libra

Repo for libra - the data toolbox for working with LC-MS based metabolomics data (cleaning, calibration and visualization)

Install

devtools::install_github("tuhulab/libra")

Before start - data transformation

Metabolomics data needs to be transformed as data_table and sample_sample to start libra calibration. These two tables can be easily generated from previous XCMS/CAMERA workflow.

data_table should compromize columns: mz, rt, X1, X2, X3, ... Xn. "Xi" denotes each sample, which each row stores intensity of each feature.

sample_table should compromize column: sample_name, code, batch (if multiple batch is involved).

Two tables

Workflow

# Load data ----------------------------------
library(tidyr)
library(dplyr)
project <- "my_project"
biospecimen <- "urine"
mode <- "pos"

data_table <- readr::read_csv(file.path("data", project, biospecimen,
                                        paste0(paste(biospecimen, mode, "raw" ,sep = "_"),".csv")))
sample_table <- readr::read_csv(file.path("data", project, biospecimen,
                                        paste0(paste(biospecimen, mode, "sampleinfo" ,sep = "_"),".csv"))) %>%
mutate(sample_name = ifelse(sample_name == "AS21-1", "AD21-1", sample_name))

# Define parameter ---------------------------
sample_rm <- c("Blank", "Blank_IS", "Assay_blank", "MetStd", "Global_pool_x3_IS", "Global_pool_x10_IS") # for removing samples
pattern.blank <- "Blank$|Assay blank"
pattern.pool <- "Global pool IS$"

# Remove features (Clean Blanks, early and late birds)  --------
d_rmF <- data_table %>% libra::rm_feature.blank() %>% libra::rm_feature.bird()

# Remove samples(rmS) ------------------------
d_rmFS <- d_rmF %>% libra::rm_sample(sample.df = sample_table, sample_rm) %>% left_join(sample_table %>% select(code, batch))

# Impute NAs ---------------------------------
d_rmSFNa <- d_rmFS %>% libra::imputeNA()

# Calibrate: batch effect (intra)-------------
d_rmSFNaC <- d_rmSFNa %>% libra::calibrateBatch.intra.rlm()

# Calibrate: batch effect (inter)-------------
d_rmSFNaC <- d_rmSFNa %>% libra::calibrateBatch.inter.rlm()

# Output data wide format---------------------
d_output_w <- d_rmSFNaC %>% select(feature, mz, rt, pcgroup, adduct, id_predret, id_kudb, code, intensity_intra_inter_calibrated) %>%
pivot_wider(names_from = code, values_from = intensity_intra_inter_calibrated)
readr::write_csv(d_output_w, file.path("data", project, biospecimen, paste0(paste(biospecimen, mode, "calibrated", "wide", sep = "_"), ".csv")))
readr::write_csv(sample_table %>% select(-mzMLs) %>% filter(code %in% (d_output_w %>% colnames() %>% stringr::str_match("X\\d{1,}") %>% purrr::discard(is.na))),
                file.path("data", project, biospecimen, paste0(paste(biospecimen, mode, "sampleinfo", "calibrated",sep = "_"), ".csv")))

# Output data long data ----------------------
readr::write_csv(d_rmSFNaC %>% select(-intensity, -injection_sequence, -intensity_intra_calibrated, -center_intensity, -intra_batch_center, -multi_batch_center, -factor),
                file.path("data", project, biospecimen,
                        paste0(paste(biospecimen, mode, "calibrated", "long",sep = "_"),".csv")))