Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5580_Библиотеки_им_академика_М_И_Перельмана
.pdf
216 K. Hori et al.
https://t.me/med1917
TS_Motif:2.8
TS271:2.4
Fig. 12.6 Results of TS conformation analysis for a Pd complex catalyzed reaction
TS7369:0.0 TS1:1.7
TS7370:2.4
TS3872:2.8
analysis have to be finished within a realistic amount of time. In the present case, all
the calculations have completed within 1.5 days.
The result for the digital screening is as follows.
i. For the reaction using styrene as the olefin, it took 12.6 h to optimize the corre-
sponding TS structure (TS_Motif ) from the initial structure, which constructed
from the cyclohexene TS structure.
ii. The conformational analysis was completed within 0.9 h. The calculations
yielded numerous conformations, many of which did not retain the geometry
required as the TS motif. Ten conformations with lower energies were extracted
and subjected to TS optimization. TS271 express the TS optimized from the
271th conformations. As the current Conflex program does not have parameters
for transition metals, it is necessary to replace some elements with other ones.
iii. One of the TS optimizations of the Conflex conformation was completed in 30–
34 h. Ten TSs with different conformations are usually optimized to obtain the
most stable TS one. If only one computer is used, the calculation takes ten times
as long as that for one optimization. However, the use of many computers at one
time allows us to optimize all the TSs with the different geometry in parallel so
that conformation searches can be completed in a similar computation time as
that for one TS optimization.
The most stable TS conformation was calculated to be TS7369, which is more
stable by 2.8 kcal/mol than TS_Motif. As expected, the order of the stability of the
Conflex conformations was different from that of the DFT ones. We may not obtain
the most stable conformer from ten stable conformers of the molecular mechanics

12 Data-Driven Chemistry for Developing Organic Synthesis Routes … 217
https://t.me/med1917
calculations. However, we obtained the conformations more stable by 0–3 kcal/mol
than that from the TS motif method.
After these searches, it is necessary to find the most stable TS conformer, which
is used for IRC calculations. Then, the reactant and the product are optimized. These
calculations ascertain that the target compound is connected with the reactant via the
most stable TS.
12.4 TSDB/QMRDB for Data-Driven Synthesis
Development
12.4.1 TSDB/QMRDB and Data in HTML Browser
The digital screening takes full advantage of the property that the TS motif method can
optimize new TS structures in a short time by utilizing similarities in the TS geometry.
We began to gather TS motifs of elementary reactions around 2010. The PostgreSQL
relational database was used to store them, and a PHP program displays retrieved
data in Internet browsers. Currently, QMRDB stores information for approximately
22,000 structures and more than 4000 TS motifs of elementary reactions. TSDB
stores more than 800 data of name reactions, which are constructed with QMRDB
data.
Figure 12.7 shows data for formation of acyl iodide-3-cyclohexylpropanoyl iodide
22]. The reaction proceeds in three steps; the first is a cyclohexyl group coordination
[
to Pd accompanying a hydrogen transfer; the second is carbonylation of the cyclohexyl group; and the third is a reductive elimination to form the product. After these
reactions, the acyl iodide reacts with H
equation as well as the energy diagram are displayed in the browser. Click a red
button in the red circle for the first step, and you will get a window with the TS
3D structure as shown in the bottom right of Fig.
constructing the TS motif are highlighted in gold.
O to form a carboxylic acid. The reaction
2
12.7. Pd, H, and two C atoms
12.4.2 Retrieval of TS Information from TSDB/QMRDB
TOSP provides the reaction name and reactants as a retrieval result. Therefore,
SMILES of reactants/products was adopted as queries for TSDB/QMRDB searches.
Internet browser displays results in the order of calculated Tanimoto coefficients.
They are usually not very large so that it is difficult to find related reactions among
the many data displayed.
