---
title: "Filtering and Subsetting"
output: rmarkdown::html_vignette
description: |
  This vignette provides a comprehensive guide to using the `filter_taxa` and 
  `subset_taxa` functions, as well as their convenience wrappers.  
vignette: >
  %\VignetteIndexEntry{Filtering and Subsetting}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

<style>
body {
text-align: justify}
</style>

```{r, include = FALSE}
knitr::opts_chunk$set(
  message = FALSE,
  digits = 3,
  collapse = TRUE,
  comment = "#>"
)
options(digits = 3)
```

The `step_filter_taxa` function is a general function that allows for flexible 
filtering of OTUs based on across-sample abundance criteria. The other 
functions, `step_filter_by_prevalence`, `step_filter_by_variance`, 
`step_filter_by_abundance`, and `step_filter_by_rarity`, are convenience 
wrappers around `step_filter_taxa`, each designed to filter OTUs based on a 
specific criterion: prevalence, variance, abundance, and rarity, respectively.

The `step_subset_taxa` function is used to subset taxa based on their taxonomic 
level.

The phyloseq or TSE used as input can be pre-filtered using methods that are 
most convenient to the user. However, the `dar` package provides several 
functions to perform this filtering directly on the recipe object.

## step_filter_taxa

The `step_filter_taxa` function applies an arbitrary set of functions to OTUs as 
across-sample criteria. It takes a phyloseq object as input and returns a 
logical vector indicating whether each OTU passed the criteria. If the "prune" 
option is set to FALSE, it returns the already-trimmed version of the phyloseq 
object.

```{r step_filter_taxa}
library(dar)
data("metaHIV_phy")

rec <- recipe(metaHIV_phy, "RiskGroup2", "Species")
rec <- 
  step_filter_taxa(rec, .f = "function(x) sum(x > 0) >= (0 * length(x))") |> 
  prep()
```

## Convenience Wrappers

### step_filter_by_abundance

This function filters OTUs based on their abundance. The taxa retained in the
dataset are those where the sum of their abundance is greater than the product 
of the total abundance and the provided threshold.

```{r step_filter_by_abundance}
rec <- recipe(metaHIV_phy, "RiskGroup2", "Species")
rec <- 
  step_filter_by_abundance(rec, threshold = 0.01) |> 
  prep()
```

### step_filter_by_prevalence

This function filters OTUs based on their prevalence. The taxa retained in the
dataset are those where the prevalence is greater than the provided threshold.

```{r step_filter_by_prevalence}
rec <- recipe(metaHIV_phy, "RiskGroup2", "Species")
rec <- 
  step_filter_by_prevalence(rec, threshold = 0.01) |> 
  prep()
```

### step_filter_by_rarity

This function filters OTUs based on their rarity. The taxa retained in the
dataset are those where the sum of their rarity is less than the provided 
threshold.

```{r step_filter_by_rarity}
rec <- recipe(metaHIV_phy, "RiskGroup2", "Species")
rec <- 
  step_filter_by_rarity(rec, threshold = 0.01) |> 
  prep()
```

### step_filter_by_variance

This function filters OTUs based on their variance. The taxa retained in the 
dataset are those where the variance of their abundance is greater than the 
provided threshold.

```{r step_filter_by_variance}
rec <- recipe(metaHIV_phy, "RiskGroup2", "Species")
rec <- 
  step_filter_by_variance(rec, threshold = 0.01) |> 
  prep()
```

## subset_taxa

The `subset_taxa` function subsets taxa based on their taxonomic level. The 
taxa retained in the dataset are those where the taxonomic level matches the 
provided taxa.

```{r subset_taxa}
rec <- recipe(metaHIV_phy, "RiskGroup2", "Species")
rec <-
  step_subset_taxa(rec, tax_level = "Kingdom", taxa = c("Bacteria", "Archaea")) |>
  prep()
```

## Conclusion

These functions provide a powerful and flexible way to filter and subset OTUs 
in phyloseq objects contained within a recipe object, making it easier 
to work with complex experimental data. By understanding how to use these 
functions effectively, you can streamline your data analysis workflow and focus 
on the aspects of your data that are most relevant to your research questions. 
The `dar` package offers the added convenience of performing these operations 
directly on the recipe object.

## Session info

```{r}
devtools::session_info()
```