Extending the small molecule similarity principle to all ... · Biological processes Interactome...
Transcript of Extending the small molecule similarity principle to all ... · Biological processes Interactome...
Patrick Aloyhttps://sbnb.irbbarcelona.org
Challenges within and between Omics data integrationNovember 19, 2018
Extending the small molecule similarity principle to all levels of biology
The ‘omics’ revolution is not yielding better drugs
�2Organizing bioactivity data
— High attrition rates in pre-clinical and clinical stages of drug discovery
— More investment did not translate into more drugs approved
— Expensive experiments are needed to elucidate the bioactivity of small molecules
— Data are sparse and dirty
— Chemistry and biology are disconnected
Dise
ase
-gene c
orre
latio
n
Disease
models P
harm
acokinetic
s
Pharm
acodyn
am
ics
DIS
EA
SE
TH
ER
AP
YM
o le c ule
s
Co
mp
lexity
2k molecules
1M m
olecules
Bioinformatics is heaven (cheminformatics is hell)
�3
— Proprietary data— Monolithic entities— No architecture— Structure ¿? Function— Optimized by people— 80M commercial molecules— Chaotic databases
— Public data— Building blocks (20 aa)— Domain architecture— Sequence - Structure - Function— Optimized by evolution— 20k proteins in Human— Beautiful databases
— Small molecules— Proteins
Organizing bioactivity data
The Chemical Checker
�4
— 1M bioactive molecules— 25 data types, from
chemistry to the clinics— Major small molecule
databases are integrated— State-of-the-art machine
learning— Possibly applicable to
“any” molecule (i.e. 80M)— Automated and flexible
3D
fingerprints Scaffolds
Structural
keys
Physico-
chemistry
Metabolic
genes Crystals
Small mol.
roles
Small mol.
pathways
Signaling
pathways
Biological
processes Interactome
Gene
expression
Cancer
cell lines
Chemical
genetics Morphology
Cell
bioassays
Therapeutic
areas Indications Side effects
Diseases and
toxicology
Drug-drug
interactions
Chemistry
Targets
Networks
Cells
Clinics
2D
fingerprints
Binding
Mechanism
of action
HTS
bioassays
Organizing bioactivity data
1M
500k
300k
10k
3k
Chemistry
�5
— What is the 2D structure of the molecule?
— And its 3D structure?— What scaffolds
(chemotypes, synthetic families)?
— Size, molecular weight, charge, lipophilicity, drug-likeness…
3D
fingerprints Scaffolds
Structural
keys
Physico-
chemistry
Metabolic
genes Crystals
Small mol.
roles
Small mol.
pathways
Signaling
pathways
Biological
processes Interactome
Gene
expression
Cancer
cell lines
Chemical
genetics Morphology
Cell
bioassays
Therapeutic
areas Indications Side effects
Diseases and
toxicology
Drug-drug
interactions
Chemistry
Targets
Networks
Cells
Clinics
2D
fingerprints
Binding
Mechanism
of action
HTS
bioassays
Organizing bioactivity data
— What the taxonomist does
— What the computer scientist does
— What the organic chemist does
— What the physicist does
Communicating chemistry to the computer
�6Organizing bioactivity data
Targets
�7
— Mode of action (when available), inhibition/activation
— Drug metabolizing enzymes, transporters and carriers
— Crystalized small molecules in the PDB, structural family of receptors
— Binding assays in the literature
— HTS (chemogenomics) binding and functional assays
3D
fingerprints Scaffolds
Structural
keys
Physico-
chemistry
Metabolic
genes Crystals
Small mol.
roles
Small mol.
pathways
Signaling
pathways
Biological
processes Interactome
Gene
expression
Cancer
cell lines
Chemical
genetics Morphology
Cell
bioassays
Therapeutic
areas Indications Side effects
Diseases and
toxicology
Drug-drug
interactions
Chemistry
Targets
Networks
Cells
Clinics
2D
fingerprints
Binding
Mechanism
of action
HTS
bioassays
Organizing bioactivity data
Networks
�8
— Popular bioactivity ontology
— Metabolic pathways (metabolites + drugs)
— Signaling cascades of the targets
— Biological processes of the targets
— Neighbors of the targets in large biological networks (interactomes)
— (Tissues where the targets are expressed)
3D
fingerprints Scaffolds
Structural
keys
Physico-
chemistry
Metabolic
genes Crystals
Small mol.
roles
Small mol.
pathways
Signaling
pathways
Biological
processes Interactome
Gene
expression
Cancer
cell lines
Chemical
genetics Morphology
Cell
bioassays
Therapeutic
areas Indications Side effects
Diseases and
toxicology
Drug-drug
interactions
Chemistry
Targets
Networks
Cells
Clinics
2D
fingerprints
Binding
Mechanism
of action
HTS
bioassays
Organizing bioactivity data
Cells
�9
— Transcriptional response in cell lines (LINCS)
— Growth inhibition in cancer cell lines (NCI60)
— Growth inhibition in a panel of yeast mutants (equivalent to genetic interactions)
— Changes in morphology, measured with a cell-painting assay
3D
fingerprints Scaffolds
Structural
keys
Physico-
chemistry
Metabolic
genes Crystals
Small mol.