To solve this problem of TSDB/TSDB, we implemented a reaction search function, which uses Reaction SMILES implemented in the RDKit [
displays a reaction search result for Beckmann rearrangement reaction (Reaction
23]. Figure 12.8

218 K. Hori et al.
https://t.me/med1917
Fig. 12.7 Display of TSDB data and 3D structure showing TS motif atoms
Smiles: CC(C)=NO >>CNC(C)=O). In this search, the Tanimoto coefficient is calculated between the reactant of the t arget reaction and those of the retrieved reactions.
Those with the coefficients larger than the threshold, in this case 0.5, are displayed
in order of their magnitude. Information in the browser is (i) the QMRDB number
of the data, (ii) the structure of the reactant and the calculated Tanimoto coefficient,
and (iii) the activation free and the reaction free energies in kcal/mol. The implementation of the search engine facilitated reaction searches for related elementary
reactions. Similar retrievals can be performed for TSDB.
12.5 Example of a Data-Driven Synthesis Route
Development
Several studies for TOSP routes were conducted on how to use digital screening for
synthetic route development [
routes AI-SRDSs (AiZyntfinder, ASKCOS) and TOSP create. Unlike TOSP, The AISRDSs propose consecutive synthesis routes of the targets from known compounds
as starting materials. We attempted to create synthesis routes for Branebrutinib 1,an
inhibitor of Bruton’s Tyrosine Kinase shown in Scheme
While AiZyntFinder created no synthetic routes to commercial compounds,
ASKCOS proposed a route starting from 2-amino-4,5-difluorobenzoic acid 2 with
24–26]. Here, we compared what different synthesis
12.2.

12 Data-Driven Chemistry for Developing Organic Synthesis Routes … 219
https://t.me/med1917
Fig. 12.8 Window showing a retrieved TS information from TSDB
Scheme 12.1 Example of relatively large transition metal complex
four-step reaction in Scheme
12.2. TOSP was able to create no synthesis routes
for the fourth step. This may be due to the fact that the present TOSP does not have
transforms enough to create synthesis routes (e.g., S
Ar reaction, Buchwald-Hartwig
N
amination).

220 K. Hori et al.
,
https://t.me/med1917
OHO
NH
2
O
HN
F
F
NH
2
F
N
NHBoc
F
O
NH
N
1
H
N
O
N
H
7
H2N
HOBt,H2O,
WSCD,
Et3N,
;NH4Cl
88
OHO
F
,
2
N
O
a
N
q
C
l
a
H
a
N
2
;
q
a
C
H
4
9
F
HN
H
3
S
O
N
N
H
2
q
a
l
%
32
NH
O
2
8
F
F
OHO
F
H
N
a2CO3
DMSO
Not detected
-
F
NHBoc
,
l
H
C
c
O
A
%
2
5
HN
H
4
O
HN
6
e
,
n
o
t
a
n
u
2
b
Scheme 12.2 Synthesis route ASKCOS created for branebrutinib 1
This is a route for a preliminary experiment and the reactant, N-(piperidin-3-yl)
but-2-ynamid 7 for the final step is poor available so that the compound with a Boc
group 8 was used for an example of the data-driven synthetic route development.
It was confirmed that all the synthetic reactions were calculated to have TSs
connecting r eactants and products with appropriate magnitudes of activation free
energies. These results suggested that it is plausible to synthesize all the intermediates
and the final product. Therefore, we conducted to settle reaction conditions of all
steps. Table
12.1 summarizes the reaction conditions for the synthesis experiments,
in which reagents and solvents considered in the digital screenings were modified
as necessary. Although their yields are unsatisfactory, all the intermediates were
synthesized in agreement with the results of the digital screenings up to the third
step. However, the fourth step produced no target compounds.
The detailed results are as follows:
The first step. The diazotization of 2 followed by treatment with sodium sulfite
produces arylhydrazine derivative 3. The activation free energies for the two reactions were calculated to be 15.2 and 14.8 kcal/mol. These results indicate that these
reactions easily proceed. The synthesis experiments yielded intermediate 3 in 49%
yield.
