The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

PepMapViz: A Versatile Toolkit for Peptide Mapping, Visualization, and Comparative Exploration ================

PepMapViz

PepMapViz is a versatile R visualization package that empowers researchers with comprehensive visualization tools for seamlessly mapping peptides to protein sequences, identifying distinct domains and regions of interest, accentuating mutations, and highlighting post-translational modifications, all while enabling comparisons across diverse experimental conditions. Potential applications of PepMapViz include the visualization of cross-software mass spectrometry results at the peptide level for specific protein and domain details in a linearized format and post-translational modification coverage across different experimental conditions; unraveling insights into disease mechanisms. It also enables visualization of MHC-presented peptide clusters in different antibody regions predicting immunogenicity in antibody drug development.

Installation

You can install the development version of PepMapViz from GitHub using the devtools package.

# Install devtools if you haven't already
install.packages("devtools")

# Install PepMapViz from the package
devtools::build()
devtools::install()

Features

  1. Mapping peptides to protein sequences
  2. Identifying distinct domains and regions of interest
  3. Accentuating mutations
  4. Highlighting post-translational modifications
  5. Enabling comparisons across diverse experimental conditions

Usage

This is a basic example which shows you how to solve a common problem:

library(PepMapViz)

# Read all files from a folder
folder_path <- system.file("extdata", package = "PepMapViz")
resulting_df <- combine_files_from_folder(folder_path)

# Strip the sequence 
striped_data_peaks <- strip_sequence(resulting_df, "Peptide", "Sequence", "PEAKS")

# Extract modifications information
PTM_table <- data.frame(PTM_mass = c("15.99", ".98", "57.02"),
                        PTM_type = c("Ox", "Deamid", "Cam"))
converted_data_peaks <- obtain_mod(
  striped_data_peaks,
  "Peptide",
  "PEAKS",
  strip_seq_col = NULL,
  PTM_table,
  PTM_annotation = TRUE,
  PTM_mass_column = "PTM_mass"
)

# Match peptide sequence with provided sequence and calculate positions
whole_seq <- data.frame(
  Epitope = c("Boco", "Boco"),
  Chain = c("HC", "LC"),
  Region_Sequence = c("QVQLVQSGAEVKKPGASVKVSCKASGYTFTSYYMHWVRQAPGQGLEWMGEISPFGGRTNYNEKFKSRVTMTRDTSTSTVYMELSSLRSEDTAVYYCARERPLYASDLWGQGTTVTVSSASTKGPSVFPLAPCSRSTSESTAALGCLVKDYFPEPVTVSWNSGALTSGVHTFPAVLQSSGLYSLSSVVTVPSSNFGTQTYTCNVDHKPSNTKVDKTVERKCCVECPPCPAPPVAGPSVFLFPPKPKDTLMISRTPEVTCVVVDVSHEDPEVQFNWYVDGVEVHNAKTKPREEQFNSTFRVVSVLTVVHQDWLNGKEYKCKVSNKGLPSSIEKTISKTKGQPREPQVYTLPPSREEMTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPMLDSDGSFFLYSKLTVDKSRWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK", 
                      "DIQMTQSPSSLSASVGDRVTITCRASQGISSALAWYQQKPGKAPKLLIYSASYRYTGVPSRFSGSGSGTDFTFTISSLQPEDIATYYCQQRYSLWRTFGQGTKLEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVDNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC"
  )
)
matching_result <- match_and_calculate_positions(
  converted_data_peaks,
  'Sequence',
  whole_seq,
  match_columns = NULL,
  sequence_length = c(10, 30),
  column_keep = c(
    "PTM_mass",
    "PTM_position",
    "reps",
    "Area",
    "Donor",
    "PTM_type"
  )
)

