---
title: "Classifiers methods"
bibliography: bibliography.bib    
vignette: >
    %\VignetteIndexEntry{10. Classifiers}
    %\VignetteEngine{knitr::rmarkdown}
---
  
```{r setup, include=FALSE}
knitr::opts_chunk$set(dpi = 300)
knitr::opts_chunk$set(cache = FALSE)
```
  
```{r, echo = FALSE,hide=TRUE, message=FALSE,warning=FALSE}
library(TCGAbiolinks)
```

```{r message=FALSE, warning=FALSE, include=FALSE}
library(SummarizedExperiment)
library(dplyr)
library(DT)
```

<br>

## Classifying gliomas samples with `gliomaClassifier`
<hr>

Classifying glioma samples with DNA methylation array based on:


**Ceccarelli, Michele, et al. "Molecular profiling reveals biologically discrete subsets and pathways of progression in diffuse glioma." Cell 164.3 (2016): 550-563.** (https://doi.org/10.1016/j.cell.2015.12.028)

Possible classifications are: 

- Mesenchymal-like 
- Classic-like
- G-CIMP-high
- G-CIMP-low
- LGm6-GBM
- Codel

## Data
<hr>

The input data can be either a Summarized Experiment object of a matrix 
(samples as columns, probes as rows) from the following platforms:

- HM27
- HM450 
- EPIC array.

In this example we will retrieve two samples from TCGA and classify them expecting 
the same result as the paper.

```{r, eval = FALSE, message = FALSE, results = "hide"}
query <- GDCquery(
    project = "TCGA-GBM",
    data.category = "DNA Methylation",
    barcode = c("TCGA-06-0122","TCGA-14-1456"),
    platform = "Illumina Human Methylation 27",
    data.type = "Methylation Beta Value"
)
GDCdownload(query)
dnam <- GDCprepare(query)
```

```{r, eval = FALSE}
assay(dnam)[1:5,1:2]
```

## Function
<hr>

```{r, eval = FALSE}
classification <- gliomaClassifier(dnam)
```

## Results
<hr>
The classfier will return a list of 3 data frames:

1. Sample final classification
2. Each model final classification
3. Each class probability of classification

```{r, eval = FALSE}
names(classification)
classification$final.classification
classification$model.classifications
classification$model.probabilities
```

## Comparing results with paper
<hr>
```{R}
TCGAquery_subtype("GBM") %>%
 dplyr::filter(patient %in% c("TCGA-06-0122","TCGA-14-1456")) %>%
 dplyr::select("patient","Supervised.DNA.Methylation.Cluster")
```