Polypharmacology - NBIC April 20, 2011

Bioinformatics Meets Systems Biology for Early Stage Drug

Discovery

Philip E. BourneUniversity of California San Diego

[email protected]://www.sdsc.edu/pb

NBIC – April 20, 2011

Big Questions in the Lab1. Can we improve how science is

disseminated and comprehended?

2. What is the ancestry of the protein structure universe and what can we learn from it?

3. Are there alternative ways to represent proteins from which we can learn something new?

4. What really happens when we take a drug?

5. Can we contribute to the treatment of neglected {tropical} diseases?

What Really Happens When We Take a Drug?

• If we knew the answer we could:

– Contribute to the design of improved drugs with minimal side effects

– Contribute to how existing drugs and NCEs might be repositioned

Motivation

Why We Think This is Important

• Ehrlich’s philosophy of magic bullets targeting individual chemoreceptors has not been realized in most cases – witness the recent success of big pharma

• Stated another way – The notion of one drug, one target, to treat one disease is a little naïve in a complex system

Motivation

Polypharmacology - One Drug Binds to Multiple Targets

• Tykerb – Breast cancer

• Gleevac – Leukemia, GI cancers

• Nexavar – Kidney and liver cancer

• Staurosporine – natural product – alkaloid – uses many e.g., antifungal antihypertensive

Collins and Workman 2006 Nature Chemical Biology 2 689-700Motivation

We Are Developing a Theoretical Approach to Address This Fundamental Shift in

Thinking

• Involves the fields of:

– Structural bioinformatics– Cheminformatics – Biophysics– Systems-level biology – Pharmaceutical chemistry

Our Approach

Our Approach

• We characterize a known protein-ligand binding site from a 3D structure (primary site) and search for similar sites (secondary sites) on a proteome wide scale independent of global structure similarity

• We try an integrated approach to understand the implications of drug binding to multiple sites

Our Approach

Which Means …

• We could perhaps find alternative binding sites (off-targets) for existing pharmaceuticals and NCEs?

• If we can make this high throughput we could rationally explore a large network of protein-ligands interactions

Our Approach

What Have These Off-targets and Networks Told Us So Far?

Some Examples…1. Nothing2. A possible explanation for a side-effect of a drug

already on the market (SERMs - PLoS Comp. Biol., 2007 3(11) e217)

3. A possible repositioning of a drug (Nelfinavir) to treat a completely different condition (under review)

4. A multi-target/drug strategy to attack a pathogen (TB-drugome PLoS Comp Biol 2010 6(11): e1000976)

5. The reason a drug failed (Torcetrapib - PLoS Comp Biol 2009 5(5) e1000387)

6. How to optimize a NCE (NCE against T. Brucei PLoS Comp Biol. 2010 6(1): e1000648)

Our Stories

Need to Start with a 3D Drug-Receptor Complex - The PDB Contains Many

ExamplesGeneric Name Other Name Treatment PDBid

Lipitor Atorvastatin High cholesterol 1HWK, 1HW8…

Testosterone Testosterone Osteoporosis 1AFS, 1I9J ..

Taxol Paclitaxel Cancer 1JFF, 2HXF, 2HXH

Viagra Sildenafil citrate ED, pulmonary arterial hypertension

1TBF, 1UDT, 1XOS..

Digoxin Lanoxin Congestive heart failure

1IGJ

Computational Methodology

Nu

mb

er o

f re

leas

ed e

ntr

ies

Year:

A Quick Aside – RCSB PDB Pharmacology/Drug View 2010-2011

• Establish linkages to drug resources (FDA, PubChem, DrugBank, ChEBI, BindingDB etc.)

• Create query capabilities for drug information

• Provide superposed views of ligand binding sites

• Analyze and display protein-ligand interactions

Drug Name Asp

Aspirin

Has Bound Drug% Similarity to Drug Molecule 100

Mockups of drug view features

RCSB PDB Ligand View RCSB PDB Team

A Reverse Engineering Approach to Drug Discovery Across Gene FamiliesCharacterize ligand binding site of primary target (Geometric Potential)

Identify off-targets by ligand binding site similarity(Sequence order independent profile-profile alignment)

Extract known drugs or inhibitors of the primary and/or off-targets

Search for similar small molecules

Dock molecules to both primary and off-targets

Statistics analysis of docking score correlations

…

Computational MethodologyXie and Bourne 2009 Bioinformatics 25(12) 305-312

• Initially assign C atom with a value that is the distance to the environmental boundary

• Update the value with those of surrounding C atoms dependent on distances and orientation – atoms within a 10A radius define i