The second step. The next reaction is an application of Fischer indole synthesis of
3 and 2-butanone to yield intermediate 4. The activation free energy was calculated
to be 37.6 kcal/mol. This magnitude is considered a little high for the reaction to
proceed. An experiment was carried out at 110 °C using acetic acid as solvent and
HCl as acid to give the target compound although in poor yield (25%).
The third step. ASKCOS proposed a reaction in which intermediate 4 is activated with thionyl chloride, followed by the reaction with ammonia to form amide
derivative 5. The activation free energies were calculated to be 20.1 and 10.3 kcal/
mol, respectively. The reaction is expected to easily proceed. Although the target
compound could not be synthesized under the conditions from literatures, the
modified conditions led to produce 5 of 88% yield.

12 Data-Driven Chemistry for Developing Organic Synthesis Routes … 221
https://t.me/med1917
Table 12.1 Reaction conditions used for synthesis of the target compound
Step Reaction name ΔGǂaReaction conditions and reagents Yields
1 1. Diazotization 15.2 conc.HCl (2 wt), NaNO2 (1.1 eq), H2O(13
2. Aryl hydrazine
synthesis
2 Fischer Indole
Synthesis
3 1. Acylation
4 SNAr reaction 28.6 (S)-3-(tert-butoxycarbonylamino)piperidine
a
kcal/mol
b
yield through two step reactions
c
No products were obtained
d
Unknown byproducts
(thionyl chloride)
2. Amide formation 10.3
14.8 Na2SO3 (4 eq), H2O (5 vol), 0~20°C then
37.6 AcOH (17 wt), MEK (2 wt), conc.HCl (0.6
20.1 SO2Cl2 (11 eq), toluene (9 vol), DMF (0.08
N.A Modified: HOBt・H2O (1.2 eq), WSCD (1.5
vol), 0°C
conc.HCl (10 wt) in H
vol), 110°C
vol), 20 to 55°C then 28%NH
eq), Et
N (1.1 eq), DMF (15 vol), 20°C then
3
NH4Cl (2 eq), Et
20°C
(3.5 eq), Na
110~130°C
2CO3
O (4 vol), 60~70°C
2
N (1.1 eq), DMF (5 vol),
3
(excess), DMSO,
aq
3
N.A
49%
b
25%
N.P
c
88%
–
d
The fourth step. The SNAr reaction substitutes the fluorine atom at the 4-position
with a piperidin-3-amine derivative. Since a reaction is also expected to proceed at
5-position, their TSs were optimized at the B3LYP/6-3G(d) level of theory. The
activation free energies were calculated to be 35.3 and 45.7 kcal/mol, respectively.
The results indicate that the reaction proceeds preferentially at the 4-position. The
activation free energy at the former position was calculated to be 28.6 kcal/mol at
the B3LYP/6-311++G level of theory.
Since the digital screening expected relatively high reaction temperatures, experiments were conducted around 110–130 °C in DMSO. The reaction produced not 6
but several unidentified compounds. In this example, side reactions are expected to
proceed at other sites, for example, the carbonyl carbon of the amide group, although
they were not evaluated. In order to reduce the time to ascertain the unidentified ones,
it is necessary to implement a function for SRDS what side reactions proceed for
the substrates under the reaction conditions. This function is being progressed in the
NEDO project.

222 K. Hori et al.
https://t.me/med1917
12.6 Concluding Remarks
The development of drugs and functional chemicals has traditionally relied on the
ability of synthetic chemists to apply the concept of retrosynthesis to target molecules
and to create synthetic routes that span multiple steps. The patience of chemists may
be the most important until the desired compound is obtained. These are talents that
only experienced and competent synthetic chemists can possess. Unfortunately, it is
impossible to place such researchers at every R&D site.
