Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса
Prediction of the toxicity of chemical compounds is one of the most important steps in drug design. The use of phenolic compounds is a promising component in the pharmaceutical industry with many possible applications. The paper fo-cuses on the application of a probabilistic neural network for class...
Збережено в:
| Дата: | 2025 |
|---|---|
| Автори: | , |
| Формат: | Стаття |
| Мова: | Англійська |
| Опубліковано: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2025
|
| Теми: | |
| Онлайн доступ: | https://journal.iasa.kpi.ua/article/view/351434 |
| Теги: |
Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
|
| Назва журналу: | System research and information technologies |
| Завантажити файл: |  |
Репозитарії
System research and information technologies| _version_ | 1867334455636525056 |
|---|---|
| author | Pushkarova, Yaroslava Zaitseva, Galina |
| author_facet | Pushkarova, Yaroslava Zaitseva, Galina |
| author_institution_txt_mv | [
{
"author": "Yaroslava Pushkarova",
"institution": "Bogomolets National Medical University, Kyiv"
},
{
"author": "Galina Zaitseva",
"institution": "Bogomolets National Medical University, Kyiv"
}
] |
| author_sort | Pushkarova, Yaroslava |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2026-02-02T20:49:24Z |
| description | Prediction of the toxicity of chemical compounds is one of the most important steps in drug design. The use of phenolic compounds is a promising component in the pharmaceutical industry with many possible applications. The paper fo-cuses on the application of a probabilistic neural network for classifying 232 phenols based on their mechanisms of toxic action. The Kruskal–Wallis test was also used to assess the influence of molecular descriptors on the reliable classification of phenolic compounds based on the mechanisms of their toxic action. It is shown that for the correct training of a probabilistic neural network and effective prediction of the mechanisms of toxic action of phenols, it is suf-ficient to use only 5 molecular descriptors. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2025.4.07 |
| first_indexed | 2026-02-08T08:06:14Z |
| format | Article |
| fulltext |
Ya.M. Pushkarova, G.M. Zaitseva, 2025
120 ISSN 1681–6048 System Research & Information Technologies, 2025, № 4
UDC 543.33 + 004.032.26
DOI: 10.20535/SRIT.2308-8893.2025.4.07
PREDICTION OF MECHANISMS OF TOXIC ACTION OF
PHENOLS BY MEANS OF PROBABILISTIC NEURAL NETWORK
IN COMBINATION WITH KRUSKAL–WALLIS TEST
Ya.M. PUSHKAROVA, G.M. ZAITSEVA
Abstract. Prediction of the toxicity of chemical compounds is one of the most im-
portant steps in drug design. The use of phenolic compounds is a promising compo-
nent in the pharmaceutical industry with many possible applications. The paper fo-
cuses on the application of a probabilistic neural network for classifying 232 phenols
based on their mechanisms of toxic action. The Kruskal–Wallis test was also used to
assess the influence of molecular descriptors on the reliable classification of pheno-
lic compounds based on the mechanisms of their toxic action. It is shown that for the
correct training of a probabilistic neural network and effective prediction of the
mechanisms of toxic action of phenols, it is sufficient to use only 5 molecular de-
scriptors.
Keywords: artificial neural network, classification, drug design, phenol, toxicity.
INTRODUCTION
Assessment of the toxicity of chemical compounds is an important and necessary
stage on the way to the creation of new medicines. It is known that the experi-
mental study of only one type of toxicity is an expensive and long-term process.
Phenolic compounds have a number of useful properties that make them interest-
ing for pharmacy: antioxidant, anti-inflammatory, antimicrobial properties, anti-
cancer activities, etc. Additionally, phenolic compounds are often found in natural
sources, such as plants, which adds to their appeal for use in pharmacy [1–4].
Overall, the diverse range of beneficial properties exhibited by phenolic
compounds makes them valuable compounds in pharmacy and medicine, with
potential applications in the treatment and prevention of various diseases. But be-
fore using phenols in pharmacy, it is important to predict possible mechanisms of
their toxic action (polar narcotics, weak acid respiratory uncouplers, pro-
electrophiles and soft electrophiles). This helps to identify risks to people and to
take measures to reduce the possible negative consequences, that is, to develop
safe medicines [5; 6].
