Module 5: Untargeted Metabolomics Processing: Peak Picking, Alignment & MSI Spectral Identification

1. Raw Feature Detection & Retention Time Alignment

Unlike peptides whose fragmentation yields predictable $b/y$ ion ladders matching an amino acid sequence, metabolites are structurally diverse small molecules ($< 1,500\text{ Da}$) comprising carbohydrates, lipids, amino acids, organic acids, and xenobiotics. Untargeted processing tracks three-dimensional features $(m/z, \text{RT}, \text{Intensity})$:

1. Centroiding & Peak Picking: Raw profile mass spectra undergo continuous wavelet transform (CentWave algorithm) to identify chromatographic peaks with signal-to-noise ratio $\text{S/N} \ge 10$ and minimum peak width ($5 - 30\text{ seconds}$).
2. Retention Time Alignment: Unavoidable column aging, mobile phase heating, and matrix pressure shifts cause retention time drift across large patient runs. Algorithms like Obiwarp utilize dynamic time warping across total ion chromatograms to align homologous peaks within $\pm 0.1\text{ min}$.
3. Isotope & Adduct Deconvolution (CAMERA): In electrospray ionization, a single metabolite generates multiple ion species: protonated $[M+H]^+$, sodium adducts $[M+Na]^+$, potassium adducts $[M+K]^+$, loss of water $[M+H-H_2O]^+$, and natural $^{13}\text{C}$ isotopes ($M+1, M+2$). Deconvolution groups co-eluting peaks with correlated intensity profiles into single compound clusters.

2. Feature Table Extraction in R with XCMS

# Complete XCMS 3 Feature Extraction Workflow in R
library(xcms)
library(MSnbase)

# Ingest mzML files from brain tumor cohort
raw_files <- list.files("./mzml_metabolomics/", pattern="*.mzML", full.names=TRUE)
pheno <- data.frame(sample_name = sub(".mzML", "", basename(raw_files)),
                    cohort = ifelse(grepl("GBM", raw_files), "Glioblastoma", "Control"))
raw_data <- readMSData(files = raw_files, pdata = new("NAnnotatedDataFrame", pheno), mode = "onDisk")

# 1. Peak Picking with centWave
cwp <- CentWaveParam(ppm = 5, peakwidth = c(5, 25), snthresh = 10, prefilter = c(3, 1000))
xdata <- findChromPeaks(raw_data, param = cwp)

# 2. Alignment with Obiwarp
xdata <- adjustRtime(xdata, param = ObiwarpParam(binSize = 0.6))

# 3. Peak Grouping across samples
pdp <- PeakDensityParam(sampleGroups = pheno$cohort, minFraction = 0.5, bw = 3)
xdata <- groupChromPeaks(xdata, param = pdp)

# 4. Fill missing peaks (Gap Filling)
xdata <- fillChromPeaks(xdata)
feature_matrix <- featureValues(xdata, value = "into")
cat("Extracted", nrow(feature_matrix), "aligned metabolic features across", ncol(feature_matrix), "patients.\n")