SwissCellPrediction

How to use SwissCellPrediction



Welcome to the documentation page of SwissCellPrediction. Here, you will find explanations and annotated images of the SwissCellPrediction input and results pages. To do a reverse virtual screening with SwissCellPrediction, you first need to input a query molecule.

Input query molecule

The query molecule can be inputted either by (1) directly writing its SMILES in the left-hand dedicated text box or (2) using the molecular editor on the right, which enables user-friendly drawing of chemical structures. The sketcher also allows to import compounds (a) from a local file or remote file through the network, (b) by its chemical or common name. Additionally, it is synchronized with the SMILES text box and has copy/paste features, providing an overall easy way to modify molecules. A few anticancer small molecules examples can be selected from the drop-down menu. A “clear” button allows the user to delete previously inputted molecules. Once the query set, the user can run the computation by clicking on the red button: “Predict cell line targets”.

Ligand-based reverse screening results

Cell lines are ranked according to their probability of sensitivity represented as a green bar. The probabilities are based on logistic regression scores obtained from the 2D and 3D similarity computation of all the known actives of each cell line with the query molecule. The table also gives Information regarding the species from which the cell line was cultivated, as well as its tissue of origin and associated disease. The last column shows the numbers of known actives that reached a similarity score of at least 0.4 in 2D (Tanimoto coefficient on FP2 fingerprints) and 0.8 in 3D (Manhattan-based similarity on ES5D vectors). The structures of the known actives for a given cell line are provided in another output page by clicking on its hypertext numbers. The whole table of the first output page can be sorted according to any column by clicking its header, and a text search box allows to filter the results. Advanced export options are provided by clicking on dedicated red icons. The table can thus be downloaded in various formats (PDF, CSV or Excel), copied into the clipboard as well as directly printed. The user can adjust the number of predicted cell lines to display to 15 (default), 25, 50 or all (maximum 100).

Additionally, a pie chart displayed in a box on the top-right corner shows the overall repartition of the primary diseases among the predicted cell lines, from the annotations of the Cancer Cell Line Encyclopedia. Links to more detailed descriptions of the cell lines are provided by clicking on their respective ChEMBL and Cellosaurus IDs (directing to the webportal of the database of origin). Another box on the top-left of the result page recalls the chemical structure of the query. A third box just below returns its estimated membrane permeation capacity. Red icons appear inside the box containing the structure of the query. These are interoperability icons to send the query molecule to other SwissDrugDesign CADD webtools developed by the Molecular Modeling Group of the SIB Swiss Institute of Bioinformatics:
  • SwissTargetPrediction (target icon for prediction of protein targets)
  • SwissSimilarity (twins icon for ligand-based virtual screening)
  • SwissADME (pill icon for calculation of parameters relevant for physchem, ADME and medicinal chemistry)
  • SwissBioisostere (hexagon icon for replacement of molecular fragment)
  • SwissParam (sigma icon for parametrizations of molecular docking) are reachable this way
  • SwissCellPrediction (cell icon for prediction of cell line targets)
The interoperability between SwissCellPrediction and SwissTargetPrediction is particularly interesting, as the complementarity of these tools has the potential to predict both the cell lines which are sensitive to a given query molecule, as well as the molecular mechanism involved, if the latter can be explained by protein binding.

Known actives on a predicted cancer cell line

Users can also access the ChEMBL Compound Report Card of any structure of the similar known actives by clicking on its ChEMBL ID. For every structure, an estimation of its membrane permeation capacity is indicated, as well as the presence of non-specific anticancer chemical patterns, if the compound was flagged as ubiquitous. The known actives are ranked according to their similarity obtained with the query, in descending order of value. Another red icon just above the table allows the user to export the list of known actives in CSV.

Case of ubiquitous query molecule

If the query molecule exhibits a chemical pattern associated to anticancer activity with non-specific mode of action, it is flagged as ubiquitous. In such cases, SwissCellPrediction purposefully does not perform ligand-based reverse virtual screening, because the anticancer activity is then assumed to be due to a particular isolated chemical pattern, rather than to the global similarity with a known active. Therefore, the user is redirected to a different output page, displaying a table containing a list of cell lines which have been experimentally shown to be sensitive against some other compounds harboring the same non-specific anticancer flags as the query molecule.

Other information

If you need help with the command-line, you can find documentation here.

If you need more help on the web interface, please have a look at the video tutorials.

Answers to frequently asked questions can be found in the FAQ.