LipidXplorer 1.5

LipidXplorer is a software that supports a variety of untargeted shotgun lipidomics experiments and experiments from LC/MS lipidomics. It is designed to support bottom-up and top-down shotgun lipidomics experiments performed on all types of tandem mass spectrometers. Lipid identification does not rely on a database resource of reference or simulated mass spectra. 

Aims

For keeping LipidXplorer maintainable and utilize updated programming features as well as libraries the code base was upgraded from Python 2 to Python 3. Official support for Python 2 is no longer available. This version is also meant as bridging element between our legacy workflow for shotgun lipidomics (LipidXplorer 1.5 –> lxPostman) and LipidXplorerNext. We further included features that are aimed to improve user guidance as well as overall performance of Lipidxplorer.  

Benchmark Data

The samples originated from mice lung tissue. The data consists of data in *.raw and *.mzML format (5 mins positive, 5 mins negative), a batch script to convert data from raw to mzML format using MSConvert, a settings (*.ini) file for LipidXplorer settings, and a standardized MFQLs for both positive and negative ion mode.

Detailed Changelog

New Features:

1.     Batch mode. Batch mode processes multiple samples in parallel using multiprocessing, then combines the per-sample results into one unified result.

When to use Batch mode: Recommended when the sample size is very large and consists of different experimental groups.

How it works: Each sample in the import directory is assigned to its own worker process (one sample per worker, run via a multiprocessing pool). Each worker independently builds that sample's MasterScan and runs the full MFQL query set against it, and finally writes its own per-sample result *_out.csv to disk. Once all samples are processed, per-sample result files are combined into a single merged table, aligned using LipidSpecies as the common key across samples.

Column requirements for merging*:

o   LipidSpecies is strictly required in each per-sample *_out.csv.  It is the key to align rows across samples. If it's missing from an output file, merging fails outright.

o   Recommended (a warning is raised if any of these columns are missing, but the merge still proceeds): LipidClass, Mass, ScanPolarity, Intensity -these are parameters used for consistent metadata reporting.

o   Also recognized and handled specially: PrecursorIntensity and any FragmentXIntensity columns (e.g. FragmentAIntensity, FragmentBIntensity …). These aren't part of the required/recommended validation check, but the merge logic specifically detects them by name pattern and treats them the same way as Intensity: kept per-sample (renamed with a sample suffix, e.g. Intensity:sample1.mzML) in contrast to parameters like PrecursorERRppm.

o   Missing or non-numeric intensity values are treated as 0 after merging.

o   You can require that a lipid species must have a positive Intensity value in at least a given fraction of samples to be kept in the final merged table. This filter requires setting an Occurrence Threshold (e.g. a value of 0.5 keeps only lipid species detected in at least half of all samples).

o   We recommend to use the standardized report structure of the MFQLs provided with this release.

Expectation from Batch Mode: Results from Batch Mode may differ from normal mode, as samples are processed differently. Using the Batch Mode is expected to report overall more lipids that may not be common across all samples (set the Occurrence Threshold accordingly). We recommend using this mode, when your experimental groups have large phenotypic differences for instance different organs need to be compared, extreme biological differences are observed and large number of samples need to be processed (computational expensive alignment processes are avoided).   

2.     Removed mzXML support.

3.     Restored DTA/CSV data support.

Text file based import was restored to provide a well-structured and easy to handle way to the improved MFQL identification engine. It further allows to use customized algorithms for peak picking / isotopic correction and MS-feature selection as import.

https://lifs-tools.org/wiki/index.php?title=LipidXplorer_Reference#Supported_file_formats

4.     Isotopic correction process was updated.

In *_dump.csv files flags are introduced:

Bug Fixes:

1.     Marking/isotopic correction was dependent in certain cases on the order of MFQL execution — this is now deterministic.

2.     MS2-level isotopic correction ignored charge state (assumed charge 1) — this is now implemented for higher charge stages.

Concessions:

1. While assignment of chemical sum compositions to m/z is correct for the regular usage of MFQLs in the ‘Run‘ panel, there remains an inconsistence with results of the ‘MS Tools‘ panel function ‘Mass-to-sum-composition‘.  ‘Mass-to-sum-composition‘ calculations will still provide compositions with the smallest fitting m/z instead of the match with the smallest relative error.

2. Isotopic correction in LipidXplorer 1.5 still has the following limitations: 

a) Isotopic pattern only calculated until M+4
b) Triggering isotopic correction is linked to the utilized MFQL scripts. For comprehensive analysis always all expected lipid classes should be included in the analysis.
c) No general isotopic model is present to deal with general overlaps that might be part of chemical background. However, wherever Precursor and Fragment ions are marked together a correction is initiated. The basis is the marked result array before application of SUCHTHAT conditions of MFQLs. This correction is not systematically tested. 

Acknowledgements:

The current maintainers, developers and contributors for LX would like to thank Ronny Herzog – initial author of the software for his continued support.

References:

1.            Schuhmann, K., et al., Intensity-Independent Noise Filtering in FT MS and FT MS/MS Spectra for Shotgun Lipidomics. Anal Chem, 2017. 89(13): p. 7046-7052.

2.            Eggers, L.F. and D. Schwudke, Shotgun Lipidomics Approach for Clinical Samples. Methods Mol Biol, 2018. 1730: p. 163-174.

3.            Herzog, R., et al., LipidXplorer: a software for consensual cross-platform lipidomics. PLoS One, 2012. 7(1): p. e29851.

4.            Herzog, R., et al., A novel informatics concept for high-throughput shotgun lipidomics based on the molecular fragmentation query language. Genome Biol, 2011. 12(1): p. R8.

5.            Herzog, R., D. Schwudke, and A. Shevchenko, LipidXplorer: Software for Quantitative Shotgun Lipidomics Compatible with Multiple Mass Spectrometry Platforms. Curr Protoc Bioinformatics, 2013. 43: p. 14 12 1-30.

6.            Chambers, M.C., et al., A cross-platform toolkit for mass spectrometry and proteomics. Nat Biotechnol, 2012. 30(10): p. 918-20.