Since the late twentieth century, computer-aided synthesis route design has
become popular. The recent SRDSs can create practical synthesis routes. However,
for target compounds whose synthesis requires multi-step synthesis, SRDSs are faced
with the problem of synthesis route divergence because they propose many possible
routes for each intermediate on the route. Although uses of synthesis robots are
expected to significantly change experimental situations, the number of experimental
validations is still limited.
The data-driven synthetic route development does not require synthetic chemists
who are very good at creating synthesis routes while the route design capability
of SRDSs is critical. Moreover, the digital screenings do not predict with 100%
probability whether or not verified reactions can synthesize the target. However, the
number of experiments can be significantly reduced since the potential synthesis
routes are ascertained to proceed. Therefore, it is clear that the data-driven synthetic
route development will shorten the period required for developing synthesis routes for
new compounds or for replacing with new ones for known compounds. An innovation
will come in the world of synthetic chemistry, a field that has been lasting for more
than 150 years.
Acknowledgements We thank the financial support (JPNP19004) from the New Energy and
Industrial Technology Development Organization (NEDO).
References
1. Gisbert S (ed) (2013) De novo Molecular Design, Wiley-VCH, Weinheim. https://doi.org/10.
1002/9783527677016
2. Isayev O, Tropsha A, Curtarolo S (eds) (2019) Materials Informatics: Methods, Tools, and
Applications. Wiley-VCH, Weinheim.
3. Pensak DA, Corey EJ (1977) LHASA Logic and Heuristics Applied to Synthetic Analysis.
In: Wipke WT, Howe WJ (eds) Computer-Assisted Organic Synthesis, American Chemical
Society, Washington DC, pp 1–32.
4. Gasteiger J, Hutchings MG, Christoph B, Gann L, Hiller C, Löw P, Marsili M, Saller H, Yuki K
(1987) A New Treatment of Chemical Reactivity: Development of EROS, an Expert System for
Reaction Prediction and Synthesis Design. In: Organic Synthesis, Reactions and Mechanisms.
Springer, Berlin, Heidelberg, pp 19–73.
5. Laird ER, Jorgensen WL (1990) Computer-assisted Mechanistic Evaluation of Organic Reactions. 17. Free-radical Chain Reactions. J Org Chem 55:9–27.
8a005
https://doi.org/10.1002/9783527802265
https://doi.org/10.1021/bk-1977-0061.fw001
https://doi.org/10.1007/3-540-16904-0_14
https://doi.org/10.1021/jo0028

12 Data-Driven Chemistry for Developing Organic Synthesis Routes … 223
https://t.me/med1917
6. Funatsu K, Sasaki S (1994) AIPHOS Computerized Organic Synthesis Route Search. Computer
Chemistry Series 2. Kyoritsu Shuppan, Tokyo.
7. Satoh K, Funatsu K (1999) A Novel Approach to Retrosynthetic Analysis Using Knowledge
Bases Derived from Reaction Databases. J Chem Inf Comp Sci 39:316–325.
10.1021/ci980147y
8. Szymkuć S, Gajewska EP, Klucznik T, Molga K, Dittwald P, Startek M, Bajczyk M, Grzybowski
BA (2016) Computer-Assisted Synthetic Planning: The End of the Beginning. Angew Chem
Int Ed 55:5904-5937.
9. Genheden S, Thakkar A, Chadimová V, Reymond J-L, Engkvist O, Bjerrum E (2020) AiZynthFinder: A Fast, Robust and Flexible Open-Source Software for Retrosynthetic Planning. J
Cheminfo 12:70.
10. Coley CW, Green WH, Jensen KF (2018) Machine Learning in Computer-Aided Synthesis
Planning. Acc Chem Res 51:1281-1289.
11. Coley CW, Thomas DA 3rd, Lummiss JAM, Jaworski JN, Breen CP, Schultz V, Hart T, Fishman
JS, Rogers L, Gao H, Hicklin RW, Plehiers PP, Byington J, Piotti JS, Green WH, Hart AJ,
Jamison TF, Jensen KF (2019) A Robotic Platform for Flow Synthesis of Organic Compounds
Informed by AI Planning. Science 365:eaax1566.
