R-based access to Mass-Spec data (RaMS)

Table of contents: Overview - Installation - Usage - File types - Contact

Overview

RaMS is a lightweight package that provides rapid and tidy access to mass-spectrometry data. This package is lightweight because it’s built from the ground up rather than relying on an extensive network of external libraries. No Rcpp, no Bioconductor, no long load times and strange startup warnings. Just XML parsing provided by xml2 and data handling provided by data.table. Access is rapid because an absolute minimum of data processing occurs. Unlike other packages, RaMS makes no assumptions about what you’d like to do with the data and is simply providing access to the encoded information in an intuitive and R-friendly way. Finally, the access is tidy in the philosophy of tidy data. Tidy data neatly resolves the ragged arrays that mass spectrometers produce and plays nicely with other tidy data packages.

Installation

To install the stable version on CRAN:

install.packages('RaMS')

To install the current development version:

devtools::install_github("wkumler/RaMS", build_vignettes = TRUE)

Finally, load RaMS like every other package:

library(RaMS)

Usage

There’s only one main function in RaMS: the aptly named grabMSdata. This function accepts the names of mass-spectrometry files as well as the data you’d like to extract (e.g. MS1, MS2, BPC, etc.) and produces a list of data tables. Each table is intuitively named within the list and formatted tidily:

msdata_dir <- system.file("extdata", package = "RaMS")
msdata_files <- list.files(msdata_dir, pattern = "mzML", full.names=TRUE)

msdata <- grabMSdata(files = msdata_files[2:4], grab_what = c("BPC", "MS1"))

Some additional examples can be found below, but a more thorough introduction can be found in the vignette.

BPC/TIC data:

Base peak chromatograms (BPCs) and total ion chromatograms (TICs) have three columns, making them super-simple to plot with either base R or the popular ggplot2 library:

knitr::kable(head(msdata$BPC, 3))

plot(msdata$BPC$rt, msdata$BPC$int, type = "l", ylab="Intensity")

library(ggplot2)
ggplot(msdata$BPC) + geom_line(aes(x = rt, y=int, color=filename)) +
  facet_wrap(~filename, scales = "free_y", ncol = 1) +
  labs(x="Retention time (min)", y="Intensity", color="File name: ") +
  theme(legend.position="top")

MS1 data:

MS¹ data includes an additional dimension, the m/z of each ion measured, and has multiple entries per retention time:

knitr::kable(head(msdata$MS1, 3))

This tidy format means that it plays nicely with other tidy data packages. Here, we use data.table and a few other tidyverse packages to compare a molecule’s ¹³C and ¹⁵N peak areas to that of the base peak, giving us some clue as to its molecular formula. Note also the use of the trapz function (available in v1.3.2+) to calculate the area of the peak given the retention time and intensity values.

library(data.table)
library(tidyverse)

M <- 118.0865
M_13C <- M + 1.003355
M_15N <- M + 0.997035

iso_data <- imap_dfr(lst(M, M_13C, M_15N), function(mass, isotope){
  peak_data <- msdata$MS1[mz%between%pmppm(mass) & rt%between%c(7.6, 8.2)]
  cbind(peak_data, isotope)
})

iso_data %>%
  group_by(filename, isotope) %>%
  summarise(area=trapz(rt, int)) %>%
  pivot_wider(names_from = isotope, values_from = area) %>%
  mutate(ratio_13C_12C = M_13C/M) %>%
  mutate(ratio_15N_14N = M_15N/M) %>%
  select(filename, contains("ratio")) %>%
  pivot_longer(cols = contains("ratio"), names_to = "isotope") %>%
  group_by(isotope) %>%
  summarize(avg_ratio = mean(value), sd_ratio = sd(value), .groups="drop") %>%
  mutate(isotope=str_extract(isotope, "(?<=_).*(?=_)")) %>%
  knitr::kable()

With natural abundances for ¹³C and ¹⁵N of 1.11% and 0.36%, respectively, we can conclude that this molecule likely has five carbons and a single nitrogen.

Of course, it’s always a good idea to plot the peaks and perform a manual check of data quality:

ggplot(iso_data) +
  geom_line(aes(x=rt, y=int, color=filename)) +
  facet_wrap(~isotope, scales = "free_y", ncol = 1)

MS¹ data typically consists of many individual chromatograms, so RaMS provides a small function that can bin it into chromatograms based on m/z windows.

msdata$MS1 %>%
  arrange(desc(int)) %>%
  mutate(mz_group=mz_group(mz, ppm=10, max_groups = 3)) %>%
  qplotMS1data(facet_col = "mz_group")

We also use the qplotMS1data function above, which wraps the typical ggplot call to avoid needing to type out ggplot() + geom_line(aes(x=rt, y=int, group=filename)) every time. Both the mz_group and qplotMS1data functions were added in RaMS version 1.3.2.

MS2 data:

DDA (fragmentation) data can also be extracted, allowing rapid and intuitive searches for fragments or neutral losses:

msdata <- grabMSdata(files = msdata_files[1], grab_what = "MS2")

For example, we may be interested in the major fragments of a specific molecule:

msdata$MS2[premz%between%pmppm(118.0865) & int>mean(int)] %>%
  plot(int~fragmz, type="h", data=., ylab="Intensity", xlab="Fragment m/z")

Or want to search for precursors with a specific neutral loss:

msdata$MS2[, neutral_loss:=premz-fragmz] %>%
  filter(neutral_loss%between%pmppm(60.02064, 5)) %>%
  head(3) %>% knitr::kable()

Minifying MS files

As of version 1.1.0, RaMS also has functions that allow irrelevant data to be removed from the file to reduce file sizes. See the vignette for more details.

tmzML documents

Version 1.2.0 of RaMS introduced a new file type, the “transposed mzML” or “tmzML” file to resolve the large memory requirement when working with many files. See the vignette for more details.

File types

RaMS is currently limited to the modern mzML data format and the slightly older mzXML format, as well as the custom tmzML format as of version 1.2.0. Tools to convert data from other formats are available through Proteowizard’s msconvert tool. Data can, however, be gzip compressed (file ending .gz) and this compression actually speeds up data retrieval significantly as well as reducing file sizes.

Currently, RaMS handles MS¹ MS², and MS³ data. This should be easy enough to expand in the future, but right now I haven’t observed a demonstrated need for higher fragmentation level data collection.

For an analysis of how RaMS compares to other methods of MS data access and alternative file types, consider browsing the speed & size comparison vignette.

Contact

Feel free to submit questions, bugs, or feature requests on the GitHub Issues page.

README last built on 2023-11-29