Transcript MBI presentation - Molecular Bioactivity Indexing

Super fast identification and optimization of
high quality drug candidates
Our Goals

Constructing highly enriched and efficient
molecular libraries for the development of new and
selective drug-like leads

Minimizing false positives by early identification of
drug failures, resulting in reduced cost/time of drug
development
Preclinical Drug Discovery
We reduce lead identification and optimization to 1-3
months, and identify highest quality drug candidates
Competing state-of-the-art computational
drug discovery technologies in Pharma

Rules for drug-like properties (Lipinski, Veber): binary, many false
positives
 Data Mining from HTS: requires innovative algortihms
 “Similarity” searches (mostly structural) : limit innovation
 Drug-target “Docking” algorithms: at their infancy, false
positives & negatives
 ADME/Tox models: can not accurately predict a molecule’s
chance to become a drug
Our Technology: what do we do best ?
Grading drug likeness and molecular bioactivity
Drug-Target: “Molecular Bioactivity Index” (MBI)
Drug-Body: “Drug Like Index” (DLI)
ISE (Iterative Stochastic Elimination) engine
Experimental
Datasets
(drugs, Non-drugs,
agonists,
antagonists, inhibitors)
DLI
and/or
MBI
MBI and DLI
 MBI is a number that expresses the chance of a
molecule being a high affinity ligand for a
specific biological target
 DLI is a number that expresses the chance of a
molecule to become a drug
 Double focusing using MBI and DLI provides:
combined target specificity and drug-likeness
MBI and DLI can make a difference in:

High Throughput Screening

Combinatorial Synthesis

Hit to lead development

Lead optimization

Construction of Focused libraries

Molecular scaffold optimization

Selectivity optimization
Iterative Stochastic Elimination:
A new tool for optimizing highly complex problems
First prize in emerging technologies symposium of ACS
 Patent in National phase examination in several countries
PCT on the derived technology of DLI
IP
A stochastic method to determine in silico the drug like
character of molecules
 By Rayan, Goldblum, Yissum (PCT stage)
 A new provisional patent application covering the
MBI algorithm will be submitted
1-2 days
ISE for identification of high quality leads
TEST SET
TRAINING SET
Validation
INPUT
ISE Engine
Database ordered
By Bioactivity
Index
Huge Commercial
Database of
chemicals
MBI MODEL
Double focusing with MBI and DLI
Database ordered
By Bioactivity
Index
MBI
MODEL
Huge Commercial
Database of
chemicals
Few hours
Assumed high
affinity leads
2 - 4 days
Validations:
Docking, Scifinder,
“fishing” tests
DLI
Optimized leads
for in vitro and
animal tests
MODELS
 Matrix metalloproteinase-2 (MMP-2)
 Endothelin receptor
 D2- dopaminergic receptor
 DHFR
 Histaminergic receptors
 HIV-1 protease
 Cannabinoid receptor
 And others..
Current technological status:

Excellent enrichment of “actives” from “nonactives” using MBI

Excellent separation of drugs from “non-drugs”
using DLI
 Discovering molecules for a known drug target,
validated by a docking algorithm

Successful validation of MBI technology by big
Pharma
Molecular Bioactivity Index (MBI):
Fishing actives from a “bath” of “non-actives”
Mix 10 in 100,000 - find 9 in best 100, 5 in best 10
Enrichment of 5000
Drug Likeness Index (DLI):
Randomly mixing 10 Drugs + 100 Non-drugs
Enrichment of ~7
DLI vs. the Medicinal Chemist-1
DLI vs. the Medicinal Chemist-2
5 top Medicinal
chemists examined
MMP-2 as a target for POC
 Identifying high affinity ligands for Matrix
metalloproteinase-2 (MMP-2) was chosen as proof of
concept for our technology
 MMP-2 (or Gelatinase A) is involved in several types of
cancer, such as Breast cancer, Hepatocellular carcinoma,
Smooth muscle hyperplasia and possibly others
 We have large datasets for training
 Chemicals easy to purchase
 In vitro assay available
 Animal model available (murine leukemia)
 Israel Science Foundation collaboration
Typical MMP-2 actives - nanomolar
Typically - hydroxamates and sulphonamides
ISE for identification of high quality leads
MBI MODEL
For MMP-2
Zinc ordered by
MBI values
ZINC database with
2 million molecules
Picking 104 molecules with top MBI values above 30
Number of molecules
Similarity between 104 prospective MMP-2 leads and the
650 MMP-2 leads used for model construction
50
40
New
Chemical
Entities
(> 90 !)
30
20
10
Less
Similar
Similar
0
0
0.1
0.2
0.3
0.4
0.5
0.6
Tanimoto Index
0.7
0.8
0.9
1
Non-typical MMP-2 suspected nanomolar
candidates
1.00
0.04
0.02
0.09
0.04
0.04
1.00
0.16
0.04
0.26
0.02
0.16
1.00
0.07
0.14
0.09
0.04
0.07
1.00
0.12
0.04
0.26
0.14
0.12
1.00
0.08
0.15
0.08
0.06
0.15
8 of highest
diversity
were picked
0.11
0.17
0.09
0.21
0.15
Scifinder – none ever examined on any MMP
0.02
0.09
0.06
0.11
0.14
0.08
0.15
0.08
0.06
0.15
1.00
0.20
0.06
0.11
0.17
0.09
0.21
0.15
0.20
1.00
0.07
0.02
0.09
0.06
0.11
0.14
0.06
0.07
1.00
The first MMP-2 candidate inhibitors picked for purchasing and testing in the
lab are devoid of the characteristics of MMP-2 or other MMP inhibitors. These
molecules are not known to have any prior biological activity and have a very
low similarity index (Tanimoto) to each other (the highest similarities are
marked in yellow in the matrix above).
Independent validation by docking
7 out of the 8 dock well to the active site of
MMP- 2
The Big Pharma technology test
Enrichment Curves
Our ISE
Our superiority claim
 Highly innovative Prize winning optimization
algorithm
 The best enrichment algorithm currently available
 MBI: “actives” from “non-actives”
 DLI: drugs from “non-drugs”
 Identification of highly diverse drug candidates
 Reduction of time for lead identification and
optimization
We vs. chemical companies selling focused libraries
Company
name
Combinat.
algorithm
Novel
detect
False
positiv
False
negat.
Enrichment
Model
speed
Virtual
screening
speed 106
3D
structure
required ?
Biofocus
-
Yes
-
-
10-100
-
-
Yes
Pharmacopeia
No
Yes
High
High
5-50
-
16,000
hours/CPU
Yes
Enamine
No
No S
Yes D
Low S
High D
High
High
10-1000 S
5-50
D
-
300-16,000
hours/CPU
TimeTec
No
No
Low
High
10-1000
-
300
hours/CPU
No
IBSinterbioscreen
No
No
Low
High
10-1000
-
300
hours/CPU
No
Comgenex
-
No
Low
High
10-1000
-
300
hours/CPU
No
OSI
pharmaceutical
-
Yes
High
High
5-50
-
16,000
hours/CPU
Yes
Our algorithm
Yes
Yes
Low
Low
200 – 5,000
1-2
days
2 hours
No
No
Yes
We vs. Docking
Docking approaches
Results reported by GSK
Average enrichment factor
found on top 10%
Ideal
10.0
Dock4
2.3
DockIt
1.9
FlexX
3.7
Flo+
3.0
Fred
2.4
Glide
3.2
Glod
2.2
LigandFit
2.6
MOEDOCK
1.2
MVP
6.3
Ours – validation test of Big Pharma
9.5