12. Satoh K, Yukimoto Y, Funatsu K (1997) Development of the Proposal Function for Retrosynthesis Using Transform. J Chem Soc Jpn Chem Ind Chem 1997:135-441.
1246/nikkashi.1997.435
13. Yamamoto H, Yamaguchi T, Yoshimura K, Sumimoto M, Hori K (2012) Towards in silico
Synthetic Route Development. Computer Aided Organic Synthesis Developments of Target
Compounds. J Synth 70:722–730.
14. Hori, K (2001) A Data Base for Transition States. Ranking of Synthesis Routes by using a
System Combined Computational with Information Chemistry. J Comp Aided Chem 2:37-44.
https://doi.org/10.2751/jcac.2.37
15. “Development of Continuous Production and Process Technologies of Fine Chemicals”,
NEDO, Project No. JPNP19004
16. Hori K, Yamaguchi T, Uezu K, Sumimoto M (2011) A Free-Energy Perturbation Method Based
on Monte Carlo Simulations using Quantum Mechanical Calculations (QM/MC/FEP method):
Application to Highly Solvent-Dependent Reactions. J Comp Chem 32:778–786.
org/10.1002/jcc.21653
17. Kaweetirawatt T, Yamaguchi T, Higashiyama T, Sumimoto M, Hori K (2012) Theoretical Study
of Keto–Enol Tautomerism by Quantum Mechanical Calculations (The QM/MC/FEP Method).
J Phys Org Chem 25:1097–1104.
18. Fukui K (1981) The Path of Chemical Reactions—The IRC Approach. Acc Chem Res 14:363–
368.
19. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, Scalmani G,
20. Conflex9 Rev. C, CONFLEX Corporation.
21. Maeyama E, Yamaguchi T, Sumimoto M, Hori K (2021) A Method to Search the Most Stable
22. Okada M, Takeuchi K, Matsumoto K, Oku T, Yoshimura T, XHatanaka, M, Cho, J-C (2022)
https://doi.org/10.1021/ar00072a001
Barone V, Mennucci B, Petersson GA, Nakatsuji H, Caricato M, Li X, Hratchian HP, Izmaylov
AF, Bloino J, Zheng G, Sonnenberg JL, Hada M, Ehara M, Toyota K, Fukuda R, Hasegawa J,
Ishida M, Nakajima T, Honda Y, Kitao O, Nakai H, Vreven T, Montgomery Jr. JA, Peralta JE,
Ogliaro F, Bearpark M, Heyd JJ, Brothers E, Kudin KN, Staroverov VN, Keith T, Kobayashi
R, Normand J, Raghavachari K, Rendell A, Burant JC, Iyengar SS, Tomasi J, Cossi M, Rega
N, Millam JM, Klene M, Knox JE, Cross JB, Bakken V, Adamo C, Jaramillo J, Gomperts
R, Stratmann RE, Yazyev O, Austin AJ, Cammi R, Pomelli C, Ochterski JW, Martin RL,
Morokuma K, Zakrzewski VG, Voth GA, Salvador P, Dannenberg JJ, Dapprich S, Daniels
AD, Farkas O, Foresman JB, Ortiz JV, Cioslowski J, Fox DJ, Gaussian 09, Revision D.01 ,
Gaussian, Inc., Wallingford CT, 2013.
Reaction Pathway and Its Application to the Pinner Pyrimidine Synthesis Reaction. J Comp
Aided Chem 22:1-7.
Hydroxycarbonylation of Alkenes with Formic Acid Catalyzed by a Rhodium (III) Hydride
https://doi.org/10.1002/anie.201506101
https://doi.org/10.1186/s13321-020-00472-1
https://doi.org/10.1021/acs.accounts.8b00087
https://doi.org/10.1126/science.aax1566
https://doi.org/10.5059/yukigoseikyokaishi.70.722
https://doi.org/10.1002/jcc.21653
https://doi.org/10.2751/jcac.22.1
https://doi.org/
https://doi.org/10.
https://doi.

