---
title: "Example of placement analysis using BoSSA"
output:
  prettydoc::html_pretty:
    theme: architect
    highlight: github
vignette: >
  %\VignetteIndexEntry{Example of placement analysis using BoSSA}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}  
---
<!--
%% \VignetteEngine{knitr::knitr}
-->

```{r style, echo = FALSE, results = 'asis'}
```

Please report comments or bugs to Pierre Lefeuvre - <pierre.lefeuvre@cirad.fr>

[BoSSA CRAN page](https://cran.r-project.org/package=BoSSA)

# Summary

A phylogenetic placement corresponds to the position of a query sequence in a reference tree. Different tools exits to infer phylogenetic placements, such as [pplacer](https://matsen.fhcrc.org/pplacer/), [EPA](http://sco.h-its.org/exelixis/web/software/epa/index.html) or [RAPPAS](https://github.com/phylo42/RAPPAS). Importantly, these three programs produce placements under a common [file format](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0031009). Placements can later be analysed using the [guppy](https://matsen.github.io/pplacer/generated_rst/guppy.html) software from the pplacer suite to obtain statistically based taxonomic classification of sequences. The BoSSA package implements functions to reads, plots and summarizes phylogentic placements. This vignette is intended to provide examples of placements analyses using BoSSA.

# Important note

- The placement mass (potentially) available in the jplace and sqlite files are imported in R (within the jplace and pplace objects) but aren't use in the analysis. You should use the "N" parameter (available in several fucntions for the package) to use different weight for each placement.
- The reference packages shiped with BoSSA are incomplete (they lack the alignment file) in order to reduce the package size. Whereas, the information available is sufficient to draw summary statistics, it won't be enough to perform actual phylogenetic placement.
- When the jplace or sqlite files are import into R, the node numbering available in the original file is converted to the class "phylo" numbering.

# How to obtain phylogentic placement file suitable for analysis with BoSSA ?

The process to obtain placement files is dependent of the program you use. Assuming you are using pplacer, the process would be (1) build a reference package that contains an align set of reference sequences and a reference phylogenetic tree, (2) align query sequences to the reference alignment, (3) use pplacer to infer placements (jplace file output, format describe [here](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0031009)) and optionally (4) infer the classification of each sequences using guppy (sqlite file output).

- The construction of the reference package could be a bit tricky. The [taxtastic](https://github.com/fhcrc/taxtastic) tool is extremely helpfull to this end. A tutorial can be find [here](http://fhcrc.github.io/taxtastic/quickstart.html). A BoSSA vignette on refpkg construction is also available.
- [HMMER](http://hmmer.org/) and [MAFFT](https://mafft.cbrc.jp/alignment/software/addsequences.html) can be use to align sequences to a reference alignment.
- For phylogenetic placement, a detailed tutorial is available [here](http://fhcrc.github.io/microbiome-demo/).

Let's say you have obtained a reference package (refpkg), a placement file (jplace file) and a guppy classification output (sqlite file). The example files presented here are derived from other [reference packages and jplaces files](https://github.com/fhcrc/microbiome-demo/zipball/master) from the [Matsen group pplacer tutorials](https://fhcrc.github.io/microbiome-demo/). The sqlite file was obtained using the following command:

```
guppy classify --multiclass-min 0 --cutoff 0.5 -c example.refpkg --sqlite example.sqlite example.jplace
```

# Exploration of a reference package

Let's start by loading the `BoSSA-package:

```{r load-packages, message=FALSE, warning=FALSE}
library("BoSSA")
```

A good practice would be to inspect the refpkg content.

```{r }
refpkg_path <- paste(find.package("BoSSA"),"/extdata/example.refpkg",sep="")
refpkg(refpkg_path)
```
It is possible to extract the taxonomy of the sequences included in the refpkg.

```{r }
taxo <- refpkg(refpkg_path,type="taxonomy")
head(taxo)
```
or display a pie chart that summarize the taxonomy...

```{r pie1, fig.width=5, fig.height=5}
refpkg(refpkg_path,type="pie",cex.text=0.5)
```

... or a subset of the taxonomy levels. Here, an example with the "class", "order" and "family" levels. Note there is a slight decay between the text labels and slices... this will need a fix in a future package update.

```{r pie2, fig.width=5, fig.height=5}
refpkg(refpkg_path,type="pie",rank_pie=c("class","order","family"),cex.text=0.6)
```

Finally, a tree display with branch colored according to a given taxonomic level is available. Here tips are colored according to the "order" classification.

```{r refseqtree, fig.width=8, fig.height=8}
refpkg(refpkg_path,type="tree",rank_tree="class",cex.text=0.5)
```

# Loading the example data

The BoSSA package comes along with examples of phylogenetic placements from the Masten group.

```{r}
sqlite_file <- system.file("extdata", "example.sqlite", package = "BoSSA")
jplace_file <- system.file("extdata", "example.jplace", package = "BoSSA")
```

To read the data, use the `read_sqlite` function.

```{r}
pplace <- read_sqlite(sqlite_file,jplace_file)
pplace
```

A summary of the object is printed with the number of runs, the command line, a short description of the phylogenetic tree, the number of placements and the number of sequences being placed. Pplace objects are stored in a list of 15 components, with 12 components being outputs from a [guppy classify](https://matsen.github.io/pplacer/generated_rst/guppy_classify.html#guppy-classify) run and 3 components corresponding to the phylogenetic tree used for placement:

```{r }
str(pplace)
```

