README.Rmd

---
title: "PepTools - An Immunoinformatics (Immunological Bioinformatics) R-package for working with peptide data"
output: github_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

# PepTools

The aim of this package is to supply a set of tools, which will ease working with peptide data within the field of immunoinformatics.

## Getting started

Using Hadley Wickham's brilliant `devtools` package, we can easily install `PepTools` like so:
```{r eval=FALSE}
install.packages("devtools")
devtools::install_github("leonjessen/PepTools")
```

Once the package has been installed, we can simply load it like so:
```{r load_libraries, message=FALSE}
library("PepTools")
```

## PSSM Examples

`PepTools` comes with an example data set of 5,000 9-mer peptides, which have been predicted by [NetMHCpan 4.0 Server](http://www.cbs.dtu.dk/services/NetMHCpan-4.0/) to be strong binders to HLA-A*02:01.

We can view the first 10 peptides like so
```{r view_test_peps}
PEPTIDES %>% head(10)
```

and derive the count matrix:
```{r pssm_counts}
PEPTIDES %>% pssm_counts %>% .[1:9,1:10]
```

and the frequency matrix:
```{r pssm_freqs}
PEPTIDES %>% pssm_freqs %>% .[1:9,1:10]
```

and the bits of information matrix;
```{r pssm_bits}
PEPTIDES %>% pssm_freqs %>% pssm_bits %>% .[1:9,1:10]
```

## Sequence Logo Examples

Using the `ggseqlogo` package, we can visualise the bits matrix:
```{r man_peps_logo, fig.width=5, fig.height=3, fig.align='center'}
PEPTIDES %>% pssm_freqs %>% pssm_bits %>% t %>% ggseqlogo(method="custom")
```

and compare with the `ggseqlogo` build in peptide-to-bits conversion:
```{r auto_peps_logo, fig.width=5, fig.height=3, fig.align='center'}
PEPTIDES %>% ggseqlogo
```

## Peptide Encoding for Deep Learning

`PepTools` also contain function for encoding peptides:
 
```{r pep_encode}
PEPTIDES %>% pep_encode %>% dim
```

As can be seen from the dimensions, `pep_encode` creates a 3D array or a tensor with 5,000 rows, 9 columns and 20 slices, corresponding to `n_peptides x l_peptide x l_enc`, where the encoding is the `BLOSUM62` matrix. This way each peptide is encoded as an'image', which can then be used as input to a Deep Learning model. To get a better understanding of what is going on, we can plot the 3 first encoded peptide 'images':
```{r plot_peptide_images, fig.width=5, fig.height=10, fig.align='center'}
PEPTIDES %>% pep_plot_images
```