Chemometric methods use mathematical and statistical models to analyze
complex data sets and extract meaningful information, making them valuable
tools in pharmaceutical research and development. Chemometric methods, in par-
ticular artificial neural networks, are widely used for prediction and classification
tasks in pharmacy. Artificial neural networks are computational models inspired
by the structure and functioning of biological neural networks in the human brain.
These methods can help predict various properties of pharmaceutical compounds,
such as their stability, toxicity, solubility and bioavailability. They are also used
for identifying different types of drugs or distinguishing between counterfeit and
authentic products [7–10].
Prediction of mechanisms of toxic action of phenols by means of probabilistic neural network …
Системні дослідження та інформаційні технології, 2025, № 4 121
MATERIALS AND METHODS
Data Set
The studied dataset consists of a training, testing and validation sub-sets with a
total of 232 phenolic compounds: training sub-set – 197 phenols, testing sub-set –
20 phenols, validation sub-set – 15 phenols. All phenolic compounds were charac-
terized by seven physical-chemical descriptors: 1) distribution coefficient; 2) en-
ergy of the lowest unoccupied molecular orbital; 3) molecular weight; 4) nega-
tively charged molecular surface area in percent’s; 5) sum of absolute charges on
nitrogen and oxygen atoms in a molecule; 6) largest positive charge on a hydrogen
atom; 7) electrotopological state index for the hydroxyl group. Values of these
descriptors and toxicity values were taken from [6].
Distribution of the studied phenolic compounds into classes according to the
mechanisms of toxic action of phenolic compounds to Tetrahymena pyriformis is
presented in Table 1. The most numerous class is class 1 of polar narcotics
(71.6% of all studied phenolic compounds), other classes are almost the same in
number of samples.
T a b l e 1 . Distribution of the studied phenolic compounds into classes accord-
ing to the mechanisms of toxic action to Tetrahymena pyriformis
Number of Phenolic Compounds
Classes According to Mechanisms
of Toxic Action Training
sub-set
Testing
sub-set
Validation
sub-set
Total
Class 1. Polar narcotics 138 16 12 166
Class 2. Weak acid respiratory uncouplers 15 1 1 17
Class 3. Pro-electrophiles 22 2 0 24
Class 4. Soft electrophiles 22 1 2 25
Applied Methods
The software package Matlab R2023b (trial individual license 11937601) was
used in the present work for realization Kruskal–Wallis test and probabilistic neu-
ral network [11].
The Kruskal–Wallis test is a non-parametric statistical test used to determine
whether there are statistically significant differences between two or more groups
of a dependent variable [12].
A probabilistic neural network is a type of artificial neural network, which
consists of following layers: input layer, pattern layer, summation layer, and out-
put layer. A brief overview of how probabilistic neural network works [13–15]:
input layer receives the input pattern;
neurons of pattern layer store the training patterns;
summation layer computes the similarity between the input pattern and
the stored patterns using Gaussian function;
output layer produces the class probability estimates.
To classify a new input pattern, the probabilistic neural network computes
the class probabilities using the summation layer and outputs the class with the
highest probability.
Ya.M. Pushkarova, G.M. Zaitseva
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 122
RESULTS AND DISCUSSION
Definition of Informative Descriptors for Classification of Phenolic Com-
pounds into Classes According to the Mechanisms of Toxic Action
The calculation of the Kruskal–Wallis test for 232 phenols characterized by
7 molecular descriptors and toxicity is given in Table 2.
