drug-drug interaction prediction through systems pharmacology analysis (poster)
DESCRIPTION
2013 Summit on Translational BioinformaticsTRANSCRIPT
Drug Drug Interaction Prediction through Systems Pharmacology Analysis
Xiaochen Sun1, Nicholas P. Tatonetti1Department of Biomedical Informatics, Columbia University, New York, NY
Introduction
Tatonetti Labat Columbia University Medical Center
3
THE UNIVERSITY IDENTITY
The design of the Columbia identity incorporates the core elements of well- thought-out branding: name, font, color, and visual mark. The logo was designed using the official University font, Trajan Pro, and features specific proportions of type height in relation to the visual mark. The official Colum-bia color is Columbia Blue, or Pantone 290. On a light color background, the logo can also be rendered in black, grey (60% black), Pantone 280, or Pantone 286; on a darker color background, the logo can be rendered in Pantone 290, 291, or 284, depending on which color works best with the overall design of your product, the media in which it will be reproduced, and its intended use.
Black
Pantone 286
4-color Process100% Cyan72% Magenta
White or Pantone 290 (Columbia Blue)Background: Pantone 286
For photographs, use the logo in white against a darker area, posi-tioning it either at top left/right or bottom left/right.
Tatonetti Labat Columbia University Medical Center
Model
Drug-drug interactions (DDI) is important!-play an important role in explaining drug side effects and facilitating drug design.
DDI mechanism-integrative pathway analysis advances our understanding of DDI by revealing the molecular basis of drug action.
DDI prediction-The emergence of pharmacogenomics knowledge bases (PharmGKB and DrugBank) and pathway databases (KEGG), provide an opportunity to improve DDI prediction methods
Method Result
Conclusions and discussion
1. Get drug, gene, pathway data from KEGG and Drugbank.
2. Integrate data
3. Find pathway pairs that share at least one gene.
4. Download detail pathway files from KEGG and convert them to text format.*The pathway are not represented as a complete path but individual interactions.
From To Relationship356 355 activation8743 8793 activation5566 57 inhibiton
... ... ...
... ... ...
5. Construct pathways as python sets with genes comprising them.*Each pathway has multiple path leading from the first upstream gene to the sharing gene for the two pathways.
hsa04012 [ [['374', '1956', '1398', '25'], ['374', '1956', '1399', '25']], [['2069', '1956', '1398', '25'], ['2069', '1956', '1399', '25']], [['685', '1956', '1398', '25'], ['685', '1956', '1399', '25']]]hsa04360 [['1945', '2046', '25'], ['1945', '285220', '25']], [['2050', '25']], [['2046', '25']], [['2047', '25']], [['2044', '25']], [['2045', '25']], [['2042', '25']], [['2043', '25']], [['2041', '25']], [['2048', '25']]]
6. Link drugs to the genes in pathways pairs and produce DDI pairs. *If the drug A targets one gene in one of the pathwaypair. Drug B targets one gene in the other pathway of the pair. Then drug A and drug B forms a DDI.
: Genes in the left pathway : Interactions between genes
: Genes in the right pathway : Drug interact with genes
: Genes shared by both pathways
Assumption:
If two pathways come across each other, the drug that target the upstream of the two pathways will likely to interact each other and form a DDI pair.
Explanation:In this case, the left pathway is targeted by drug A and right pathway is targeted by drug B. Since drug effect signal pass downstream, the shared genes by the two pathways get mixed signals from two drugs. This may cause unexpected changes in the drug response. The two drugs here are hypothesized to form a drug-drug interaction pair.
A
B
Department of Biomedical Informatics Columbia University Medical Center 622 West 168th St. VC5 New York, NY 10032
Contacts:Xiaochen Sun Nicholas P [email protected] [email protected] 212-305-2055
Conclusions:
1.Two random drugs will have somewhat interactions more than half of the time. Drug of similar class tend to interact with each other.
2.The incompleteness of pathway data is limiting the research result.
3.Standardization is still a big issue in data integration.
Future Work:
1. Filter results, improve model.
2. Combine EHR clinical data to confirm the hypothesis.
3. Look into other discoveries from the results.
Limitation:1.Data from KEGG and Drugbank are biased and sparse. Model is not determinant.
2.Didn’t filter compounds (not drugs) from the database.
Pathway Gene Gene ID Drug Drug ID
hsa00010 HK1 3098 Alpha-‐D-‐Glucose-‐6-‐Phosphate DB02007
hsa00010 HK1 3098 Beta-‐D-‐Glucose DB02379
hsa00010 HK1 3098 Adenosine-‐5'-‐Diphosphate DB03431
hsa00010 GPI 2821 Erythose-‐4-‐Phosphate DB03937
hsa00010 FBP1 2203 Adenosine monophosphate DB00131
hsa00010 GAPDH 2597 NADH DB00157
hsa00010 GAPDH 2597 NicoInamide-‐Adenine-‐DinucleoIde DB01907
hsa00010 PKM2 5315 Pyruvic acid DB00119
hsa00010 PKLR 5313 Pyruvic acid DB00119
gene_id a_pathwayid b_pathwayid18 hsa00250 hsa0028025 hsa04012 hsa0411026 hsa00330 hsa0034027 hsa04012 hsa0541631 hsa00061 hsa0062032 hsa00061 hsa0062034 hsa00071 hsa0028035 hsa00071 hsa0028036 hsa00071 hsa0028038 hsa00071 hsa0007239 hsa00071 hsa0007290 hsa04060 hsa0435091 hsa04010 hsa0406094 hsa04060 hsa04350100 hsa00230 hsa05340107 hsa00230 hsa04020113 hsa00230 hsa04020124 hsa00010 hsa00071125 hsa00010 hsa00071
Drug-drug interactions121 unique drugs5202 unique DDIs
drug1 drug2
Desipramine Labetalol
Desipramine Isoetharine
Desipramine Levobunolol
Desipramine Esmolol
Desipramine Oxprenolol
Desipramine Risperidone
Desipramine Penbutolol
Desipramine Isoproterenol
Desipramine Nadolol
Atropine Danazol
Atropine Desipramine
Atropine Dibucaine
Atropine Diprenorphine
Atropine Dobutamine
5202$
3943$
2726$
143$0$
1000$
2000$
3000$
4000$
5000$
6000$
>0$ >2$ >4$ >8$
Repeat&Times&
>0$
>2$
>4$
>8$
False positive rate
ROC area=0.67
Drug-Class Interactions
True positive
rate
labetalol
Class-class Interactions
N06 (Psychoanaleptics) tends to interact with nerve system drugsL01 (Antineoplastic agents) tends to interact with antineoplastic agents
Wednesday, April 3, 13