx3a.jpg")
background-size: cover
class: center, bottom, inverse

---
background-image: url("https://github.com/yihui/xaringan/releases/download/v0.0.2/karl-moustache.jpg")
background-size: cover
class: center, bottom, inverse

# And now something completely different!

---

# Mediation

### http://github.com/simecek/intermediate

---

# What is mediation?

1) `$X \rightarrow Y$` 
 (i.e. `$X \not\mkern-5mu{\perp\mkern-10mu\perp} Y$`)

2) `$M \rightarrow Y$` 
 (i.e. `$M \not\mkern-5mu{\perp\mkern-10mu\perp} Y$`)
 
 3) `$X \rightarrow M \rightarrow Y$` 
 (i.e. `$X {\perp\mkern-10mu\perp} Y | M$`)
 
 4) But not `$X \rightarrow Y \rightarrow M$`
 (i.e. `$X \not\mkern-5mu{\perp\mkern-10mu\perp} M | Y$`)

---

# Mediation issues

1) Shared QTL vs. True mediation

2) LOD drop threshold, the minimizer is not necessarily the true mediator
 
 3) `$M \leftarrow Q \rightarrow Y$` vs. `$Q \rightarrow M \rightarrow Y$`

---
background-image: url("http://img0.etsystatic.com/069/1/8631557/il_fullxfull.827661664_2oto.jpg")
background-size: cover
class: center, top

# Thank you for your attention!