PDID Help and Tutorial

This page explains how to use Protein-Drug Interaction Database (PDID) and how to obtain result for your query. Introduction to the PDID is available here.


Table of content


Main page

The main page provides summary of the PDID resource and access to search forms and data relavant to this database.



1 the About link provides access to the page that introduces and explains PDID. This page includes information about data sources and methods that were used to derive information that is stored in PDID.

2 the Help&Tutorial link provides access to this tutorial page.

3 the Statistics section provides summary of the data that are included in PDID. The table shows the number of included drugs, proteins (unique human proteins) and protein structures (structures or structure fragments of human proteins), and the number of putative interactions. These putative interactions are predicted by three methods that are described here. The table also provides the total and median numbers of known and putative targets that are aggregated per proteins. These statistics reflect the most recent version of PDID.

4 the Search by drug name option allows searching the database for a specific drug. The list of included drugs is available here. First, select the desired drug with the help of the pull-down menu and next click GO!. A new window will be opened with the detailed information about known and putative interactions of the selected drug and each of the protein structures that are included in PDID. See results page for drug molecule.

5 the Search by PDB ID of Protein Target option allows searching the database for a specific protein. The list of included proteins is available here. First, type the protein ID and next click GO!. The protein ID is in the format of the Protein Data Bank, i.e., with 4 characters followed by one character that denotes chain identifier (e.g., 12CA_A). See results page for protein target.

6 the Search by Protein Sequence option allows searching the database for a protein most similar to the query sequence. The sequence alignment is based on BLAST. First, enter the query protein sequence in the FASTA format and set the E-value threshold for BLAST alignment, and then click GO!. See results page for protein sequence query.

7 the Datasets section provides access to the list of proteins and drugs that are included in PDID.




Back to table of content

Results page for drug molecule

This page summarizes data for the selected drug molecule. This includes links to relevant pages in the Protein Data Bank, DrugBank and BindingDB, and tabulated detailed information concerning the known and putative interactions (or lack of them) between the selected drug and each of the human proteins included in PDID. The proteins in the table are sorted by the likelihood that they interact with the selected drug, from the highest to the lowest.



1 the More information section provides links to the relevant pages in the Protein Data Bank, DrugBank and BindingDB.

2 the PDB ID column gives the identifier, using the format from Protein Data Bank (PDB), of a protein for which the interaction (or lack of it) with the given drug is summarized in the corresponding row. The ID is linked to the corresponding record in the Protein Data Bank. The proteins are sorted by the likelihood that they interact with the selected drug, from highest to lowest, which is quantified with the ILbind score.

3 the Protein Name {Synonym} column gives the name and the corresponding synonyms for a protein for which the interaction (or lack of it) with the given drug is summarized in the corresponding row. The name is linked to the page that describes results per protein, see results page for protein target. The proteins are sorted by the likelihood that they interact with the selected drug, from highest to lowest, which is quantified with the ILbind score.

4 the Sequence and structure files columns provide links to the files with the protein sequence (using the FASTA format) and protein structure (using the Protein Data Bank format).

5 the Type of Annotation column provides information whether a given protein is:
- Known to bind the selected drug. This is denoted using annotation In Complex that corresponds to the fact that the protein-drug complex was solved structurally and is available in the Protein Data Bank (the adjacent column provides link to this structure), and annotation Known to Bind that means that this protein-drug interaction was deposited into the DrugBank or BindingDB (the adjacent column provides link to the corresponding record in the DrugBank and/or BindingDB).
- Predicted to bind the selected drug. This is denoted using annotation Predicted to Bind and is based on the prediction from the ILbind method, i.e., this method has to provide sufficiently high score (>0.75) that is shown in the last column and color-coded in green.
- Predicted not to bind the selected drug. This is denoted using annotation No Interaction which means that this protein is not known and was not predicted to bind the selected drug. Note that some of these protein targets may have high prediction scores (shown in the last column and color-coded in green) from the SMAP and/or eFindSite methods, which indicates that its possible that this interaction occurs.

6 the Source Database column provides links to the corresponding information from the Protein Data Bank (PDB), DrugBank and BindingDB for proteins that are known to bind the selected drug.

