A Systems Pharmacology Knowledgebase: On- and Off-Target Effects of Beta-BlockersAnton Yuryev and Ilya Mazo

In silico analysis can provide valuable insights to how candidate drug compounds may modulate protein targets and can predict potential drug effects, both on- and off-target. Using MedScan, a proprietary, high-content linguistic tool, a database of protein networks associated with different diseases and small molecule effects has been compiled by extracting the information from scientific literature and further complimented with a curated collection of ‘canonical’ signaling, metabolic and toxicity pathways.

The resulting database has been interrogated to find shared pathways modulated by 19 known beta-blockers (including Propranolol, Brevibloc, Carvedilol and Atenolol). Beta blockers, also known as beta-adrenergic blocking agents, are drugs that block norepinephrine and epinephrine from binding to beta

Conclusionsn MedScan compiles a comprehensive knowledgebase of protein/drug/disease networksn Gene set/sub-network enrichment analysis assigns high scores to the biologically meaningful findings: known signaling pathways and disease indications

affected by beta-blockers n Analysis of two independent databases: (i) curated pathways and (ii) disease sub-networks, cross-validates the relevance of beta-blockers to effects on

inflammation and kidney disease.n Several beta-blockers may have side effect of diabetes and hyperalgesia, two conditions not found in the literature to be associated with these drugs.

MethodMedScan Technology works with abstracts, articles and other types of scientific text. It recognizes proteins, chemicals, cell processes, diseases, and other con-cepts using proprietary dictionaries to capture relationships between them such as binding, regulation, protein modification, and others. Curation of the extracted data ensures high quality of Pathway Studio databases. Ariadne databases

MedScan® (Dictionaries & Rules)

MedScan extracts relationships from text in three major steps:

n Entities are recognized in a sentence using dictionaries

n Algorithms perform noun phrase normalization decomposing complicated sentences into primitive semantic blocks

n Patten matching algorithms find semantic blocks describing biological interactions and records them in XML files classifying them according to predesigned ontologies

Results

Abstract

Beta-blockers are suggested to be effective in inflammation and kidney disease. Beta-blocker compounds inhibit (activate) proteins activated (inhibited) in the disease.

Suggested side-effects for beta-blockers Compounds activate (inhibit) proteins activated (inhibited) in the toxicity.

Workflow 1:

Top-ranking pathways as identified by enrichment analysis (Fisher exact test) of beta-blocker protein targets vs. Ariadne collection of 300+ signal transduction, metabolic and toxicity pathways.

-protein

-drug

-disease

Beta-blocker protein targets are highlighted in red.

Workflow 2:

Top-ranking disease-centric protein sub-networks where beta blocker compounds modulate proteins affected in a disease/toxicity. Identified by searching the Ariadne database of 3000+ disease/toxicity networks for the “signed” overlap with the list of beta-blocker protein targets.• Information extracted by MedScan:

– Entities: e.g..protein, small molecule,etc – Type: e.g.. binds, inhibits– Relation: e.g.. Axin -- beta-catenin (relation: Binding) Axin --| GSK-3beta (relation: Regulation)

– Effect: e.g..negative, positive– Context (tissue, organ, cell type, cell line, organism) e.g., tissue: amygdala

Example sentence:

“Axin binds beta-catenin and inhibits GSK-3beta in amygdala.”

Monday, June 14, 2010

Workflow Overview

receptors on nerves. Used to treat conditions such as cardiac arrhythmias, heart attack, and hypertension, they are also known to cause a wide variety of side-effects.

Using this information, we predicted anti-inflammatory and prevention drug effects for beta-blockers in renal failure. Although many of these predic-tions have been confirmed by clinical studies of several beta-blockers, we predicted additional effects for diabetes and hyperplasia that were not previously reported. These effects include both harmful toxicities, as well as novel disease indications. This analysis provides a framework to generate and test hypotheses of mechanism of drug action that can be used to refine further experiments and interpret experimental design.

have been created with MedScan from the entire PubMed and over 60 full text journals. As PubMed grows, regular updates keep the databases refreshed.

An integral part of Pathway Studio, MedScan can be used to update database daily. Additional public and proprietary sources can also be also imported to customize databases. Database customizations such as organism or topic- specific datasets can also be ordered from Ariadne.

Vascular inflammation pathway

Canonical pathway for kidney toxicity

Angiotensin receptor, Tachykinin receptor and Adrenergic receptor signal transduction pathways.

Workflow 1: 1. Start with the list of all known beta-blockers (19 drugs and compounds).2. Identify all downstream protein targets for compounds inhibited by at least

2 beta-blockers (77) and activated by at least 2 beta-blockers (27).3. Identify signaling and toxicity pathways enriched with beta-blocker targets.

Workflow 2: 1. Find disease-centric protein sub-networks where beta-blocker targets are either

inhibited or activated in the disease. 2. Rank disease based on number of targets whose synergistic effects were considered

“side-effects”, and opposing effects were considered possible “indications”.