Prediction of protein location by cross-linking mass spectrometry data

The scripts in this repository utilize cross-linking mass spectrometry (XL-MS) data to predict the location and topology of proteins by direct interactions and defined localization marker.

Summary and script execution

• Python v 3.10
• script can be run in Python (e.g. Python IDE)
• prediction is split into two steps
  1. data preparation script: gets localization and topology information from uniprot for cross-linked proteins in XL-MS data and assigns localization marker; resulting script can be edited manually
  2. prediction script: predicts the localization and topology of proteins by pre-defined localization marker
• for correct execution, please adjust the path of your input data, uniprot.tsv and preferred output in the main function of each script and then run the code

Data preparation script

Input

• XL-MS information, the following columns are needed:

  gene_a	gene_b	Protein1	Protein2	n_score_a	n_score_b	LinkPos1	LinkPos2	crosslink_score	crosslink_ab	crosslinks_a	crosslinks_b

• crosslinks_a and crosslinks_b columns must contain all crosslinks found in XL-MS data for Protein1 and Protein2

• reviewed uniprot information downloaded as tsv with the following columns:

  Entry	Protein names	Gene Names	Subcellular location [CC]	Topological domain	Transmembrane

Method description

• define localization marker by checking which proteins in XL-MS data have either only one subcellular location or, in case of membrane proteins, a defined topology and localization for non-transmembrane regions in uniprot data

• accumulate transmembrane regions for membrane proteins

• combine protein information in XL-MS data, subcellular locations and membrane topology

  • for each protein_a, crosslinks will be accumulated by taking all (inter-protein) interactions where it is either protein_a or protein_b (this will be done for protein_b in data accordingly)
  • if protein is not first part of crosslink (PB-1-PROTEIN-5), the crosslink will be reversed (PROTEIN-5-PB-1)
  • if this protein has a transmembrane region, it will get multiple rows splitting crosslinks and locations/topology in each row according to residue numbers of the transmembrane region

    Protein	Crosslink	Transmembrane region
    Protein1	P1-3-P5-6
    Protein1		5..9
    Protein1	P1-11-P8-5

• topology information for proteins with transmembrane regions will be added from uniprot data and ordered according to their transmembrane regions

• a localization marker column is added, indicating proteins with an already known and thus starting location/topology as True, unknown proteins or transmembrane regions as False

• these localization marker cancidates can be checked for their consistency and crosslinks between them

• the output is a combined table, which will be saved and can be edited manually

  gene	protein	crosslinks	topology	subcellular_location	transmembrane	is_localization_marker

Prediction script

Input

• combined table (output of the data preparation script)

Method description

• if transmembrane regions were changed, crosslink, subcellular location and topology information will be reorderd and updated for each protein
• the prediction is done by using the crosslinked residue of each row of a protein to identify the residue of a linked protein, and assigning topology as well as subcellular location
• afterwards, all predicted residues of a protein will be aggregated and ordered via their residue numbers
• the following output table will be saved

  predicted_gene	predicting_gene	predicted_gene_residue	predicting_gene_residue	predicted_subcellular_location	predicted_topology	predicting_crosslinks	predicted_by_transmembrane	transmembrane_regions	predicting_gene_is_lm

Interpretation of results

In the output table, the "predicted_subcellular_location" and "predicted_topology" columns show the sub-compartment localization or the topology information of the protein. If there is no transmembrane information for a protein in the "Transmembrane region" column, hence a soluble protein, the predicted information of "predicted_subcellular_location" and "predicted_topology" should be consistent. Otherwise, the sub-compartment localization of this protein is ambiguous. If there is transmembrane information in the "Transmembrane region" column, hence a membrane protein, the predicted information needs to be interpreted according to the transmembrane region, and for ease of interpretation, the predicted information for each protein is ordered via their residue numbers.

Example

• the example folder contains exemplary input and output tables