Click the + button on the input table below to create new lines.
For each line, fill in the required information (compound name and its human protein targets) about the compounds you want predictions for.
The 'Autofill' button reads the compound names and tries to retrieve their targets from DrugBank, if the drug name is recognized (a "Not found" message appears in the target fields otherwise). Warning: this will remove data previously included in the targets field.
Inputting multiple compound names in a single entry, as a semicolon-separated list, and clicking the Autofill button will result in an instance that combines the list of targets it finds for each compound in the list. Note that, although this allows for testing predictions for combinations of compounds, our models were trained with only single compound data.
Protein targets may be represented by their STRING Target IDs (e.g.: 9606.ENSP00000345659) or by their names (e.g.: CA7, for the same protein). Any field can be left blank.
Our models are trained on annotations from human proteins, so the inputs must use human protein targets.
Clicking the Make Predictions button will load a result table with the number of valid targets (those with annotations in our source data) and the prediction results for male and female mice.
If any entry has only compound names and no info for targets when clicking Make Predictions, the system will atempt to 'Autofill' the targets for those entries.
Prepare a text file (.txt or .tsv) where each line corresponds to an entry for a compound. Click the Load from File button and select the input file.
Each line should have 3 tab-separated values: The compound's name (tab) a comma-separated list of STRING IDs (tab) a comma-separated list of gene names.
Any element of the line can be left blank, as long as the tab separations are present.
After loading the file, the input table will be automatically filled with data.
The 'Autofill' button reads the compound names and tries to retrieve their targets from DrugBank, if the drug name is recognized (a "Not found" message appears in the target fields otherwise). Warning: this will remove data previously included in the targets fieds.
Our models are trained on annotations from human proteins, so the inputs must use human protein targets.
Clicking the Make Predictions button will load a result table with the number of valid targets (those with annotations in our source data) and the prediction results for male and female mice.