7 the Sequence Similarity to Known Target column gives the sequence similarity [%] between the protein identified in the first column by the Protein Data Bank (PDB) ID and the protein identified in the BindingDB, DrugBank, or PDB databases which were used to annotate known protein-drug interactions. Similarity was measured using BLAST-based alignment. Higher value of similarity denotes a more accurate match.

8 the Binding Prediction Score column gives scores generated by the three methods used to perform the prediction of protein-drug interactions: ILbind, SMAP, and eFindSite (details are described here). The scores and color-coded where green/blue/red denotes that the corresponding protein is likely/possibly/unlikely to bind the selected drug. Among several scores that ILbind and SMAP generate, we provide one score (binding propensity for ILbind and raw score for SMAP) that was empirically shown to provide the best predictive performance [Structure 2012; 20:1815-22]. eFindSite generates only one propensity score.




Back to table of content

Results page for protein target

This page provides information concerning the known and putative protein-drug interactions for the selected protein target. This includes links to the corresponding protein sequence and structure files, visualization of the protein structure in complex with drugs that are known and that are predicted to bind the selected proteins, and tabulated detailed information concerning the known and putative protein-drug interactions. The drugs in the table are sorted by the likelihood that they interact with the selected proteins, starting with the drugs that are known to interact and following with the drugs predicted to interact, from higher to lower likelihood of interaction. Only the drugs that are known to interact or are predicted by at least one method to interact are included.



1 the top left section provides
- the name of the selected protein (with synonyms listed inside the curly brackets)
- the ID and the link to the corresponding record in the Proitein Data Bank (PDB)
- the ID and the link to the corresponding record in the UniProt database
- sequence similarity between the protein collected from PDB and the corresponding protein from the UniProt
2 the top right section shows structure of the selected protein target (using gray trace of the backbone), position of ligands that are in complex with this protein (using blue sticks; they are included in the structure file), and the predicted center of the drug molecule (using red balls) that are summarized in the 'Binding to drugs' table. The drug molecules can be identified by the drug IDs. The structure can be manipulated (rotated, zoomed, redrawn) using a mouse. The figure was drawn using the JSMOL package available from http://sourceforge.net/projects/jsmol/.
3 the Annotated as Known Target columns identify drugs that are known to interact with the selected target protein. Two types of interactions are possible: In Complex that corresponds to the fact that the protein-drug complex was solved structurally and is available in the Protein Data Bank (the adjacent column provides link to this structure), and Known to Bind that means that this protein-drug interaction was deposited into the DrugBank and/or BindingDB (the adjacent column provides link to the corresponding record in the DrugBank or BindingDB).
4 the Predicted as Target columns give scores generated by the three methods used to perform the prediction of protein-drug interactions: ILbind, SMAP, and eFindSite (details are described here). The scores and color-coded where green/blue/red denotes that the selected protein target is likely/possibly/unlikely to bind the corresponding drug. Among several scores that ILbind and SMAP generate, we provide one score (binding propensity for ILbind and raw score for SMAP) that was empirically shown to provide the best predictive performance [Structure 2012; 20:1815-22]. eFindSite generates only one propensity score.
5 the Predicted Coordinates column provides coordinates (in the coordinates system based on the structure file that is linked at the top of the page) of the predicted positions of the centers of the drug molecules. These coordinates are shown using red balls in the structure shown in the top right section of the page.
6 the Binding Summary columns show whether a given drug is know to interact with the selected protein target, and how many prediction methods (out of 3) predict a given interaction.



Back to table of content

Results page for protein sequence query

This page shows the most significant alignment between the query sequence and the structural human-like proteome, and provides information concerning the known and putative protein-drug interactions for the protein most similar to query sequence like the results page for protein target.



1 the top section shows the protein from the structural human-like proteome which is most similar to query sequence. The similarity is measured by sequence identity and E-value based on BLAST. The result of alignment is shown below.
2 the bottom section shows the ligand binding information for the protein found similar to query sequence. The information is provided in the same way as results page for protein target.

Back to table of content