Baseline correction¶
Info
This documentation covers the MS and Other detector workflow. The difference is either selecting a feature lists or other raw files + traces.
Info
To process data from other detectors, the Other detector service of mzmine is required. This feature is included in the mzmine PRO subscription. Academic program and other users can reach out to inquire access options.
Baseline correction is a useful tool to correct for large drifts in baseline during chromatographic separation. However, it can be tricky to optimize the baseline correction.
The module fits an estimated baseline to each selected trace and subtracts it. Corrected intensities are clipped at zero, so the baseline itself may drop below zero (e.g. for detector drifts) without producing negative signals. Four baseline correctors are available in the user interface: arPLS (default), LOESS, Polynomial and Spline.
Info
While baseline correction can correct some drifts in the baseline, the better approach is usually to acquire a reference wavelength and reference the acquired data on that during the data acquisition.
Recommended citations¶
Info
When using the arPLS baseline corrector, please also cite:
Baek, S.-J., Park, A., Ahn, Y.-J., Choo, J. Baseline correction using asymmetrically reweighted penalized least squares smoothing. Analyst 140, 250-257 (2015). https://doi.org/10.1039/c4an01061b
Parameters¶
Raw data files¶
Select the MS data files to baseline correct the selected traces of.
Trace selection¶
Select the specific traces you want to baseline correct. Typically, it is not necessary to baseline correct all existing traces, but only the ones you want to correlate with the MS, e.g., the UV ( =Absorption) traces
For a detailed description of all sub-parameters see trace selection parameter.
Suffix¶
A suffix that will be appended to the name of the individual trace.
Baseline corrector¶
Select the baseline correction algorithm. The available options are arPLS (default), LOESS, Polynomial and Spline. Each corrector exposes its own sub-parameters below the selection.
Shared parameters¶
Exclude peaks¶
If selected, the baseline is estimated in two internal passes. First, the baseline is corrected for the full chromatogram. Then that intermediate baseline is subtracted and a peak detection is run to locate chromatographic peaks. In the second pass the baseline is fitted again, this time excluding the detected peak regions so that peaks do not pull the baseline upwards.
This option is available for all four correctors. It is most useful for the interpolation-based correctors (LOESS, Polynomial, Spline); arPLS already suppresses peaks on its own and usually works best with Exclude peaks disabled.
Percentage of baseline samples¶
Specify how many of the data points of the chromatogram are sampled to fit the baseline (default 5%). This value depends on the acquisition rate of the detector and whether previous binning has been applied. The higher the acquisition rate, the lower the percentage of samples can be.
Info
This parameter is used by the LOESS, Polynomial and Spline correctors. arPLS fits the full trace and does not use it.
Correctors¶
arPLS¶
The arPLS (asymmetrically reweighted penalized least squares) baseline corrector is the default and recommended corrector. It fits a smooth baseline to the whole trace and iteratively down-weights data points that lie above the current baseline, so peaks are pushed out of the fit while the baseline settles onto the lower envelope of the signal. It does not require an explicit peak detection and therefore works well with Exclude peaks turned off.
Smoothness (lambda)¶
Controls how stiff the baseline is. Larger values yield a stiffer, smoother baseline that ignores narrow signals; smaller values let the baseline follow the data more closely, but may start to follow peaks. Typical range 1e3 - 1e7, default 1e5.
Max iterations¶
The maximum number of reweighting iterations (default 50). The fit usually converges well before this limit, so this value rarely needs to be changed.
LOESS¶
The LOESS (locally estimated scatterplot smoothing) baseline corrector fits the sampled baseline points using the LOESS algorithm, which smooths the baseline locally. It is a robust choice when arPLS is not suitable.
Bandwidth¶
Specify how wide the smoothing window is at any point of the chromatogram. E.g., 10% means that 10% of the sampled data points are used at any specific retention time. For a chromatographic run of 10 min, a window of about 1 min around each point is taken into account. Larger values create a smoother baseline but react more slowly to sharp shifts.
Iterations¶
Specify how many robustness iterations the LOESS algorithm performs (default 2). This is independent of the iterations performed by the Exclude peaks option and can increase robustness against sudden changes in the baseline.
Polynomial¶
The polynomial baseline corrector fits the sampled baseline points with an nth degree polynomial.
Degree of polynomial¶
Specify the degree of the polynomial (default 4). The higher the degree, the more the baseline will resemble the original data, which may be a disadvantage.
Spline¶
The Spline baseline corrector fits a cubic (3rd degree) spline through the sampled baseline points. The spline passes exactly through the sampled points and interpolates between them, so — unlike LOESS — the sampled points are not additionally smoothed.
Optimization¶
The show preview option is recommended to optimize the baseline correction.
Percentage of samples¶
The percentage of samples specifies how many data points of the baseline are included in the baseline correction. This parameter depends vastly on the acquisition rate. The influence of this parameter is shown here:
Smoothness (arPLS)¶
If the arPLS baseline follows the peaks too closely, increase Smoothness (lambda); if it is too stiff and floats above a curved baseline, decrease it.
FAQ¶
Q: Ghost peaks appear at the start/end of my chromatogram when I apply baseline correction.
A: The ghost peaks are most often triggered by rapid shifts in the baseline, e.g. at the start of a chromatogram during the dead volume or at the end when the column is reequilibrated. The best solution is to restrict the retention time range to the relevant parts of the chromatogram by trimming.
In UV: trimming traces
In MS: Chromatogram builder
Both options are controlled by the Wizard Crop retention time