CellWhere is a data combining and visualization tool that enables bench researchers to quickly explore the reported subcellular locations of a list of genes/proteins, and to put these subcellular locations into the context of previously identified physical interactions that could be occurring between these proteins and others within the cell.

CellWhere retrieves localizations from UniProt and/or The Gene Ontology, and retrieves interactions from the Mentha server. It graphs the resulting network to resemble a physical map of the cell, placing proteins in a way that helps biologists to hypothesize and interpret mechanistic links between their genes/proteins of interest. It produces an interactive display of the graph using the Cytoscape.js library.

A researcher can begin with one or more genes/proteins of special interest to her, or may have a longer list resulting from some screening or omics analysis. In either case:



and, if the researcher has used some other process to arrive at a gene association network in Cytoscape 3 (such as IntAct or GeneMANIA results):

Below are a few ways that you can benefit from CellWhere (there are almost certainly other uses that we have not thought of!)

The table below shows the localization procedure for 3 example queries:


For the three query genes, reviewed Swiss-Prot (UniProt) protein accessions were retrieved (in this example, gene names are queried, but other identifiers can also be used), along with localization terms from The Gene Ontology and UniProt (actually, CellWhere retrives the localization text field from UniProt and parses it into phrases). These terms (of which ~3000 are in use) are then mapped to CellWhere localizations (of which there are 50 - see "What are CellWhere's default localizations?"). The relative frequency of each CellWhere localization is calculated for a given protein.

CellWhere currently maps all Uniprot and GO localization terms that have been applied to more than 25 proteins. This covers more than 99% of all protein localization annotations (1,258,337 out of a total of 1,269,645), and includes the most frequently used 1013 of the 3812 terms that comprise the Gene Ontology Cellular Component namespace, and 422 of the 1283 terms parsed from the Uniprot Subcellular location field.

If you select "Generic" as the localization flavor then it is the frequency percentage that is used to select the localization at which the protein will be shown on the final graph. In this case, RRAD, EMILIN2, and ACTC1, would be placed at the Membrane, ECM, and Cytoplasm, respectively.

Alternatively, a ranked list of all the CellWhere localizations is consulted and the highest ranking for a given protein is chosen for that proteins location on the graph. We include several ranked list 'flavors', and by using one as a template you can create your own flavor according to your research interests (please see the front page for template and upload instructions, and also the downloadable files available from the menu). By using our muscle flavor, in the example above ACTC1 would be placed into the 'Focal adhesion' location, because the muscle flavor sets a high priority score on this location, due to it being of special interest to muscle researchers.

You are not obliged to use both the UniProt and GO localizations, and can choose either alone. The UniProt localization field is carefully (and conservatively) curated to contain the classicly known locations of a protein, whereas The Gene Ontology is aimed more towards a comprehensive listing of all locations at which a protein has been observed. Therefore, in our context, GO is useful for screening proteins against localizations of interest, whereas UniProt is useful if you are interested to know the 'typical' location of a protein.

The "spatial relation" tells CellWhere how to place a localization on the output graph. CellWhere currently supports the following spatial relations. Each CellWhere localization must be associated with one of these in the mapping file.



* "Nucleus", "Cytoplasm", "Membrane", and "Extracellular", are the primary compartments of CellWhere's visual display. A localization that is designated to one of these will not be independently labeled. Instead, proteins carrying this localization will float freely within the primary compartment - the localization will NOT be given a separate box (these boxes are referred to in Cytoscape.js as 'compound nodes').

† A location designated as "IN Nucleus", "IN Cytoplasm", "IN Membrane", or "IN Extracellular", will be given it's own box, and member proteins will be displayed within this box. The box will be located somewhere within the appropriate primary compartment (Nucleus, Cytoplasm, Membrane, or Extracellular).

‡ A location designated as "UNDER Membrane", "SURFACE Membrane", or ""ACROSS MEMBRANE", will be given it's own box. The box will be placed, respectively: touching the interior of the cell membrane, touching the exterior of the cell membrane, or traversing entirely the membrane.

When you localize proteins by annotation frequency, CellWhere uses its generic mapping file. This file maps Uniprot and GO localization terms to one or more of the 50 CellWhere localization terms shown in the following table. As described above, you can download an example mapping file by following the link on the front page, and you can create your own mappings by creating and uploading a pre-made flavor.