T a b l e 2 . Results of the Kruskal–Wallis test calculation for 7 descriptors and
toxicity
P
ar
am
et
er
T
ox
ic
it
y
D
is
tr
ib
ut
io
n
co
ef
fi
ci
en
t
E
ne
rg
y
of
t
he
lo
w
es
t
un
oc
cu
pi
ed
m
ol
ec
ul
ar
o
rb
it
a
M
ol
ec
ul
ar
w
ei
gh
t
N
eg
at
iv
el
y
ch
ar
ge
d
m
ol
ec
ul
ar
s
ur
fa
ce
ar
ea
in
p
er
ce
nt
’s
Su
m
o
f a
bs
ol
ut
e
ch
ar
ge
s
on
n
it
ro
ge
n
an
d
ox
yg
en
at
om
s
in
a
m
ol
ec
ul
e
L
ar
ge
st
p
os
it
iv
e
ch
ar
ge
o
n
a
hy
dr
og
en
at
om
E
le
ct
ro
to
p
ol
og
ic
al
st
at
e
in
de
x
fo
r
th
e
hy
-
d
ro
xy
l g
ro
u
p
χ2 17.80 54.32 104.90 35.78 70.24 31.71 4.34 18.56
Critical value of χ2 at the significance level of 5% with 3 degrees of freedom
is 7.82 [16].
It was established some dependences between studied descriptors and classi-
fication of phenolic compounds according to the mechanisms of their toxic action:
1) descriptor largest positive charge on a hydrogen atom is not influenced on
classification of phenolic compounds according to the mechanisms of toxic
action, because experimental value of χ2 is less than critical value (4.34 < 7.82);
2) descriptor energy of the lowest unoccupied molecular orbital has the
greatest influence on the phenols classification according to the mechanisms of
toxic action (maximum experimental value of χ2 is established for this descrip-
tor — 104.90);
3) the studied parameters can be conventionally divided into three groups
according to their influence on the classification of phenols:
weak influence: toxicity and electrotopological state index for the hy-
droxyl group;
moderately strong influence: molecular weight and sum of absolute
charges on nitrogen and oxygen atoms in a molecule;
strong influence: distribution coefficient, energy of the lowest unoccupied
molecular orbital and negatively charged molecular surface area in percent’s.
Application of Probabilistic Neural Network
In the context of the probabilistic neural network, the spread of the Gaussian
function is an important parameter for its construction. Choosing the right spread
parameter is crucial for the performance of the probabilistic neural network. If the
spread is too small, the network may over fit to the training data and perform
poorly on new data. If the spread is too large, the network may under fit and fail
to capture the underlying patterns in the data [8; 13].
In the present work it was investigated the applicability of probabilistic neu-
ral network at different values of the spread of the Gaussian function: 0.1; 0.2;
Prediction of mechanisms of toxic action of phenols by means of probabilistic neural network …
Системні дослідження та інформаційні технології, 2025, № 4 123
0.3; 0.4; 0.5; 0.6; 0.7; 0.8; 0.9; 1.0. It should be noted that the probabilistic neural
network is trained with zero error at spread values from 0.1 to 1.0 for different
sets of descriptors. Results of prediction of the mechanisms of toxic action of
phenols for testing and validation sub-sets are also the same for spread values
from 0.1 to 1.0 for different sets of descriptors.
The unreliability of the prediction was estimated as the part of incorrectly
classified phenols of the testing or validation sub-sets in percent’s [8]:
%100‧
N
n
P ,
where n is the number of incorrectly classified phenols in the testing or validation
sub-set; N is the total number of phenols in the testing or validation sub-set.
Results of prediction of the mechanisms of toxic action of phenolic com-
pounds by means of probabilistic neural network based on a set of all 7 molecular
descriptors and toxicity are shown in Table 3.
T a b l e 3 . Unreliability values of the prediction based on a set of all 7 molecu-
lar descriptors and toxicity
Sub-set P, %
Testing 10.0
Validation 6.7
Results of prediction of the mechanisms of toxic action of phenolic com-
pounds by means of probabilistic neural network based on a set of 5 molecular
descriptors (distribution coefficient, energy of the lowest unoccupied molecular
orbital, molecular weight, negatively charged molecular surface area in percent’s
and sum of absolute charges on nitrogen and oxygen atoms in a molecule) are
shown in Table 4.
