In leonjessen/PepTools: PepTools - An R-package for making immunoinformatics accessible

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:

install.packages("devtools")
devtools::install_github("leonjessen/PepTools")

Once the package has been installed, we can simply load it like so:

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 to be strong binders to HLA-A*02:01.

We can view the first 10 peptides like so

PEPTIDES %>% head(10)

and derive the count matrix:

PEPTIDES %>% pssm_counts %>% .[1:9,1:10]

and the frequency matrix:

PEPTIDES %>% pssm_freqs %>% .[1:9,1:10]

and the bits of information matrix;

PEPTIDES %>% pssm_freqs %>% pssm_bits %>% .[1:9,1:10]

Sequence Logo Examples

Using the ggseqlogo package, we can visualise the bits matrix:

PEPTIDES %>% pssm_freqs %>% pssm_bits %>% t %>% ggseqlogo(method="custom")

and compare with the ggseqlogo build in peptide-to-bits conversion:

PEPTIDES %>% ggseqlogo

Peptide Encoding for Deep Learning

PepTools also contain function for encoding peptides:

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':

PEPTIDES %>% pep_plot_images

leonjessen/PepTools documentation built on May 29, 2019, 3:40 a.m.