CellWhere localization	Spatial relation	CellWhere localization	Spatial relation
1. Acrosome	UNDER Membrane	26. Lysosome	IN Cytoplasm
2. Actin cytoskeleton	IN Cytoplasm	27. Melanosome	IN Cytoplasm
3. Amyloplast	IN Cytoplasm	28. Membrane	Membrane
4. Apoplast	SURFACE Membrane	29. Microtubule cytoskeleton	IN Cytoplasm
5. Autophagosome	IN Cytoplasm	30. Mitochondrion	IN Cytoplasm
6. Caveolae	UNDER Membrane	31. Motile parts	ACROSS Membrane
7. Cell cortex	UNDER Membrane	32. Nucleoid	IN Cytoplasm
8. Cell junction	ACROSS Membrane	33. Nucleolus	IN Nucleus
9. Cell surface	SURFACE Membrane	34. Nucleus	Nucleus
10. Cell wall	SURFACE Membrane	35. Outer membrane	IN Membrane
11. Chloroplast	IN Cytoplasm	36. Periplasm	IN Membrane
12. Cyanelle	IN Cytoplasm	37. Peroxisome	IN Cytoplasm
13. Cytoplasm	Cytoplasm	38. Plasmodesma	SURFACE Membrane
14. Cytoskeleton	IN Cytoplasm	39. Plastid	IN Cytoplasm
15. Endoplasmic reticulum	IN Cytoplasm	40. Podosome	SURFACE Membrane
16. Endosome	IN Cytoplasm	41. Proteasome	IN Cytoplasm
17. ERMES complex	IN Cytoplasm	42. Ribosome	IN Cytoplasm
18. Extracellular	Extracellular	43. Sarcomere	IN Cytoplasm
19. Extracellular matrix	IN Extracellular	44. Sarcoplasmic reticulum	IN Cytoplasm
20. Focal adhesion	UNDER Membrane	45. Spectrin cytoskeleton	UNDER Membrane
21. Gap Junction	ACROSS Membrane	46. Synapse	ACROSS Membrane
22. Glycosome	IN Cytoplasm	47. Vacuole	IN Cytoplasm
23. Golgi	IN Cytoplasm	48. Vesicle	IN Cytoplasm
24. Inner membrane	IN Membrane	49. Vesicular exosome	IN Extracellular
25. Intermediate filaments	IN Cytoplasm	50. Virion	IN Extracellular

Many proteins have been experimentally observed at more than one location and carry mutliple localization annotations in Uniprot and/or GO. Examples include proteins that shuttle between organelles, that perform different functions at different points in development or cell differentiation, or that behave differently depending on cell type.

Cellwhere now includes an advanced feature to display duplicate copies of a protein node at alternative locations. The feature can be observed by running the default query but selecting “Annotation frequency” under localization options. Duplicate nodes are connected to their parent node by a green edge labelled with a question mark.

This feature is switched on by default when using "annotation frequency" but off by default when using priority flavors. It can be modified under advanced options at the bottom of the front page. By default, when the feature is selected, an alternative location will be shown if it has a frequency score >0.33 (i.e. if >33% of the protein's Uniprot/GO annotations map to this CellWhere localization). The user can set this cut-off value as desired, but low values will be refused if they result in a too large and over-populated graph. If using the feature with a priority flavor rather than annotation frequency, the score should be adjusted accordingly (e.g. to 7000 or some other high value).

CellWhere first attributes subcellular locations, as described above, for either the uploaded gene list or an uploaded pre-made network.

CellWhere will also query the Mentha server to retrieve known interactions between query proteins, together with the score allocated by Mentha as a measure of the strength of the evidence supporting each interaction.

If the option to grow the network using Mentha is selected, then CellWhere will request additional interactors of the query proteins. It will rank these by their score and retain all of the interacting proteins up to the maximum number set by the user.

The resulting network, with subcellular locations added is then organized for visualization.

Proteins are grouped according to their location, and locations are organized graphically.

A limited vocabulary (as described above) specified in the mapping file tells CellWhere where to place each location relative to the cell and to the cell membrane (for example, the Gap Junction may be marked 'ACROSS Membrane', or the Peroxisome marked 'IN Cytoplasm').

Co-ordinates are recalculated and the network is displayed using Cytoscape.js, on top of a membrane background.