Among these:

- the `run` element contains the run id and the command line summary

- the `taxa` element is a data frame with the whole taxonomy available in the reference package 

- the `multiclass` element is a data frame with the taxonomic assignation of each placement

- the `placement_positions` element is a data frame with the position of each placement over the reference phylogenetic tree

- the `arbre` element is the class `phylo` object of the reference phylogenetic tree

# Some plots

Four different plots are available to display placements on a phylogenetic tree:

- the `number` plot. Placement number associated to each branch is indicated. Note that this representation may be hard to read due to overlaps between number boxes. Placement numbers are obtained after the multiplication of their weights with the ML ratio of the placement probabilities. Placement sizes are later round. A zero indicates a size superior to 0 but inferior to 1.

```{r test1, fig.width=9, fig.height=9}
plot(pplace,type="number",main="number",cex.number=1.5)
```

- the `color` plot is the best option. Branches with placement are colored according to the number of sequences they bear.

```{r test2, fig.width=9,fig.height=9}
plot(pplace,type="color",main="color",edge.width=2)
```

- in the `fattree` plot, branch wicth is proportionnal to the number of sequences they bear.

```{r testfat, fig.width=9,fig.height=9}
plot(pplace,type="fattree",main="fattree")
```

- in the `precise` plot dots are drawn at the exact placement positions. Whereas the color of the dots depend of the pendant branch length, their sizes depend on the placement sizes. Note that placements are drawn one above the other.

```{r test3, fig.width=9,fig.height=9}
plot(pplace,type="precise",main="precise")
```

Note that it is possible to apply a function to modify the dot size using the `transfo` option. In the following example, the dot size is multiplied by 2. In some other cases `log` or `log10` transformations could be usefull. Beware that when using the `transfo` option, the legend does not anymore correspond to the placement size but to the transform dot size (*i.e.* the transform function applied to the dot size).

```{r test4, fig.width=9,fig.height=9}
plot(pplace,type="precise",main="precise",transfo=function(X){X*2})
```

# Subsetting the pplace object

Placement object can be subseted. This could be done using placements ids...

```{r }
sub1 <- sub_pplace(pplace,placement_id=1:100)
sub1
```

...or using placements names.

```{r }
ids <- sample(pplace$multiclass$name,50)
sub2 <- sub_pplace(pplace,ech_id=ids)
sub2
```

# Conversion

### To a table

Using the `pplace_to_table` function produces a table that contains the placement information along with the classification for each sequence. The output can be limited to the "best" placement (as in the example, i.e. the placements with the highest likelihood for each sequence).

```{r }
pplace_table <- pplace_to_table(pplace,type="best")
head(pplace_table,n=3)
```

### To a contingency matrix

The `pplace_to_matrix` produces a contingency table. Let say the first 50 sequences in the multiclass table correspond to sequence from "sample 1" and the following 50 correspond to "sample 2", the function output a contingency table for these two samples. You can either have the taxonomic names (tax_name=TRUE, in the example) or keep the taxonomic ids (tax_name=FALSE).

```{r}
example_contingency <- pplace_to_matrix(pplace,c(rep("sample1",50),rep("sample2",50)),tax_name=TRUE)
example_contingency
```

### To a taxonomy

Using the `pplace_to_taxonomy` function, a taxonomy table is obtained for each sequences with the taxonomy levels defined in the reference package. The taxonomy levels can be limited to a set of levels using the `rank` option.

```{r}
example_taxo <- pplace_to_taxonomy(pplace,taxo,tax_name=TRUE,rank=c("order","family","genus","species"))
head(example_taxo)
```

### Make a phyloseq object

Assuming the sequences in the pplace object represent centroids of sequence cluster obtained from multiple samples, using the taxonomy table and an appropriate OTU file, you can create a phyloseq object.

```{r}
example_OTU <- matrix(sample(1:100, 500, replace = TRUE), nrow = 100, ncol = 5,dimnames=list(pplace$multiclass$name,paste("sample",1:5,sep="_")))
head(example_OTU)
```
The exemple below is not run (commented) due to errors/warnings triggered by the used of Bioconductor packages (i.e. phyloseq) in CRAN vignette on some platform. Just uncomment the code if you like to have a try.

```{r}
#library(phyloseq)
#example_phyloseq <- phyloseq(otu_table(example_OTU,taxa_are_rows=TRUE),tax_table(example_taxo))
#example_phyloseq
```

# Citation

If you find BoSSA and/or its tutorials useful, you may cite:

```{r}
citation("BoSSA")
```

# Other resources

### On phylogenetic placements
[pplacer website](https://matsen.fhcrc.org/pplacer/) and [documentation](http://matsen.github.io/pplacer/)

[taxtastic](https://github.com/fhcrc/taxtastic)

### Other R package with a related topic
[ggtree](https://bioconductor.org/packages/release/bioc/html/ggtree.html) and [clstutils](https://bioconductor.org/packages/release/bioc/html/clstutils.html)

### R packages used by BoSSA
[RSQLite](https://cran.r-project.org/package=RSQLite) and [jsonlite](https://cran.r-project.org/package=jsonlite) to read files, [ape](https://cran.r-project.org/package=ape) and [phangorn](https://cran.r-project.org/package=phangorn) to manipulate phylogenetic trees and [plotrix](https://cran.r-project.org/package=plotrix) for pie charts.