Augmented base pairing networks encode RNA-small molecule binding preferences

<p>Dataset used to train and validate the RNAmigos model from &quot;Augmented base pairing networks encode RNA-small molecule binding preferences&quot;.</p> <p>&nbsp;</p> <p>This will give you a cleaned up version of the data used to train the RNAmigos 1.0 models.</p> <p>&nbsp;</p> <p>If you run `python make_nice.py` you will generate a CSV file `rnamigos1_dataset.csv` which contains all the info you need.</p> <p>The script will also use DecoyFinder to generate the decoys for each pocket.</p> <p>&nbsp;</p> <p>&nbsp;</p> <p>### Pockets</p> <p>&nbsp;</p> <p>The CSV has one row for each binding pocket.</p> <p>&nbsp;</p> <p>The columns are:</p> <p>&nbsp;</p> <p>* pdbid: the PDBID this pocket belongs to</p> <p>* model_num: the model number inside the PDB we took</p> <p>* chain: the chain the pocket belongs to</p> <p>* ligand_id: the 3-letter code of the ligand (e.g. ATP) which you can look up on RCSB.org</p> <p>* ligand_resnum: the residue number of the ligand in the PDB</p> <p>* nodelist: a list of nodes separated by &#39;;&#39; in the pocket as a string in the format `&lt;node1 pdbid&gt;.&lt;node1 chain&gt;.&lt;node1 position&gt;-&lt;nucleotide type&gt;;&lt;node2 pdbid&gt;...`</p> <p>* edgelist: a list of edges separated by &#39;;&#39; in the pocket as a string in the format nodes are in the same format as above, and connected by a &#39;-&#39; char, with an additional label field. e.g. of a two edge list `1aju.A.1-1aju.A.5-CWW;1aju.A.1-1aju.A.2-B53`</p> <p>* fp_native_maccs: bit string of the MACCS for the native ligand</p> <p>* split_{k}_train: one col for all the splits we ran (k \in {0-9}) contains True if this pocket was in the train set for this split</p> <p>* split_{k}_test: one col for all the splits we ran (k \in {0-9}) contains True if this pocket was in the test set for this split</p> <p>&nbsp;</p> <p>### Decoys</p> <p>&nbsp;</p> <p>The folder `decoy_db/` has the following structure:</p> <p>&nbsp;</p> <p>```</p> <p>decoy_db</p> <p>&nbsp;&nbsp; &lt;pdbid&gt;_&lt;chain&gt;_{ligand_id}_{ligand_position}</p> <p>&nbsp;&nbsp; &nbsp; &nbsp; decoyfinder</p> <p>&nbsp;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; actives.txt</p> <p>&nbsp;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; decoys.txt</p> <p>&nbsp;&nbsp; &nbsp; &nbsp; pdb</p> <p>&nbsp;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; actives.txt</p> <p>&nbsp;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; decoys.txt</p> <p>&nbsp;</p> <p>&nbsp;</p> <p>Each `actives.txt` and `decoys.txt` is a file with one SMILES per line. &nbsp;</p> <p>&nbsp;</p> <p>`decoyfinder/` has decoys computed by DecoyFinder and the acvtives are just the native ligands.</p> <p>`pdb/` has decoys taken from other pockets in the PDB and actives are just the native ligands.</p>