1 / 20

Novel Approach to Retrosynthesis Automatic Generation of Transform Libraries

V. Eigner-Pitto, J. Eiblmaier, H. Kraut, H. Saller and P. Loew.

InfoChem GmbH, Landsberger Strasse 408, Munich, 81241, Germany

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

2 / 20

Why Retrosynthesis?
Is there a shorter synthesis path to my target? How can I synthesize this compound?

Can we increase the yield of this process by choosing another synthesis strategy?

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

3 / 20

Retrosynthesis: Expected Features

Main Challenge: Automatic generation of chemically meaningful
precursors and their reasonable validation 1. Evaluation of Alternative Synthesis Pathways
Sorting criteria are necessary for broad representations with many expanding paths: Yield Type of transformation (e.g. ring-closures or construction reactions preferred, etc.)

2. Support of Multistep Synthesis

Necessary to guide the synthesis from the target to all existing precursors that can be obtained from commercially available substances

3.

Organized Visualization
Allows a broad view over the different strategies. A tree view, i.e., gives an organized representation of the synthesis paths to the target

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

4 / 20

Synthesis-Planning Tools
1. Use of known compounds and published reactions
i.e. MDL (CrossFire Commander), Symyx (Reaction Pathfinder), InfoChem (STS) ! Handle multistep synthesis ! Order in synthesis paths - Sorting criteria ! Shows synthesis options in your database (tree representation) " Structure and corresponding reaction(s) must be in database

2.

Intellectual retrosynthesis systems

! New molecules (not in the database) ! Alternative routes to known molecules ! Reaction transformation / Synthesis strategy to specify ! Transform based

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

5 / 20

Intellectual Retrosynthesis Approaches

2a. Transforms compiled manually from literature
! accurate, highly specific " tedious, expensive, restrictive

2b.

Transforms based on enumerated bond changes

! fully automated " unspecific, combinatorial explosion

Examples:
LHASA, SECS / CASP ... (Corey, Wipke) WODCA, EROS ... (Gasteiger) SYNGEN ... (Hendrickson) AIPHOS ... (Sasaki) .....
ICIC Nmes, October 22-25, 2006

InfoChem GmbH 2006

Dr. V. Eigner-Pitto

6 / 20

RetroSynthesis Concept Overview

1. Pre-processing
Template (2 levels):
1 14 2 4 7 13 8

11

12

R1
10 6 3

R2
9

Automatic transform extraction (ICMAP/CLASSIFY)

Transform:

R e m o v e b o n d 1 -2 M a k e n e w s in g le b o n d b e t w e e n a to m s 3 a n d 4
Decrease bond order of double bond 2=4 by 1. Make new single bond between atoms 2 and 6. Add group to atom 3: -OH

R e m o v e b o n d 1 -3 Remove atoms M ake ne w s in bond g le between b ond b e t w3 eand e n 6. atom s 3 a nd 5

Reaction database
Transform Library

2. RetroSynthesis
O H H O

e. g. SPRESI proprietary databases commercial databases

RetroSyn
Lookup stored examples

Target molecule

Reaction 1: Reaction 2:

Precursor(s) 1 Precursor(s) 2

Transform x
O

Transform x
H H O

Transform y

Transform y

...
Reaction n:

Precursor(s) n

Transform z

Target molecule

Transform z

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

7 / 20

Reaction Database Requirements

1. Correct Reaction Mapping
Description of the transformation occuring during the reaction Determination of reaction centers

2.

Reaction Classification
Automatic identification of the structural environment around the reacting centers Determination of unique identifiers (ClassCodes)

O O + N N

Reaction Type Amination:

.3. .3.

O
.2. .1.

O
.5.

+ O
.4.

N
.1.

.5. .2.

F F F

O O O + N

F F F

O N

N
.4.

