In general, the ion identity molecular networking (IIMN) workflow starts with LC-MS2 data processing in one of the supported open source feature-finding tools. After the creation of an aligned feature list of all samples, ion species that originate from the same analyte are grouped and annotated by MS1 criteria, such as their retention time, feature shape correlation, and m/z difference. Here, such groups are named ion identity networks. Subsequently, information of detected features and their representative MS2 spectra, ion identities, and connections to other ion identities are exported and transferred to the GNPS web server for the molecular networking part (refer to tool-specific sections for details). After the construction of ion identity molecular networks, features share connectivity based on MS2 spectral cosine similarity and MS1-based feature shape correlation. In addition to this combined network, GNPS calculates a version with collapsed IIN, where one node represents multiple ions of the same molecule. Results are available in the GNPS web interfaces and as downloads in various open formats as tables and networking files to allow local visualization, reviewing, and post-processing.

The IIMN workflow aids the feature-based molecular networking workflow7 by adding MS1 specific information, which is provided as new columns in the quantification table and as additional edges in a Supplementary Pairs text file within the GNPS-FBMN workflow. The option of additional edges from other tools was introduced to stimulate and facilitate the development of new computational methods that link nodes in the resulting molecular networks and was initially developed for IIMN. The text format follows a generic comma-separated style with the columns ID1 and ID2 (matching the feature IDs in the feature quantification table and mgf), EdgeType (defining the method), Score (numerical), and Annotation. To enable a broad user base to employ ion identity molecular networking in their studies, three popular mass spectrometry processing tools, namely, MZmine17, MS-DIAL30, and XCMS( + CAMERA)13,16, were modified or adapted with additional export scripts or modules. In comparison to FBMN, IIMN can include features that are lacking MS2 fragmentation spectra but are connected to other feature nodes by MS1 IIN edges. Regarding a higher detectability by MS1 compared to triggered MS2 acquisition, the additional nodes with ion identities complement the resulting networks with information otherwise lost in FBMN or classical MN.

If needed, convert the spectral data files to an open format (e.g., mzML)

Import the data into one of the open source tools: MZmine, MS-DIAL, or XCMS

Process the data to create a feature list (aligned overall samples)

Perform MS1-based feature grouping and ion identity annotation

Export the feature list as a feature quantification table (.csv), an MS2 spectral summary file (.mgf), which contains a representative fragmentation spectrum for each feature, and a supplementary edges files (IIN files, .csv) (more information in the tool-specific workflow sections)

Create a metadata file to group samples for statistics (optional)

Upload all files to GNPS and start a new feature-based molecular networking job (MZmine can directly submit and start a new IIMN job on GNPS)

Download and visualize the results in a network analysis software (e.g., Cytoscape31,

The option Download Cytoscape Data provides two.graphml networking files

The standard FBMN and IIMN networks (base directory)

IIMN networks with collapsed ion identity networks (in the gnps_molecular_network_iin_collapse_graphml directory)

The option Direct Cytoscape Preview/Download provides the IIMN network and its collapsed version as Cytoscape projects with various style presets

Refer to the documentation on how to run FBMN within GNPS and multiple mass spectrometry data processing tools.

For IIMN, refer to the related part of the GNPS documentation.

One result of the GNPS-IIMN workflow is the combined networks with IIN collapsed into single nodes. For this, all ion nodes with the same IIN ID are merged into a representative node based on the feature with the highest library match score, if available, or otherwise the feature with the maximum abundance. While all IIN edges are collapsed, MN edges of all ion identities are redirected to their representative nodes so that duplicates replace existing edges if their edge score (cosine similarity) is higher. Limiting the number of MN edges to the one with the highest cosine similarity. Furthermore, representative collapsed nodes are extended by multiple attributes, including the intensity of each ion identity and their summed intensity. This enables the direct comparison of ionization tendencies and provides new visualization options. An example with pie charts of the ion abundances is demonstrated in Supplementary Fig. 3.

In IIMN, nodes may combine annotations from MS2 spectral library matching and MS1 ion identity networking. As cross-validation, GNPS parses and harmonizes the ion species string of both the detected ion identity and matching spectral library entry before checking for equality. The results are reported as an additional column in the node table. This equality check facilitates manual reviewing and the spotting of discrepancies between the MS1 and MS2 annotations.

The ion string parser harmonizes an input (e.g., [M − H2O + 2H]2+) in the following steps:

Spaces are removed

Charge state is detected and removed from the input (2+)

Brackets are removed ([]())

Input is split into added (+2H) and removed (−H2O) parts

Both lists are sorted alphabetically (+2H sorted by letter H)

If the charge state is missing, it is calculated for all parts that are listed in a lookup table (e.g., +Na or +H correspond to charge 1+)

The harmonized string is constructed by concatenation of [M-all removed parts + all added parts]charge state.

As an example, the harmonized string [M + H]+ is produced by the input strings M + H, M + H + , and [M + H]+, which are all commonly found in the GNPS spectral libraries and as an output of various software tools.

The full open source code of the ion string parser and its latest charge lookup table can be found on GitHub (