T a b l e 4 . Unreliability values of the prediction based on a set of 5 molecular
descriptors
Sub-set P, %
Testing 20.0
Validation 6.7
One can see, that results of prediction of the mechanisms of toxic action of
phenolic compounds based on a set of all 7 molecular descriptors with toxicity
and based on a set of 5 molecular descriptors are differed by two incorrectly clas-
sified phenols. This confirms, the verity of calculation results of the Kruskal–
Wallis test: largest positive charge on a hydrogen atom, toxicity and electroto-
pological state index for the hydroxyl group are weakly influenced on assignment
of phenols to one or another class according to mechanisms of their toxic action.
Decreasing the number of descriptors into 3 (distribution coefficient, energy
of the lowest unoccupied molecular orbital and negatively charged molecular sur-
face area in percent’s) resulted in an increasing the part of incorrectly classified
phenols of the testing sub-set from 20% till 40% (Table 5). It means, that molecu-
lar weight and sum of absolute charges on nitrogen and oxygen atoms in a mole-
Ya.M. Pushkarova, G.M. Zaitseva
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 124
cule are moderately strong influenced for classification of phenols according to
mechanisms of their toxic action and can’t be ignore.
T a b l e 5 . Unreliability values of the prediction based on a set of 3 molecular
descriptors
Sub-set P, %
Testing 40.0
Validation 6.7
Detailed information about prediction of the mechanisms of toxic action of
phenolic compounds of testing and validation sub-sets are shown in Tables 6 and
7, correspondingly: 1 — polar narcotics; 2 — weak acid respiratory uncouplers;
3 — pro-electrophiles; 4 — soft electrophiles. Incorrect predictions are indicated
in bold text.
T a b l e 6 . Results of prediction of the mechanisms of toxic action of phenols of
the testing sub-set
N
Phenol
compound
Predicted mech-
anism of toxic
action using 7
descriptors and
toxicity (0.1 ≤
spread ≤ 1.0)
Predicted
mechanism of
toxic action us-
ing 5 descriptors
(0.1 ≤ spread
≤ 1.0)
Predicted
mechanism of
toxic action
using 3 de-
scriptors (0.1 ≤
spread ≤ 1.0)
True
mechanism
of toxic
action
[5, 6]
1 2-Fluorophenol 1 1 1 1
2 2-Allylphenol 1 1 1 1
3 3-Chlorophenol 1 1 1 1
4
4,6-
Dichlororesorcinol
1 1 3 1
5 4-Benzyloxyphenol 1 1 1 1
6 3-Iodophenol 1 1 1 1
7 2,3-Dichlorophenol 1 1 1 1
8 4-Phenylphenol 1 1 1 1
9 4-Hexyloxyphenol 1 1 3 1
10 4-Hexylresorcinol 1 1 1 1
11
2,4,5-
Trichlorophenol
1 1 1 1
12 2,4-Diaminophenol 3 3 1 3
13 Methylhydroquinone 3 1 1 3
14 3-Nitrophenol 4 4 1 4
15 4-Ethoxyphenol 1 3 3 1
16
4-Bromo-2,6-
dimethylphenol
1 1 1 1
17 4-Methoxyphenol 1 1 1 1
18
2,6-Diiodo-4-
nitrophenol
1 1 4 2
19
2-Methyl-3-
nitrophenol
4 4 4 1
20 4-Isopropylphenol 1 1 1 1
Prediction of mechanisms of toxic action of phenols by means of probabilistic neural network …
Системні дослідження та інформаційні технології, 2025, № 4 125
T a b l e 7 . Results of prediction of the mechanisms of toxic action of phenols of
the validation sub-set
N Phenol
compound
Predicted mecha-
nism of toxic
action using 7
descriptors and
toxicity (0.1 ≤
spread ≤ 1.0)
Predicted mech-
anism of toxic
action using 5
descriptors
(0.1 ≤ spread ≤
1.0)
Predicted
mechanism of
toxic action using
3 descriptors
(0.1 ≤ spread
≤ 1.0)
True
mecha-
nism of
toxic ac-
tion [5, 6]
1 4-Hydroxypropiophenone 1 1 1 1
2 3-Hydroxybenzaldehyde 1 1 1 1
3 4-(4-Hydroxyphenyl)-
2-butanone 1 1 1 1
4 4-Hydroxybenzaldehyde 1 1 1 1