Cl

Cl

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

8 / 20

Definition of Template (Retron)

SPRESI Reaction Reg-No.: 3179546
Kita Yasuyuki, Yoshida Yutaka, Mihara Sachiko, Fang Dai-Fei, Higuchi Kazuhiro, Furukawa Akihiro, Fujioka Hiromichi Efficient pinacol rearrangement mediated by trimethyl orthoformate Tetrahedron Lett., 38 (1997) 48, 8315-8318

CH3 HO OH O

CH3

CH3 O O

CH3 O

CH3

Template (level 2)

Template (level 3)
ICIC Nmes, October 22-25, 2006

Template (level 4)

InfoChem GmbH 2006

Dr. V. Eigner-Pitto

9 / 20

Definition of Transform

Template:
1 2 3 5

CH3
8

O
4

Indications:

Remove bond between atom 5 and 6 Decrease bond order of double bond 3=4 by 1 Make new single bond between atoms 3 and 6 Add group to atom 5: -OH

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

10 / 20

Pre-processing Step (Transform Library Building)

Reaction Database Transform Library

O O + N

O N

11

12

R1
10

R2
9 6

O O + N

O N

Transform Generator
O

Template (2 levels):
1 14 2

3 7 4

5 8 13

R em ove bo nd 1-3 M a k e n e w s in g le b o n d b e t w e e n a t o m s 3 a n d 5
Transform:

b on d 1-2 M a k e ne w s Decrease in g le b bond o nd order b eof tw double e e n bond a t o2=4 m sby 31.a n d 4
Make new single bond between atoms 2 and 6. Add group to atom 3: -OH

R e3m ov e Remove bond between atoms and 6.

F F F

O O O + N

F F F

O N

Cl

Cl

Generation of substructure index

Template Index

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

11 / 20

Templates Search
Rated Templates

Template Index Target Molecule

ICFSE
Rating C Rating B Rating A

Strategy Parameters

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

12 / 20

Strategy Parameters and Precursors Rating

Strategy parameters for precursor-rating consider the following criteria:
formation of a new carbon-carbon bond bond change within a functional group formation of a new ring/aromatic bond formation of a new bond between two symmetric fragments formation of a new bond near the center of the molecule number of generated precursors

These criteria are weighted depending on the chosen configuration:

sets of pre-defined strategy parameters are available sets may be defined from user

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

13 / 20

Precursors Generation from Target

Rated Templates Target Precursors

RetroSyn

Transform Transform

Transform
Rating c Rating B Rating A

Transform Library

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

14 / 20

Multistep Synthesis
Best First Search
Rating A

Precursors Step 1
O Br O O

Precursors Step 2
Rating AA

Mg

Br

Catalog
OH

Target
O

OH

O
O

O
+

O
Rating AB

NH2

OH

Limits:
steps precursors
Rating B

OH
O

N N
Rating BA

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

15 / 20

Prototype: Query Form

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

16 / 20

Reaction Databases Used in Prototype

SPRESI Synthesis Name Reactions EROS SOS CAC
2.5 mio reactions from SPRESI reactions with 50% minimum yield 140.000 reactions from SPRESI all examples for a large set of Name Reactions 68.000 reactions from J. Wileys Encylopedia of Reagents for Organic Synthesis 120.000 reactions from Thiemes Science of Synthesis 6.600 reactions from Springers Comprehensive Asymmetric Catalysis

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

17 / 20

Prototype: Suggested Precursors

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

18 / 20

Prototype: Synthesis Tree

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

19 / 20

Conclusions
Very large pool of data available (SPRESI 4 million reactions) Automatic transform generation from reaction databases Suggestions based on real (analogous) reactions Verification of suggested synthesis reactions, through link to reaction database / literature Availability-check of precursors in catalogs System will be integrated in a retrieval system providing common search tools

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006

Dr. V. Eigner-Pitto

20 / 20

Thank you!

InfoChem GmbH:

www.infochem.de www.spresi.de info@infochem.de

Dr. V. Eigner-Pitto

InfoChem GmbH 2006

ICIC Nmes, October 22-25, 2006