0.2

0.1)cos(

0.1

i

Di

PiPGP

neighbors

Conceptually similar to hydrophobicity or electrostatic potential that is dependant on both global and local environments

Characterization of the Ligand Binding Site - The Geometric Potential

Xie and Bourne 2007 BMC Bioinformatics, 8(Suppl 4):S9Computational Methodology

Discrimination Power of the Geometric Potential

0

0.5

1

1.5

2

2.5

3

3.5

4

0 11 22 33 44 55 66 77 88 99

Geometric Potential

binding site

non-binding site

• Geometric potential can distinguish binding and non-binding sites

100 0

Geometric Potential Scale

Computational Methodology Xie and Bourne 2007 BMC Bioinformatics, 8(Suppl 4):S9

For Residue Clusters

Local Sequence-order Independent Alignment with Maximum-Weight Sub-Graph Algorithm

L E R

V K D L

L E R

V K D L

Structure A Structure B

• Build an associated graph from the graph representations of two structures being compared. Each of the nodes is assigned with a weight from the similarity matrix

• The maximum-weight clique corresponds to the optimum alignment of the two structures

Xie and Bourne 2008 PNAS, 105(14) 5441Computational Methodology

Similarity Matrix of Alignment

Chemical Similarity• Amino acid grouping: (LVIMC), (AGSTP), (FYW), and

(EDNQKRH)• Amino acid chemical similarity matrix

Evolutionary Correlation• Amino acid substitution matrix such as BLOSUM45• Similarity score between two sequence profiles

ia

i

ib

ib

i

ia SfSfd

fa, fb are the 20 amino acid target frequencies of profile a and b, respectivelySa, Sb are the PSSM of profile a and b, respectively Computational Methodology Xie and Bourne 2008 PNAS, 105(14) 5441








Our Stories

Selective Estrogen Receptor Modulators (SERM)

• One of the largest classes of drugs

• Breast cancer, osteoporosis, birth control etc.

• Amine and benzine moiety

Side Effects - The Tamoxifen Story PLoS Comp. Biol., 2007 3(11) e217

Adverse Effects of SERMs

cardiac abnormalities

thromboembolic disorders

ocular toxicities

loss of calcium homeostatis

?????


Ligand Binding Site Similarity Search On a Proteome Scale

• Searching human proteins covering ~38% of the drugable genome against SERM binding site

• Matching Sacroplasmic Reticulum (SR) Ca2+ ion channel ATPase (SERCA) TG1 inhibitor site

• ER ranked top with p-value<0.0001 from reversed search against SERCA

ER

0 20 40 60 80

0.0

00

.02

0.0

40

.06

Score

De

nsi

ty

SERCA


Structure and Function of SERCA

• Regulating cytosolic calcium levels in cardiac and skeletal muscle

• Cytosolic and transmembrane domains

• Predicted SERM binding site locates in the TM, inhibiting Ca2+ uptake


Binding Poses of SERMs in SERCA from Docking Studies

• Salt bridge interaction between amine group and GLU

• Aromatic interactions for both N-, and C-moiety

6 SERMS A-F (red)


Off-Target of SERMscardiac abnormalities

thromboembolic disorders

ocular toxicities

loss of calcium homeostatis

SERCA !

in vivo and in vitro Studies TAM play roles in regulating calcium uptake activity of cardiac SR TAM reduce intracellular calcium concentration and release in the platelets Cataracts result from TG1 inhibited SERCA up-regulation EDS increases intracellular calcium in lens epithelial cells by inhibiting SERCA

in silico Studies Ligand binding site similarity Binding affinity correlation PLoS Comp. Biol., 2007 3(11) e217

The Challenge

• Design modified SERMs that bind as strongly to estrogen receptors but do not have strong binding to SERCA, yet maintain other characteristics of the activity profile









Our Stories

Nelfinavir

• Nelfinavir may have the most potent antitumor activity of the HIV protease inhibitors

Joell J. Gills et al, Clin Cancer Res, 2007; 13(17) Warren A. Chow et al, The Lancet Oncology, 2009, 10(1)

• Nelfinavir can inhibit receptor tyrosine kinase(s)• Nelfinavir can reduce Akt activation

• Our goal: • to identify off-targets of Nelfinavir in the human

proteome• to construct an off-target binding network • to explain the mechanism of anti-cancer activity

Possible Nelfinavir Repositioning PLoS Comp. Biol., 2011 To Appear

Possible Nelfinavir Repositioning

binding site comparison

protein ligand docking

MD simulation & MM/GBSABinding free energy calculation

structural proteome

off-target?

network construction & mapping

drug target

Clinical Outcomes

1OHR


Binding Site Comparison

• 5,985 structures or models that cover approximately 30% of the human proteome are searched against the HIV protease dimer (PDB id: 1OHR)