5 4-Isopropylphenol 1 1 1 1
6 3-Fluoro-4-nitrophenol 4 4 4 4
7 Benzyl-4-hydroxybenzoate 1 1 1 1
8 5-Pentylresorcinol 1 1 1 1
9 2-Hydroxy-4-
methoxyacetophenone 1 1 1 1
10 3-Methyl-2-nitrophenol 1 1 1 1
11 2-Ethylhexyl-4/-
hydroxybenzoate 1 1 1 1
12 2,3-Dinitrophenol 2 2 1 2
13 2-Nitrophenol 4 4 4 4
14 3-Methoxyphenol 1 1 1 1
15 4-Chlororesorcinol 3 3 1 1
CONCLUSIONS
A set of five molecular descriptors (distribution coefficient, energy of the lowest
unoccupied molecular orbital, molecular weight, negatively charged molecular
surface area in percent’s and sum of absolute charges on nitrogen and oxygen at-
oms in a molecule) is sufficient for correct classification of phenolic compounds
by mechanisms their toxic effects.
The application of probabilistic neural network provides a reliable classifica-
tion of phenolic compounds by mechanisms of their toxic action, as well as pre-
diction of the mechanisms of their toxic action with high accuracy.
The proposed procedure for predicting the mechanisms of toxic action of
phenolic compounds can be useful at the stage of development of medicines.
REFERENCES
1. R.E. Mutha, A.U. Tatiya, S.J. Surana, “Flavonoids as natural phenolic compounds and
their role in therapeutics: An overview,” FJPS, vol. 7, article no. 25, pp. 1–13, 2021. doi:
https://doi.org/10.1186/s43094-020-00161-8
2. A. Durazzo et al., “Polyphenols: A concise overview on the chemistry, occurrence, and
human health,” Phytother. Res., vol. 33(9), pp. 2221–2243, 2019. doi:
https://doi.org/10.1002/ptr.6419
3. M.M. Rahman et al., “Role of phenolic compounds in human disease: current knowledge
and future prospects,” Molecules, vol. 27(1), pp. 233, 2021. doi:
https://doi.org/10.3390/molecules27010233
4. Y. Pushkarova, G. Zaitseva, M.Al Saker, “Prediction of Toxicity of Phenols Using Arti-
ficial Neural Networks,” IEEE — 12th International Conference on Advanced Computer
Information Technologies (ACIT), Spisska Kapitula, Slovakia, 26–28 September 2022,
pp. 493–496. doi: https://doi.org/10.1109/ACIT54803.2022.9913174
Ya.M. Pushkarova, G.M. Zaitseva
ISSN 1681–6048 System Research & Information Technologies, 2025, № 4 126
5. N. Aptula et al., “Multivariate discrimination between modes of toxic action of phenols,”
QSAR, vol. 21(1), pp. 12–22, 2002. doi: https://doi.org/10.1002/1521-
3838(200205)21:1<12::AID-QSAR12>3.0.CO;2-M
6. M.T. Cronin et al., “Comparative assessment of methods to develop QSARs for the pre-
diction of the toxicity of phenols to Tetrahymena pyriformis,” Chemosphere, vol. 49(10),
pp. 1201–1221, 2002. doi: https://doi.org/10.1016/s0045-6535(02)00508-8
7. Y. Pushkarova, V. Panchenko, Y. Kholin, “Application an Artificial Neural Network for
Prediction of Substances Solubility,” IEEE EUROCON 2021 − 19th International Con-
ference on Smart Technologies, Lviv, Ukraine, 6–8 July 2021, pp. 82–87. doi:
https://doi.org/10.1109/EUROCON52738.2021.9535593
8. Y. Pushkarova, G. Zaitseva, A. Kaliuzhenko, “Classification of Residual Solvents by
Risk Assessment Using Chemometric Methods,” IEEE — 13 th International Confer-
ence on Advanced Computer Information Technologies (ACIT), Wroclaw, Poland, 21–22
September 2023, pp. 562–565. doi: https://doi.org/10.1109/ACIT58437. 2023.10275405
9. E. Havránková, E.M. Pena-Mendez, J. Csöllei, J. Havel, “Prediction of biological activity
of compounds containing a 1, 3, 5-triazinyl sulfonamide scaffold by artificial neural net-
works using simple molecular descriptors,” Bioorg. Chem., vol. 107, pp. 104565, 2021.
