pelmesha 0.4.4

A package for proccesing and aligning peaklist Mass-spectrometry imaging data from .imzml files

Package pelmesha

pelmesha (Peak Extraction Library for Mass spectrometry Enhanced by Statistical High-throughput Analysis) is a Python package that allows users to process Mass Spectrometry Imaging (MSI) data from imzml files. It provides several features, including:

1. Conversion of raw imzml data to the hdf5 format (function imzml2hdf5 from module pspectra)
2. Processing of mass spectra (function Raw2proc from module pspectra)
3. Creation of peaklists from the data (functions proc2peaklist and Raw2peaklist from module pspectra)
4. Generation of a feature matrix from a single image's peaklist by grouping peaks (function Pgrouping_KD from module pfeats)
5. Generation of a feature matrix for multiple images from the peaklists of each image (function Roi_Pgrouping_KD from module pfeats)

The processing of mass spectra includes several steps:

• Data resampling — This process allows you to bring data to a uniform scale between points on the mz and to a single scale on the mz.
• Alignment of spectra relative to reference peaks using the msalign tool. It should be noted that msalign does a worse job with non-continuous and non-uniform data, so it is strongly recommended to perform a resampling process before using it. Also msalign is modified in this package for correct work with other steps.
• Baseline correction using the pybaselines package.
• Smoothing — based on code snippets from the mMass library, which provide smoothing using the moving average, Gaussian, and Savitsky-Goley algorithms.
• Peak-picking — peaks in the spectrum are searched and filtered.

In the mass spectra in the image, there is a slight difference in the peak values, even after alignment during processing. To further analyze these spectra, it will be necessary to group the peaks based on their relative positions and create a feature matrix as a result.

The Pgrouping_KD and Roi_Pgrouping_KD functions combine the wandering peak values from signals in mass spectra into a single mz value based on their kernel density estimation. This is achieved by determining the centers around which the peak values are located.

To ensure accurate results, high-quality bandwidth selection is required. This can be done manually or automatically. The probability density is estimated using the FFTKDE function from the kdepy package, which is extremely fast and uses Fourier transforms for calculation.