README for Chapter 7
Augmenting Drug Mechanism Prediction with Text Mining
Emily K Mallory

Description:
Directory contains supplemental files for Chapter 7 of dissertation.


##########################################

Directories and Files

A. auc (directory): 
	Directory with auc results for semantic motifs in PD and PK pathways.

B. supplemental_semantic_motif_results (directory):
	Directory with semantic motif files for PD and PK pathways. Includes raw results 
	and PD/PK score comparison files.

C. Semantic Motifs for Risperidone GNBR Query (file):
	File with GNBR semantic motif query results for risperidone .


##########################################

File Descriptions


A. auc directory:

File names contain pathway and skeleton type. For example: 

PD_LINE_ProteinSmallMoleculeProtein_within_auc_lines_all.txt

refers to LINE skeleton in PD pathway where the entities are Protein -> SmallMolecule -> Protein. Theme1 in the AUC file is for the first edge and theme2 is for the second edge.

File format:
Column1: theme pair separated by ";". Format is Theme1;Theme2. Short codes used for theme descritions: see Theme Description section below. 
Column2: Number of positive motifs
Column3: Number of negative motifs
Column4: Area under the Receiver Operating Characteristic curve (AUROC or AUC)


B. supplemental_semantic_motifs directory:

Directory contains three primary types of files: those ending in "_lines_all.txt", "_collapsed.txt", and "_comparison".

As with the auc directory, filenames contain the pathway type (PD or PK) and the skeleton type (e.g., LINE) for the comparison files.

Example filenames:
edge_motifs_3_PK_within_vec_positives_lines_all.txt - detect semantic motifs
edge_motifs_3_PK_within_OUT_LINE_lines_all_comparison.txt - comparison of average semantic motifs between PD and PK for a given skeleton
edge_motifs_3_PK_within_vec_positives_lines_all_collapsed.txt - generic semantic motifs for a given pathway

File format for edge_motifs_3_(PD/PK)_within_vec_positives_lines_all.txt:
column1: skeleton type
column2: skeleton entities
column3: theme pair separated by ";". Format is Theme1;Theme2. Short codes used for theme descritions: see Theme Description section below.
column4: combined theme score
column5: pathway entity 1
column6: theme1 and score, separated by "|"
column7: pathway entity 2
column8: theme2 and score, separated by "|"
column9: pathway entity 3

File format for edge_motifs_3_(PK/PD)_within_vec_positives_lines_all_collapsed.txt:
column1: skeleton type
column2: skeleton entities
column3: theme pair separated by ";". Format is Theme1;Theme2. Short codes used for theme descritions: see Theme Description section below.
column4: count of matching semantic motifs in the "_all.txt" file
column5: average combined theme score for a combination skeleton and theme pair

File format for edge_motifs_3_(PD/PK)_within_(skeleton)_lines_all_comparison.txt:
Filename contains pathway type and skeleton.

Columns:
skeleton: skeleton entities
theme: theme pair separated by ";". Format is Theme1;Theme2. Short codes used for theme descritions: see Theme Description section below.
score:	average combined theme score for a combination skeleton and theme pair
rank change: difference in rank between PD and PK semantic motifs
score change: difference in score between PD and PK semantic motifs
percentile change: difference in percentile between PD and PK semantic motifs

C. Semantic Motifs for Risperidone GNBR Query (file):

Semantic Motif: entity type and themes
Entity1: entity1 matching the query in GNBR. Entry is a combination of database identifier and common name, separated by ";"
Dependency Path 1: max dependency path score for the theme1 in the query. Used for normalization and final score.
Entity2: entity2 matching the query in GNBR. Entry is a combination of database identifier and common name, separated by ";"
Dependency Path 2: max dependency path score for the theme2 in the query. Used for normalization and final score.
Entity3: entity3 matching the query in GNBR. Entry is a combination of database identifier and common name, separated by ";"
Theme 1 Score: normalized score for dependency path 1 for theme1
Theme 2 Score: normalized score for dependency path 2 for theme2
Combined Score: sum of theme1 and theme2 score

NOTE: risperidone will show up in either entity1, 2 or 3 as MESH:D018967;risperidone or MESH:D018967.


##########################################

Theme descriptions

Reference with description of the EBC algorirthm:
Percha B, Altman RB. Learning the Structure of Biomedical Relationships from
Unstructured Text. PLoS Comput Biol. 2015 Jul 28;11(7):e1004216. doi:
10.1371/journal.pcbi.1004216. eCollection 2015 Jul. PubMed PMID: 26219079; PubMed
Central PMCID: PMC4517797.

Reference for the expanded GNBR themes:
Percha B, Altman RB. A global network of biomedical relationships derived from
text. Bioinformatics. 2018 Aug 1;34(15):2614-2624. doi:
10.1093/bioinformatics/bty114. PubMed PMID: 29490008.

Theme Codes:
A+: agonism, activation
A-: antagonism, blocking
B: binding, ligand
E+: increases expession/production
E-: decreases expression/production
E: affects expression/production (neutral)
N: inhibits
K: metabolism, pharmacokinetics
Z: enzyme activity
O: transport, channel