lipyd – A Python module for lipidomics LC MS/MS data analysis

This module implements methods and workflows for MS/MS lipidomics data analysis. Runs primarily in Python 3 but also in Python 2.7.x.

Input and preprocessing

At reading raw mass spec data from mzML files, peak picking and feature detection we rely on the OpenMS library. This ensures computationally efficient processing by well established methods. As our OpenMS integration is not yet complete we provide a temporary solution to read already preprocessed features from CSV files exported by the PEAKS software. We are not comfortable with the idea of building on expensive proprietary software and in the near future we will provide complete integration with OpenMS.

Metabolite database lookup

The lipyd.modb module provides an unified interface to standard databases like SwissLipids and LipidMaps In addition it is able to generate custom metabolite masses. With the default settings the database consists of more than 100 thousands of lipid species. The lipyd.lipid module contains more than 150 predefined lipid classes and it's easy to define new ones. The Sample and SampleSet objects in lipyd.sample, which represent a series of features, support the automatic lookup in the databases.

MS2 spectrum identification

The lipyd.ms2 module contains generic classes to support the analysis and identification of MS2 spectra. Based on around 50 standards run by our group and reviewing many spectra from publications and databases we created bult in rules for identification of more than 80 lipid classes. You can modify the methods or create new ones by writing Python methods. However we are working on MFQL integration to provide a more standard way of defining rules. Also we will introduce similarity search against spectrum databases.

Feature filtering, post-processing

The lipyd.sample and lipyd.feature modules provide classes for analysis of features optionally in relation to other variables and filter them. Analysis and filtering of the features can be done before or after the lipid identification. Doing it before reduces the number of MS2 spectra to be analyzed this way saving time. In the future we will add more utilities to build arrays of features and also MS2 fragments across arbitrary number of experiments to provide opportunities for higher level analysis.