4.8. MS Data Processing

Information related to mass spectra (m/z and intensity) was exported and analyzed using the online platform GeenaR [43]: a target file, that is, a tab-separated value file containing the spectra and sample names, the number of replicates and the sample group, was created and loaded on the platform, together with the dataset folder containing the sample spectra in the form of text files. The following parameters were applied for the pre-processing of spectra: trimming from 0 to 15,000 m/z; variance stabilization: square root (sqrt); smoothing: Savitzky–Golay with half window size of 10; baseline removal: TopHat, with half window size of 75; normalization: total ion current (TIC). An average mass spectrum for each sample was obtained by calculating the mean of intensity for each m/z value of the technical replicates. Peak alignment was performed using MAD (median absolute deviation) for noise estimation (half window size: 20, SNR: 2) and LOWESS (local weight scatterplot smoothing) for phase correction (mass tolerance: 0.002). Then, peaks were selected by applying the strict method for binning and a coverage (minimum frequency of peaks for their selection) of 0.5. A feature (peak) matrix was generated, and a heat map was also provided to see the peak distribution. Principal component analysis was applied on the pre-processed mass spectra. Hierarchical clustering analysis was performed using the average linkage method and the gap statistic to estimate the number of clusters. The dendrogram was created by using the cosine correlation as a similarity measure between two samples (peak list).

Data are expressed as mean ± standard deviation of the mean. Statistical analysis of data was performed by using one-way analysis of variance (ANOVA) and Tukey’s test, with p-values < 0.05 considered statistically significant.