• Structures with SMAP p-value less than 1.0e-3 were retained for further investigation

• A total 126 structures have significant p-values < 1.0e-3


Enrichment of Protein Kinases in Top Hits

• The top 7 ranked off-targets belong to the same EC family - aspartyl proteases - with HIV protease

• Other off-targets are dominated by protein kinases (51 off-targets) and other ATP or nucleotide binding proteins (17 off-targets)

• 14 out of 18 proteins with SMAP p-values < 1.0e-4 are protein kinases


p-value < 1.0e-3

p-value < 1.0e-4

Distribution of Top Hits on the Human Kinome

Manning et al., Science, 2002, V298, 1912


1. Hydrogen bond with main chain amide of Met793 (without it 3700 fold loss of inhibition)2. Hydrophobic interactions of aniline/phenyl with gatekeeper Thr790 and other residues

H-bond: Met793 with quinazoline N1 H-bond: Met793 with benzamidehydroxy O38

EGFR-DJKCo-crys ligand

EGFR-Nelfinavir

Interactions between Inhibitors and Epidermal Growth Factor Receptor (EGFR) – 74% of binding site resides

are comparable

DJK = N-[4-(3-BROMO-PHENYLAMINO)-QUINAZOLIN-6-YL]-ACRYLAMIDE

Off-target Interaction Network

Identified off-target

Intermediate protein

Pathway

Cellular effect

Activation

Inhibition


Other Experimental Evidence to Show Nelfinavir inhibition on EGFR, IGF1R, CDK2 and Abl is Supportive

The inhibitions of Nelfinavir on IGF1R, EGFR, Akt activitywere detected by immunoblotting.

The inhibition of Nelfinavir on Akt activity is less than a known PI3K inhibitor

Joell J. Gills et al.Clinic Cancer Research September 2007 13; 5183

Nelfinavir inhibits growth of human melanoma cellsby induction of cell cycle arrest

Nelfinavir induces G1 arrest through inhibitionof CDK2 activity.

Such inhibition is not caused by inhibition of Aktsignaling.

Jiang W el al. Cancer Res. 2007 67(3)

BCR-ABL is a constitutively activated tyrosine kinase that causes chronic myeloid leukemia (CML)Druker, B.J., et al New England Journal of Medicine, 2001. 344(14): p. 1031-1037

Nelfinavir can induce apoptosis in leukemia cells as a single agentBruning, A., et al. , Molecular Cancer, 2010. 9:19

Nelfinavir may inhibit BCR-ABL


Summary

• The HIV-1 drug Nelfinavir appears to be a broad spectrum low affinity kinase inhibitor

• Most targets are upstream of the PI3K/Akt pathway

• Findings are consistent with the experimental literature

• More direct experiment is needed









Our Stories

The Future as a High Throughput Approach…..

The Problem with Tuberculosis

• One third of global population infected• 1.7 million deaths per year• 95% of deaths in developing countries• Anti-TB drugs hardly changed in 40 years• MDR-TB and XDR-TB pose a threat to

human health worldwide• Development of novel, effective and

inexpensive drugs is an urgent priority

Repositioning - The TB Story

The TB-Drugome

1. Determine the TB structural proteome

2. Determine all known drug binding sites from the PDB

3. Determine which of the sites found in 2 exist in 1

4. Call the result the TB-drugome

A Multi-target/drug Strategy Kinnings et al 2010 PLoS Comp Biol 6(11): e1000976

1. Determine the TB Structural Proteome

284

1, 446

3, 996 2, 266

TB proteome

homology

models

solve

d

structu

res

• High quality homology models from ModBase (http://modbase.compbio.ucsf.edu) increase structural coverage from 7.1% to 43.3%


2. Determine all Known Drug Binding Sites in the PDB

• Searched the PDB for protein crystal structures bound with FDA-approved drugs

• 268 drugs bound in a total of 931 binding sites

No. of drug binding sites

MethotrexateChenodiol

AlitretinoinConjugated estrogens

DarunavirAcarbose


Map 2 onto 1 – The TB-Drugomehttp://funsite.sdsc.edu/drugome/TB/

Similarities between the binding sites of M.tb proteins (blue), and binding sites containing approved drugs (red).