doi: https://doi.org/10.1016/j.bioorg.2020.104565
10. S. Gummadi, P.K. Chandaka, “Chemometrics approach to drug analysis – An overview,” Am.
J. Pharm. Tech. Res., vol. 9, pp. 1–13, 2019. doi: https://doi.org/10.46624/
ajptr.2019.v9.i1.001
11. “MATLAB and Simulink for Engineered Systems,” mathworks.com. Available:
https://www.mathworks.com/
12. E. Ostertagova, O. Ostertag, J. Kováč, “Methodology and application of the Kruskal–
Wallis test,” Appl. Mech. Mater., vol. 611, pp. 115–120, 2014. doi:
https://doi.org/10.4028/www.scientific.net/AMM.611.115
13. D.F. Specht, “Probabilistic Neural Networks,” Neural Netw., vol. 3, pp. 109–118, 1990.
14. B. Mohebali, A. Tahmassebi, A. Meyer-Baese, A.H. Gandomi, “Probabilistic neural net-
works: a brief overview of theory, implementation, and application,” Handbook of probabilis-
tic models, pp. 347–367, 2020. doi: https://doi.org/10.1016/B978-0-12-816514-0.00014-X
15. Y. Zeinali, B.A. Story, “Competitive probabilistic neural network,” ICAE, vol. 24(2), pp.
105–118, 2017. doi: https://doi.org/10.3233/ICA-170540
16. J. Miller, J.C. Miller, Statistics and chemometrics for analytical chemistry. Pearson edu-
cation, 2018.
Received 15.06.2024
INFORMATION ON THE ARTICLE
Yaroslava M. Pushkarova, ORCID: 0000-0001-9856-7846, Bogomolets National
Medical University, Ukraine, e-mail: yaroslava.pushkarova@gmail.com
Galina M. Zaitseva, ORCID: 0000-0003-3138-6324, Bogomolets National Medical
University, Ukraine, e-mail: galinazaitseva777@gmail.com
ПРОГНОЗУВАННЯ МЕХАНІЗМІВ ТОКСИЧНОЇ ДІЇ ФЕНОЛІВ ЗА
ДОПОМОГОЮ ЙМОВІРНІСНОЇ НЕЙРОННОЇ МЕРЕЖІ В ПОЄДНАННІ
З ТЕСТОМ КРАСКЕЛА–УОЛЛІСА / Я.М. Пушкарьова, Г.М. Зайцева
Анотація. Прогнозування токсичності хімічних сполук є одним із
найважливіших етапів розроблення лікарських засобів. Використання
фенольних сполук є перспективним компонентом у фармацевтичній
промисловості з багатьма можливими застосуваннями. Працю присвячено
застосуванню ймовірнісної нейронної мережі для класифікації 232 фенолів за
механізмами їх токсичної дії. Для встановлення впливу молекулярних
дескрипторів на достовірну класифікацію фенольних сполук за механізмами їх
токсичної дії використали тест Краскела–Уолліса. Показано, що для
коректного навчання ймовірнісної нейронної мережі та ефективного
прогнозування механізмів токсичної дії фенолів достатньо використовувати
лише 5 молекулярних дескрипторів.