roles
Small mol.
pathways
Signaling
pathways
Biological
processes Interactome
Gene
expression
Cancer
cell lines
Chemical
genetics Morphology
Cell
bioassays
Therapeutic
areas Indications Side effects
Diseases and
toxicology
Drug-drug
interactions
Chemistry
Targets
Networks
Cells
Clinics
2D
fingerprints
Binding
Mechanism
of action
HTS
bioassays
Organizing bioactivity data
Clinics
�10
— Therapeutic areas (ATC codes)
— Indications (disease terms)
— Side effects in drug package labels
— Drug-drug interactions— (Pharmacogenomics)— (General toxicology)
3D
fingerprints Scaffolds
Structural
keys
Physico-
chemistry
Metabolic
genes Crystals
Small mol.
roles
Small mol.
pathways
Signaling
pathways
Biological
processes Interactome
Gene
expression
Cancer
cell lines
Chemical
genetics Morphology
Cell
bioassays
Therapeutic
areas Indications Side effects
Diseases and
toxicology
Drug-drug
interactions
Chemistry
Targets
Networks
Cells
Clinics
2D
fingerprints
Binding
Mechanism
of action
HTS
bioassays
Organizing bioactivity data
A few (ahem) straight applications to biopharma…
�11Organizing bioactivity data
Characterization of chemical collections
�12
— A front-end to rapidly characterize chemicals
— Global view of the chemical and biological space
— Popularity, singularity and mappability scores help contextualize compounds
Organizing bioactivity data
�13Organizing bioactivity data
Proparacaine Propantheline
Scaffolds
3D fingerprints
Mechs. of action
Mechs. of action
Pathways
Gene expression
Cancer cell lines
Therap. areas
Indications
Side effects
1
1
1
1
1
1
2
2
2
2
22
3
3
3
3
33
4
4
4
4
4
4
5
5 5
55
5
6
6
6
6
6
6
Erlotinib Vandetanib Gefitinib Ponatinib Semaxanib Dasatinib
1 2 3 4Rimantadine Tetracaine
1
1
1
11
2
2
2
2
23
3
3
3
34
44
4
4
Complex queries to compound libraries— Similar MoA/indication, diverse chemistry and pharmacogenomics
— Different therapeutic areas, similar gene expression and side effects
Proparacaine Propantheline
Scaffolds
3D fingerprints
Mechs. of action
Mechs. of action
Pathways
Gene expression
Cancer cell lines
Therap. areas
Indications
Side effects
1
1
1
1
1
1
2
2
2
2
22
3
3
3
3
33
4
4
4
4
4
4
5
5 5
55
5
6
6
6
6
6
6
Erlotinib Vandetanib Gefitinib Ponatinib Semaxanib Dasatinib
1 2 3 4Rimantadine Tetracaine
1
1
1
11
2
2
2
2
23
3
3
3
34
44
4
4
Large-scale target prediction
�14
— Based on chemical and biological similarities
Target 1 > Target 2 > Target 3
Ranked list of similar molecules in different chemical and biological spaces
Organizing bioactivity data
ChemistryBinding
NetworksCells
Clinics
Based onchemistry
Addingphenotypicdata
Addingchemogenomicsdata
BromperidolBenperidol
Spiperone
F
N
O
Chemical sim.
DRD2Prediction
(2018)
Knowledge(bef. 2016) Tacedinaline
HDAC1
HDAC2
HDAC5 HDAC6
HDAC8
Cross-pharmacology
AC1061JE Nortryptiline HRH1Yeast chemical genetics
Chemical prediction Chemogenomic prediction
Phenotypic prediction
Clinical trial toxicity failures
�15
— Flexible knowledge-based clinical predictors
Organizing bioactivity data
Black-box predictor Based on toxicity panels Based on liable targets
Clinical trial toxicity failures
�16
— Flexible knowledge-based clinical predictors
Organizing bioactivity data
Black-box predictor Based on toxicity panels Based on liable targets
BIA10-2474
VX-745
Tozadenant
CH3
N
NN
N+
O
O-
Cl Cl
F
F
NN
N
O
S
CH3
CH3
NN
N
NOHOH
O
O
S
Concluding remarks (wrap-up)
�17Organizing bioactivity data
Drug discovery is highly inefficient and secretive, and current computational tools are insufficient and isolated
Miscellanea of Nat Rev Drug Discov articles
Drug discovery pipeline
The accumulation of omics data is not yielding better drugs
Duran-Frigola et al, Curr Opinion Syst Biol (2018)
Computational tools are focused on curating and organizing the data
The Omics revolution
Structural Bioinformatics & Network Biology group
�18Organizing bioactivity data
Miquel Duran-Frigola Carles Pons Eduard Pauls Sergi Bayod Francesco Sirci Martino Bertoni Lídia Mateo Csaba Fehér Adrià Fernández-Torras Víctor Alcalde Oriol Guitart
We are looking for postdocs !!
Patrick Aloyhttps://sbnb.irbbarcelona.org
Challenges within and between Omics data integrationNovember 19, 2018
Extending the small molecule similarity principle to all levels of biology