LipidHome: a database of theoretical lipids optimized for high throughput mass spectrometry lipidomics

Abstract

Protein sequence databases are the pillar upon which modern proteomics is supported, representing a stable reference space of predicted and validated proteins. One example of such resources is UniProt, enriched with both expertly curated and automatic annotations. Taken largely for granted, similar mature resources such as UniProt are not available yet in some other "omics" fields, lipidomics being one of them. While having a seasoned community of wet lab scientists, lipidomics lies significantly behind proteomics in the adoption of data standards and other core bioinformatics concepts. This work aims to reduce the gap by developing an equivalent resource to UniProt called 'LipidHome', providing theoretically generated lipid molecules and useful metadata. Using the 'FASTLipid' Java library, a database was populated with theoretical lipids, generated from a set of community agreed upon chemical bounds. In parallel, a web application was developed to present the information and provide computational access via a web service. Designed specifically to accommodate high throughput mass spectrometry based approaches, lipids are organised into a hierarchy that reflects the variety in the structural resolution of lipid identifications. Additionally, cross-references to other lipid related resources and papers that cite specific lipids were used to annotate lipid records. The web application encompasses a browser for viewing lipid records and a 'tools' section where an MS1 search engine is currently implemented. LipidHome can be accessed at http://www.ebi.ac.uk/apweiler-srv/lipidhome.

Figure 1. The structural hierarchy of lipid records.

a) The structural hierarchy of lipid records in the Lipid Maps Structural Database (LMSD). The Lipid Maps classification system organises all lipids into “Categories”, “Main Classes” and “Sub Classes”. The lipid records are stored at the “Geometric Isomer” level where the total number of carbons, total number of double bonds, the position of double bonds and the stereochemistry of double bonds are defined for each fatty acid. The transparent lipid identification hierarchy levels are not supported by the LMSD. b) The structural hierarchy of lipid records in the LipidHome database. Similar to the LIPID MAPS classification system, lipids are organised into “Categories”, “Main Classes” and “Sub Classes”. Lipid records are stored at four levels. Each level relates to a typical type of identification from a high throughput mass spectrometry experiment. From structurally undefined “Species” typically identified from a single precursor ion mass, to structurally resolved “Isomer” level identifications.

Figure 2. A diagram of in silico construction of theoretical diradyl lipid “Sub Species”.

Steps: 1. All viable potential fatty acids are generated from a set of starting parameters; 2. They are combined all against all; 3. The head groups with alpha-carbons and linkages are generated; 4. The head groups are crossed with the fatty acid pairs to produce all viable lipid structures within the predefined chemical space.

Figure 3. Screenshot of the LipidHome “Browser” view.

The LipidHome structural hierarchy can be navigated in the far left tree panel. Clicking on a lipid record produces two vertically stacked panels in the right hand panel. The top panel shows general information about the selected record including an image. The bottom panel displays a table of the selected records’ children lipids, i.e. selecting the “Sub Class” “Diacylglycerophosphocholines” provides a list of its “Species”. These lists are exportable to a number of file formats.

Figure 4. Screenshot of the output of an “MS1 search engine”.

Results are split into each input search mass and their corresponding lipid “Species” identifications. The columns are filterable and sortable so that results may be organised prior to exporting them in a number of available file formats. In addition, whole lipid “Categories”, “Main Classes” and “Sub Classes” may be filtered out for the purpose of simplifying the results or replicating any step in the experimental protocol that may have isolated specific sets of lipids prior to the MS analysis.