# Quantify matched peptide sequences by PSM
matching_columns = c("Chain", "Epitope")
distinct_columns = c("Donor")
data_with_psm <- peptide_quantification(
  whole_seq,
  matching_result,
  matching_columns,
  distinct_columns,
  quantify_method = "PSM",
  with_PTM = TRUE,
  reps = TRUE
)
region <- data.frame(
  Epitope = c("Boco", "Boco", "Boco", "Boco", "Boco", "Boco"),
  Chain = c("HC", "HC", "HC", "HC", "LC", "LC"),
  Region = c("VH", "CH1", "CH2", "CH3", "VL", "CL"),
  Region_start = c(1,119,229,338,1,108),
  Region_end = c(118,228,337,444,107,214)
)
result_with_psm <- data.frame()
for (i in 1:nrow(region)) {
  chain <- region$Chain[i]
  region_start <- region$Region_start[i]
  region_end <- region$Region_end[i]
  region_name <- region$Region[i]

  temp <- data_with_psm[data_with_psm$Chain == chain & 
                          data_with_psm$Position >= region_start & 
                          data_with_psm$Position <= region_end, ]
  temp$Region <- region_name

  result_with_psm <- rbind(result_with_psm, temp)
}
  
head(result_with_psm)
##   Character Position Chain Epitope PSM Donor   PTM PTM_type Region
## 1         Q        1    HC    Boco   0    D1 FALSE     <NA>     VH
## 2         V        2    HC    Boco   0    D1 FALSE     <NA>     VH
## 3         Q        3    HC    Boco   0    D1 FALSE     <NA>     VH
## 4         L        4    HC    Boco   0    D1 FALSE     <NA>     VH
## 5         V        5    HC    Boco   0    D1 FALSE     <NA>     VH
## 6         Q        6    HC    Boco   0    D1 FALSE     <NA>     VH
# Plotting peptide in whole provided sequence
domain <- data.frame(
  domain_type = c("CDR H1", "CDR H2", "CDR H3", "CDR L1", "CDR L2", "CDR L3"),
  Region = c("VH", "VH", "VH",  "VL", "VL", "VL"),
  Epitope = c("Boco", "Boco", "Boco", "Boco", "Boco", "Boco"),
  domain_start = c(26, 50, 97,  24, 50, 89),
  domain_end = c(35, 66, 107,  34, 56, 97)
)
x_axis_vars <- c("Region")
y_axis_vars <- c("Donor")
column_order <- list(
    Donor = "D1,D2,D3,D4,D5,D6,D7,D8",
    Region = "VH,CH1,CH2,CH3,VL,CL"
)
domain_color <- c(
"CDR H1" = "#F8766D",
"CDR H2" = "#B79F00",
"CDR H3" = "#00BA38",
"CDR L1" = "#00BFC4",
"CDR L2" = "#619CFF",
"CDR L3" = "#F564E3"
)
PTM_color <- c(
  "Ox" = "red",
  "Deamid" = "cyan",
  "Cam" = "blue",
  "Acetyl" = "magenta"
)
label_value = list(Donor = "D1")

p_psm <- create_peptide_plot(
  result_with_psm,
  y_axis_vars,
  x_axis_vars,
  y_expand = c(0.2, 0.2),
  x_expand = c(0.5, 0.5),
  theme_options = list(legend.box = "horizontal"),
  labs_options = list(title = "PSM Plot", x = "Position", fill = "PSM"),
  color_fill_column = 'PSM',
  fill_gradient_options = list(limits = c(0, 160)),  # Set the limits for the color scale
  label_size = 1.9,
  add_domain = TRUE,
  domain = domain,
  domain_start_column = "domain_start",
  domain_end_column = "domain_end",
  domain_type_column = "domain_type",
  domain_color = domain_color,
  PTM = TRUE,
  PTM_type_column = "PTM_type",
  PTM_color = PTM_color,
  add_label = TRUE,
  label_column = "Character",
  label_value = label_value,
  column_order = column_order
)

Getting Started

For a detailed guide on how to use PepMapViz, please refer to our vignette and docuemntation under inst/doc.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.