05-alpha.Rmd

```{r, echo=FALSE}
library(mia)
library(miaViz)
library(dplyr)
se <- readRDS("data/se.rds")
tse <- readRDS("data/tse.rds")
```



# Alpha diversity

This section demonstrates the analysis of alpha diversity. This
quantity measures microbial diversity within each sample. 

Alpha diversity is a key quantity in a microbiome research. The [_mia_
package](https://microbiome.github.io/mia/) provides access to a wide
variety of alpha diversity indices. For more background information
and examples with various alpha diversity indices, see the [online
book](https://microbiome.github.io/OMA/microbiome-diversity.html#alpha-diversity).

There are different alpha diversity implementations that relate to:
 - unit level: general, species or ASVs? prehaps genes?
 - unit detection: precence / absence or abundances, weighing of rare units
 - is phylogenic information incorporated, i.e. how well the sample covers the phylogenetic tree?`

Popular framework to capture the different aspects of alpha diversity is [**Hill numbers**](https://doi.org/10.1111/1755-0998.13014)

Higher numbers of unique taxa, and more even abundance distributions within a
sample yield larger values for alpha diversity.

 - richness depicts the number of units / species in sample and is intuitive
 - Shannon Index accounts also for the evenness
 - Simpson Index weighs the abundant or dominating units more than Shannon Index



```{r}
# Calculates indices
tse <- estimateDiversity(tse, index = "shannon", name = names)

```

Next we can visualize Shannon index with histogram.

```{r}
# ggplot needs data.frame format as input.
# Here, colData is DataFrame, therefore it needs to be converted to data.frame
shannon_hist <- ggplot(as.data.frame(colData(tse)), 
                       aes(x = shannon)) + 
  geom_histogram(bins = 20, fill = "gray", color = "black") +
  labs(x = "Shannon index", y = "Sample frequency")

shannon_hist
```


## Visualization

Next let us compare indices between different patient status and
cohorts. Boxplot is suitable for that purpose.

```{r}
# Creates Shannon boxplot 
shannon_box <- ggplot(as.data.frame(colData(tse)),
  aes(x = patient_status, 
      y = Shannon_diversity,
      fill = cohort)) + 
  geom_boxplot() +
  theme(title = element_text(size = 12)) # makes titles smaller

```

For an alternative visualization, see examples with [scater::plotColData](https://microbiome.github.io/OMA/microbiome-diversity.html#alpha-diversity).


## Statistical testing and comparisons

To further investigate if patient status could explain the variation
of Shannon index, let's do a Wilcoxon test. This is a non-parametric
test that doesn't make specific assumptions about the distribution,
unlike popular parametric tests, such as the t test, which assumes
normally distributed observations.

Wilcoxon test can be used to estimate whether the differences between
two groups is statistically significant. 


```{r}
# Wilcoxon test, where Shannon index is the variable that we are comparing. 
# Patient status - ADHD or control - is the factor that we use for grouping. 
wilcoxon_shannon <- wilcox.test(Shannon_diversity ~ patient_status, data = colData(tse))

wilcoxon_shannon
```

For more examples, see a dedicated section on alpha diversity in the
[online book](https://microbiome.github.io/OMA/microbiome-diversity.html#alpha-diversity).

## Exercises

Add the following in the reproducible summary report.

* Estimate alpha diversity for each sample and draw a histogram. Tip:
  estimateDiversity

* Compare the results between two or more
  alpha diversity indices (visually and/or statistically).

* See [online book](https://microbiome.github.io/OMA/microbiome-diversity.html#alpha-diversity) 
for further examples.