CzeekD: Fragment-based de novo Drug Design System for Drug Discovery

Edmund Taylo Kyoto Constella Technologies Co., Ltd.

Kyoto, Japan

Corporate Information

Registered  Name:  Kyoto  Constella  Technologies  Co.,  Ltd  Establishment:  March  31,  2008  LocaBon:  Kyoto,  Japan  Business:  Contract  computaBonal  services,  

 chemoinformaBcs  related  soIware    development  

Rapid  screening  for  candidate  compound

Compound  design  and  opBmizaBon

Drug  related  adverse  reacBon  informaBon  retrieval  system

Basic  Research

Non-­‐clinical  experiments

Clinical  experiments  （clinical  trial）

Approval  and  

markeBng  applicaBon

Grant  of  approval

Post-­‐markeBng  surveillance

From screening to de novo design

Efficient creation of new active compound structures through the combination of ligand-target interaction machine-learning and optimization algorithms

SO

OO

O

OCl

zzzz

Limited  number  of  compounds  (approx.  6  million  compounds)

SelecBon  of  acBve  compounds  from  the  compound  library  

In  silico  screening

•  Commercially  available  compound  libraries  

•  In-­‐house  libraries�

InteracBon  machine-­‐learning

OpBmizaBon  algorithm  

SO

OO

O

OCl

Over  1060  compounds  

Highly  diverse  selecBon  of  acBve  novel  compounds  

Random  genera8on  of  compounds�

Efficiently  search  large  

chemical  space

SelecBon  of  high  quality  lead  compounds

Enormous  amount  of  computaBonal  Bme

Over  1060  compounds  

SO

OO

O

OCl

Highly  diverse  selecBon  of  acBve  novel  compounds  

Random  genera8on  of  compounds�

De novo compound design (CzeekD)

Fragment Binding • RECAP • BRICS

Structure substitution,

addition, and deletion • Bioisosteres • MMP

Without protein structure • CGBVS • QSAR

With protein structure

• docking • binding free energy

Optimization method • Full search • Tabu search • Evolution algorithm （GA, etc.）

• Swarm intelligence（Particle Swarm Optimization; PSO）

Crea%on  of  compounds

Evalua%on  Search  of  chemical  space

1  =  amide   2  =  ester   3  =  amine   4  =  urea  

5  =  ether   6  =  olefin  7  =  quaternary                nitrogen  

9  =  lactam  N  carbon  

Compounds  are  fragmented  based  on  RECAP  rules  

Fragmentation method

8  =  aroma%c  N  carbon  

10  =  aroma%c  carbon  –                      aroma%c  carbon   11  =  sulphonamide  

Example:

4  types  of  units  can  be  joined  together  to  create  a  target  synthesis  frame

Fragment  units

○ ○

x Units  with  dangling  bonds  are  not  allowed

Virtual design of compounds (CzeekD)

Appropriate  fragments  are  aPached  based  on  the  RECAP  rules

× Fragment  combina%ons  not  conforming  to  the  RECAP  rules  are  discarded.

CGBVS (Chemical Genomics-based VS)

0.10 －0.2 －0.4

NN

N

SOO

NN

N

O

N

NN

N

SO O

high  score  comp.

0.98 0.97

NN

N

O O

ON

N

N

O

O

SN O

O

Ligand-­‐target  predicBon  

high  score  comp.

Compound optimization

NN

N

O

O

O

ON

SNN

N

OS

NO

O

NN

N

OS

N

N

O

ON

N

N

OO

ONN

N

O

O

O

O

0.54 0.65 0.95   1.12 1.25

structural  features  of  compounds  with  high  scores  are  conserved  while  creaBng  successive  generaBons  of  compounds

NN

N

O

O

SN O

ON

N

N

OS

N O

ON

N

N

OS

N O

ON

N

N

OS

NO

O

N O

NN

N

OS

N O

OF

1.51 1.52 1.52   1.53 1.53

repeat  of  generaBon  cycle  (5000x)  

Enriched  collecBon  of  compounds  possessing  chemical  structures  that  suggest  pharmacological  acBvity      

Evolution of molecules within the machine

Over  1015  possible  combinaBons  

105  combinaBons

105  combinaBons 105  combinaBons random  generaBon  of  128  compounds  

Candidate compounds： β2AR: 29,815 V1bR: 38,088

A case study in de novo design(from library design to chemical synthesis)

CGBVS

PSO※

OpBmizaBon

Filter  by  constraints

Synthesis  frame Core  structure

Commercially  available  building  blocks

？ ？

#  of  virtual  compounds：  β2AR  ：  9.6  ×  108  V1bR  ：  2.3  ×  109

Target Evaluated Results

IC50≦30μM IC50≦100μM

β2AR 104 40 65

V1bR 105 24 77

ImplementaBon  of  compound  design,  chemical  synthesis  and  validaBon  of  lead  discovery,  shown  here  for  GPCRs

Combinatorial  chemistry

Compound  hits  with  novel  structures

Structure-activity relationship (SAR) data

β2AR  antagonist  SAR  table V1bR  antagonist  SAR  table

Substance  Patent  ApplicaBon  (V1b  antagonist)  ApplicaBon  No.:  2012-­‐177572  (August  9,  2012)    Applicants： ①  Kyoto  University  

             ②  Kyoto  Constella  Technologies  Co.,  Ltd.                ③  KNC  Laboratories  Co.,  Ltd.  Ø  Useful data for subsequent chemical synthesis

Ø  Results of validation yielded hit compounds now under patent application

AcBvity  (IC50) Side  chain  structure  with                                              effects  to  acBvity posi%ve nega%ve

Evaluation of specificity towards V1bR

3 compounds are evaluated for specificity against the GPCR family of proteins

Assay  performed

V1bR

GPCR名

活性評価実験（＋/－：50%阻害活性＠30μM） A-Ac-11 C-Ac-11

A9-Ac-11

① ADRB2 － － － ② CHRM1 ＋ － － ③ PTGER1 － － － ④ NPY2R － － － ⑤ V1AR ＋ ＋ － ⑥ V1BR ＋ ＋ ＋ ⑦ V2R － － － ⑧ EDNRB － － － ⑨ CCR1 － － － ⑩ SSTR1 － － － ⑪ LHCGR － － － ⑫ P2Y1 － － － ⑬ VIPR1 － － － ⑭ GRM1 ＋ ＋ ＋

N N

O

N

ONO2

N N

O

N

OO

O2N

+/- 50% Inhibitory Activity at 30 uM

Settings Window

De novo compound design system: CzeekD

SelecBon  of  target  proteins

Sefng  the  synthesis  frame

SelecBon  of  target  family

De novo design system: CzeekD

Results Analysis Window

Structure  display

Substructure  search

Summary

l  Features  v Fragment-­‐based,  syntheBcally  accessible  (RECAP)  v Unique  and  rapid  scoring  funcBon  (CGBVS)  v Efficient  search  of  chemical  space  (PSO)  v Screening  against  mulBple  target  proteins  v Capable  of  scaffold  hopping  v User-­‐friendly  interface  v Models  for  GPCRs,  Kinases,  Ion  channels  ,  Transporters,  Nuclear  receptors  

and  Proteases    

l  Proven  performance  v GPCRs  v Kinases  v  Ion-­‐channels  

l  Service  v  ASP  　→　SoIware  demos  (2  weeks)  v  Contract  service  

Arigatou gozaimasu! Thank you!