Data-driven Drug Discovery for TB (D4TB)

2nd BD2K / 4th Network of BioThings Hackathon 2015

Motivation Problem Statement

Challenges - Amount of Data

• Biological data is increasing exponentially and difficult to cope up with.

​

​

Challenges - Data Analysis

• Literature on TB has increased to 75% over last 10 years

​

​

Challenges - Disparate Datasets

• Disparate and Heterogeneous data exists that needs to communicate

​

​

Objectives

• Finding alternate/adjunct TB drug therapy
• Integrate datasets to assist in TB drug discovery
• Provide an interface for biologists to interact with the data

​

Overall Approach

Data

Integration

RDF/SPARQL

Results1

Results 2

Possible

Drug Candidates

TB Drugs

Chemical Similarity

Side Effect Similarity

Approach 1

Approach 2

Approach - 1 - Linked Data driven discovery

Approach 1 - Results

Query 1 - Find associated diseases for TB drugs

Query 2 - Find drugs associated with diseases from Query 1

Approach 2 - Approach

Possible Candidates

TB Drugs

23 known TB Drugs

220,116 drug pairs

7890 drug pairs

4707 drug pairs

1109 drug pairs

19 drug pairs

SIDER side effect similarity

MACCS mol fingerprint similarity

Results2

SideEffect Similarity Score (<0.3)

Chemical Similarity (>0.75)

Results 1

Overall Results

Demo

Future directions

• Convert and integrate more biologically relevant datasets
• Clinical Trials, Genomics and Proteomics Data
• Improve user interface for biologists to explore the data
• Continue collaborations between the groups

Thank you !

Questions?