224 K. Hori et al.
https://t.me/med1917
Diiodide Complex Bearing a Bidentate Phosphine Ligand. Organometal 41: 1640–1468. https://
doi.org/10.1021/acs.organomet.2c00138
23. RDKit: Open-source Cheminformatics. https://www.rdkit.org. Accessed 3 Aug 2022
24. Hori K, Sumimoto M, Murafuji T (2015) Quantum Chemistry-Assisted Synthesis Route
Development. AIP Conf Proc 1702:20.
25. Hori K, Sadatomi H, Miyamoto A, Kuroda K, Sumimoto M, Yamamoto H (2010) Towards the
Development of Synthetic Routes Using Theoretical Calculations: An Application of In Silico
Screening to 2,6-Dimethylchroman-4-one. MOL 15:8289–8304.
ecules15118289
26. Hori K, Sadatomi H, Okano K, Sumimoto M, Miyamoto A, Hayashi S, Yamamoto H (2007) An
Attempt Method for Developing New Synthetic Routes by Fusing Computational Chemistry
and Chemoinformatics: Syntheses of Ethyl and Benzyl Methacrylates J Comp Aided Chem
8:12-18.
https://doi.org/10.2751/jcac.8.12
https://doi.org/10.1063/1.4938827
https://doi.org/10.3390/mol

Chapter 13
https://t.me/med1917
“Quantum-Chemoinformatics”
for Design and Discovery of New
Molecules and Reactions
Hiroko Satoh, Vincenz-Maria Steiner, and Jürg Hutter
13.1 Introduction
Chemical reaction design and prediction are important steps in the basic research
stage of drug development. They aim for efficient synthesis of designed molecules
with high yield, fewer side products, and reducing the cost and environmental impact
13.1c). Since the 1960s, many software systems for these purposes have been
(Fig.
developed (Fig.
2], SYNCHEM [3], SYNCHEM2 [4, 5
using rule- or logic-based approaches (Fig.
methods (Fig. 13.2b) using knowledgebase or machine learning (ML) models, which
are automatically built from databases (DBs), have been mainstream. The first datadriven systems were AIPHOS [
12], and SOPHIA [13], which initiated a first surge of data-driven approaches in
[
this area in the 1990s. The SECS system, which is one of the pioneering works
on computer-assisted synthetic design and prediction in the 1970s, started with a
logic approach [
from DB in 1990 [
generates reactions from first principles by formal bond- and electron-shift processes
] and employed automatic derivation of rules from reaction DB starting with its
16
[
version6[
13.2). In the early time, all of the systems, e.g., OCCS (LHASA) [1,
], SYNGEN [
13.2a). Since the late 1980s, data-driven
9], WODCA [10], SECS in 1990 [11], EROS6.0
14, 15] and introduced a knowledge base automatically derived
11
]. The first version of EROS was a rule-based system, which
12
]. Interest in these approaches has been revived in the 2010s, when
6, 7], and CAMEO [8], were
H. Satoh (B) · V. - M . S te i n e r · J. Hutter
Department of Chemistry, University of Zurich (UZH), Winterthurerstrasse 190, CH-8057 Zurich,
Switzerland
e-mail: hiroko.satoh@chem.uzh.ch
J. Hutter
e-mail:
hutter@chem.uzh.ch
H. Satoh
Joint Support-Center for Data Science Research, Research Organization of Information and
Systems (ROIS), Midorimachi 10-3, Tachikawa, Tokyo 190-0014, Japan
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024
H. Satoh et al. (eds.), Drug Development Supported by Informatics,
https://doi.org/10.1007/978-981-97-4828-0_13
225
Соседние файлы в папке Библиотека им академика М.И. Перельмана