From a Drug Repositioning Perspective

• Similarities between drug binding sites and TB proteins are found for 61/268 drugs

• 41 of these drugs could potentially inhibit more than one TB protein

No. of potential TB targets

raloxifenealitretinoin

conjugated estrogens &methotrexate

ritonavir

testosteronelevothyroxine

chenodiol


Top 5 Most Highly Connected Drugs

Drug Intended targets Indications No. of connections TB proteins

levothyroxine transthyretin, thyroid hormone receptor α & β-1, thyroxine-binding globulin, mu-crystallin homolog, serum albumin

hypothyroidism, goiter, chronic lymphocytic thyroiditis, myxedema coma, stupor

14

adenylyl cyclase, argR, bioD, CRP/FNR trans. reg., ethR, glbN, glbO, kasB, lrpA, nusA, prrA, secA1, thyX, trans. reg. protein

alitretinoin retinoic acid receptor RXR-α, β & γ, retinoic acid receptor α, β & γ-1&2, cellular retinoic acid-binding protein 1&2

cutaneous lesions in patients with Kaposi's sarcoma 13

adenylyl cyclase, aroG, bioD, bpoC, CRP/FNR trans. reg., cyp125, embR, glbN, inhA, lppX, nusA, pknE, purN

conjugated estrogens estrogen receptor

menopausal vasomotor symptoms, osteoporosis, hypoestrogenism, primary ovarian failure

10

acetylglutamate kinase, adenylyl cyclase, bphD, CRP/FNR trans. reg., cyp121, cysM, inhA, mscL, pknB, sigC

methotrexatedihydrofolate reductase, serum albumin

gestational choriocarcinoma, chorioadenoma destruens, hydatidiform mole, severe psoriasis, rheumatoid arthritis

10

acetylglutamate kinase, aroF, cmaA2, CRP/FNR trans. reg., cyp121, cyp51, lpd, mmaA4, panC, usp

raloxifeneestrogen receptor, estrogen receptor β

osteoporosis in post-menopausal women 9

adenylyl cyclase, CRP/FNR trans. reg., deoD, inhA, pknB, pknE, Rv1347c, secA1, sigC

Vignette within Vignette

• Entacapone and tolcapone shown to have potential for repositioning

• Direct mechanism of action avoids M. tuberculosis resistance mechanisms

• Possess excellent safety profiles with few side effects – already on the market

• In vivo support• Assay of direct binding of entacapone and tolcapone

to InhA reveals a possible lead with no chemical relationship to existing drugs

Kinnings et al. 2009 PLoS Comp Biol 5(7) e1000423

Summary from the TB Alliance – Medicinal Chemistry

• The minimal inhibitory concentration (MIC) of 260 uM is higher than usually considered

• MIC is 65x the estimated plasma concentration

• Have other InhA inhibitors in the pipeline

Repositioning - The TB Story Kinnings et al. 2009 PLoS Comp Biol 5(7) e1000423








Our Stories

The Future as a Dynamical Network Approach

Drug Failure - The Torcetrapib Story PLoS Comp Biol 2009 5(5) e1000387

Cholesteryl Ester Transfer Protein (CETP)

• collects triglycerides from very low density or low density lipoproteins (VLDL or LDL) and exchanges them for cholesteryl esters from high density lipoproteins (and vice versa)

• A long tunnel with two major binding sites. Docking studies suggest that it possible that torcetrapib binds to both of them.

• The torcetrapib binding site is unknown. Docking studies show that both sites can bind to torcetrapib with the docking score around -8.0.

HDLLDL

CETP

CETP inhibitor

X

Bad Cholesterol Good Cholesterol

PLoS Comp Biol 2009 5(5) e1000387Drug Failure - The Torcetrapib Story

Computational Evaluation of Drug Off-Target Effects

Proteome

Drug binding site alignments

SMAP

Predicted drug targets

Drug and endogenous substrate binding site analysis

Competitively inhibitable targets

Inhibition simulations in context-specific model

COBRA Toolbox

Predicted causal targets and genetic risk factors

Metabolicnetwork

Scientificliterature

Tissue and biofluid localization data

Gene expression

data

Physiologicalobjectives

System exchange constraints

Flux states optimizing objective

Physiological context-specific

model

Influx

Efflux

Drug response phenotypes

Dru

g ta

rget

s

Physiologicalobjectives

Causal drug targets

All targets

336 genes1587 reactions

Plos Comp. Biol. 2010 6(9): e1000938

Summary

• Generated a large number of testable hypotheses regarding off-targets of known drugs

• A few have been tested experimentally (e.g., encapalone against InhA)

• A possible methodology for use in early stage drug discovery

Acknowledgements

Sarah Kinnings

Lei Xie

Li Xie

http://funsite.sdsc.edu

Roger ChangBernhard Palsson

Jian Wang

Polypharmacology - NBIC April 20, 2011

Education

drug information

drug binds

drug nelfinavir

drug molecule

implications of drug

drug resources fda

d drugreceptor complex

alternative binding