Ключові слова: штучна нейронна мережа, класифікація, дизайн ліків, фенол,
токсичність.
|
| id | journaliasakpiua-article-351434 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2026-02-08T08:06:14Z |
| publishDate | 2025 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/0e/47ae68582b2ec09780b553ec95bf8f0e.pdf |
| spelling | journaliasakpiua-article-3514342026-02-02T20:49:24Z Prediction of mechanisms of toxic action of phenols by means of probabilistic neural network in combination with Kruskal–Wallis test Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса Pushkarova, Yaroslava Zaitseva, Galina artificial neural network classification drug design phenol toxicity штучна нейронна мережа класифікація дизайн ліків фенол токсичність Prediction of the toxicity of chemical compounds is one of the most important steps in drug design. The use of phenolic compounds is a promising component in the pharmaceutical industry with many possible applications. The paper fo-cuses on the application of a probabilistic neural network for classifying 232 phenols based on their mechanisms of toxic action. The Kruskal–Wallis test was also used to assess the influence of molecular descriptors on the reliable classification of phenolic compounds based on the mechanisms of their toxic action. It is shown that for the correct training of a probabilistic neural network and effective prediction of the mechanisms of toxic action of phenols, it is suf-ficient to use only 5 molecular descriptors. Прогнозування токсичності хімічних сполук є одним із найважливіших етапів розроблення лікарських засобів. Використання фенольних сполук є перспективним компонентом у фармацевтичній промисловості з багатьма можливими застосуваннями. Працю присвячено застосуванню ймовірнісної нейронної мережі для класифікації 232 фенолів за механізмами їх токсичної дії. Для встановлення впливу молекулярних дескрипторів на достовірну класифікацію фенольних сполук за механізмами їх токсичної дії використали тест Краскела–Уолліса. Показано, що для коректного навчання ймовірнісної нейронної мережі та ефективного прогнозування механізмів токсичної дії фенолів достатньо використовувати лише 5 молекулярних дескрипторів. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2025-12-29 Article Article Peer-reviewed Article application/pdf https://journal.iasa.kpi.ua/article/view/351434 10.20535/SRIT.2308-8893.2025.4.07 System research and information technologies; No. 4 (2025); 120-126 Системные исследования и информационные технологии; № 4 (2025); 120-126 Системні дослідження та інформаційні технології; № 4 (2025); 120-126 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/351434/338454 |
| spellingShingle | штучна нейронна мережа класифікація дизайн ліків фенол токсичність Pushkarova, Yaroslava Zaitseva, Galina Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса |
| title | Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса |
| title_alt | Prediction of mechanisms of toxic action of phenols by means of probabilistic neural network in combination with Kruskal–Wallis test |
| title_full | Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса |
| title_fullStr | Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса |
| title_full_unstemmed | Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса |
| title_short | Прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом Краскела–Уолліса |
| title_sort | прогнозування механізмів токсичної дії фенолів за допомогою ймовірнісної нейронної мережі в поєднанні з тестом краскела–уолліса |
| topic | штучна нейронна мережа класифікація дизайн ліків фенол токсичність |
| topic_facet | artificial neural network classification drug design phenol toxicity штучна нейронна мережа класифікація дизайн ліків фенол токсичність |
| url | https://journal.iasa.kpi.ua/article/view/351434 |
| work_keys_str_mv | AT pushkarovayaroslava predictionofmechanismsoftoxicactionofphenolsbymeansofprobabilisticneuralnetworkincombinationwithkruskalwallistest AT zaitsevagalina predictionofmechanismsoftoxicactionofphenolsbymeansofprobabilisticneuralnetworkincombinationwithkruskalwallistest AT pushkarovayaroslava prognozuvannâmehanízmívtoksičnoídíífenolívzadopomogoûjmovírnísnoínejronnoímerežívpoêdnanníztestomkraskelauollísa AT zaitsevagalina prognozuvannâmehanízmívtoksičnoídíífenolívzadopomogoûjmovírnísnoínejronnoímerežívpoêdnanníztestomkraskelauollísa |