CoAbundantTaxa.Rmd

---
title: "Identify Co-abundant taxa groups"
author: "Leo Lahti, Sudarshan Shetty et al."
bibliography: 
- bibliography.bib
output:
  BiocStyle::html_document:
    number_sections: no
    toc: yes
    toc_depth: 4
    toc_float: true
    self_contained: true
    thumbnails: true
    lightbox: true
    gallery: true
    use_bookdown: false
    highlight: haddock
---
<!--
  %\VignetteIndexEntry{microbiome tutorial}
  %\VignetteEngine{knitr::rmarkdown}
  %\usepackage[utf8]{inputenc}
  %\VignetteEncoding{UTF-8}
-->

**Co-abundant groups of taxa**    

Get data  
```{r cag-1}
library(microbiome)
pseq <- atlas1006
pseq.rel <- microbiome::transform(pseq, "compositional")
```

Identify co-abundant taxa. 
```{r cag-2}
# get median abundances 
median_abund = apply(abundances(pseq.rel), MARGIN = 1, FUN = median)

# select taxa with median abundance of more than or equal to 0.01
abundant_taxa = abundances(pseq)[median_abund >= 0.01, ]
# check how many taxa
nrow(abundant_taxa)
# check which taxa
rownames(abundant_taxa)
```


Use Spearman's correlation to identify which of the co-abundant groups show correlation.   
```{r cag-1b}
# cor() is from `stats` package
spearman_matrix = cor(t(abundant_taxa), method = "spearman")
# hclust() is from `stats` package
spearman_tree = hclust(dist(spearman_matrix), method = "ward")
plot(spearman_tree)

```

Using the `heat` function in `microbiome pkg` now compare the correlation with the tree above.  
```{r plot-heat}
spearman_long_df <- reshape2::melt(spearman_matrix)
head(spearman_long_df)

heat(spearman_long_df, "Var1","Var2", 
     order.rows = TRUE, order.cols = TRUE) + 
  theme_bw() + theme(axis.text.x = element_text(angle = 90))
```

Validation of CAGs   
```{r}
# Validation of CAGs 
# Randomly split the dataset, compute a correlation matrix and run a Mantel test
subsample = sample(c(1:300), size = 220, replace = F)
train_data = abundant_taxa[,subsample]
test_data = abundant_taxa[,-subsample]
train_spearman = cor(t(train_data), method = "spearman")
test_spearman = cor(t(test_data), method = "spearman")
vegan::mantel(as.dist(train_spearman), as.dist(test_spearman), permutations=999)
```

Further reading:  

[de la Cuesta-Zuluaga, J., Corrales-Agudelo, V., Velásquez-Mejía, E.P. et al. Gut microbiota is associated with obesity and cardiometabolic disease in a population in the midst of Westernization. Sci Rep 8, 11356 (2018).](https://doi.org/10.1038/s41598-018-29687-x)   

[Claesson, M., Jeffery, I., Conde, S. et al. Gut microbiota composition correlates with diet and health in the elderly. Nature 488, 178–184 (2012).](https://doi.org/10.1038/nature11319)