Computational Modeling and Biological Evaluation of Benzophenone Derivatives as Antileishmanial Agents

Abstract

Leishmaniasis is a neglected tropical disease with limited therapeutic options characterized by high toxicity, adverse side effects, and growing resistance to existing treatments. In this study, machine learning (ML) methods were employed to design and evaluate benzophenone and xanthone derivatives as potential antileishmanial agents. A dataset of 73 compounds was curated, and Quantitative Structure-Activity Relationship (QSAR) models were developed using artificial neural networks (ANN), Random Forest (RF), and J48 decision tree classifiers. The ANN model achieved the highest accuracy (86.2%) in predicting antileishmanial activity, validated through in vitro assays. Among 14 newly synthesized benzophenones, compounds 5 and 7 demonstrated significant biological activity with inhibitory concentration 50 (IC₅₀) values of 10.19 and 14.35 μM, respectively, and favorable selectivity indices compared to reference drugs pentamidine and amphotericin B. Structural analysis highlighted the importance of thiosemicarbazone and 4-methyl groups, alongside electronegative substituents at position 11, in enhancing activity. This study underscores the potential of computational tools to streamline the discovery of novel, effective, and selective antileishmanial agents.

Keywords:
benzophenone; xanthone; leishmaniasis; machine learning techniques

Introduction

Leishmaniasis, a neglected disease, disproportionately affects populations living in regions with tropical and subtropical climates.¹ According to the World Health Organization (WHO),² an estimated 50 000 to 90 000 new cases of visceral leishmaniasis occur worldwide annually. In 2023, approximately 83% of these cases were reported in seven countries: Brazil, Ethiopia, India, Kenya, Somalia, South Sudan, and Sudan.³,⁴ These data underscore the critical need to implement effective and sustainable public health strategies in these regions, aiming to control, prevent, and eventually eliminate disease transmission. The firstline treatment for leishmaniasis is based on pentavalent antimonial drugs, such as Pentostan® and Glucantime®. Amphotericin B is considered the second line therapeutic option.⁵ However, both treatments present significant challenges, including high toxicity and difficulties in administration, which often lead patients to discontinue treatment. Other drugs, such as pentamidine, miltefosine, and paromomycin, are also available, but all are associated with high toxicity and adverse side effects.

Furthermore, the high cost of treatments and the growing prevalence of resistant strains highlight the urgent need for innovative therapeutic alternatives.³ In this context, the search for more effective drugs with lower toxicity and fewer side effects is crucial, and computational methods, such as machine learning and artificial intelligence, have played a pivotal role in this endeavor.⁶,⁷,⁸

Benzophenones and xanthones are naturally occurring compounds widely distributed in the flowers, fruits, seeds, and leaves of numerous plant species.⁹,¹⁰,¹¹ For instance, benzophenones (1a-2a), isolated from the fruits of Garcinia brasiliensis, demonstrated significant activity against both amastigote and promastigote forms of Leishmania amazonensis. Additionally, the simplest synthetic derivatives of benzophenones (3a-4a) not only inhibited promastigote growth but also exhibited activity as papain inhibitors and were effective in inhibiting the cysteine proteases B rCPB2.8 and rCPB3.0, which are critical to the lifecycle of the parasite (Figure 1).⁷,⁸,⁹,¹⁰,¹¹,¹² Xanthones (5a-6a), another class of natural products, structurally related to benzophenones, have also demonstrated promising antileishmanial properties.¹³,¹⁴,¹⁵ These findings highlight the potential of benzophenones and xanthones as promising therapeutic agents against Leishmania spp., emphasizing their critical role in advancing the search for innovative and effective antiparasitic treatments.

Figure 1
Examples of benzophenones and xanthones with antileishmanial activity.

Machine learning (ML) techniques, based on artificial intelligence, enable computer systems to learn from data by identifying patterns to make decisions or predictions with minimal human intervention. These techniques have been transformative across various domains, including search engines, financial services, healthcare, drug discovery, and design applications.¹⁶ In the field of drug discovery, ML has revolutionized the process by significantly accelerating the identification and optimization of therapeutic compounds. These methodologies are highly effective in processing large datasets, identifying complex relationships between chemical structures and biological activities, and accurately predicting molecular properties. For example, algorithms such as Random Forest, Support Vector Machines, and Neural Networks have proven fundamental in predictive modeling for QSAR (Quantitative Structure-Activity Relationship) studies, providing critical insights into chemical interactions and enhancing the efficiency of drug development efforts.⁸,¹⁶

In this research, a QSAR prediction model based on machine learning was developed to forecast the antileishmanial activity of benzophenone and xanthone derivatives. This approach employed advanced computational methods to analyze molecular descriptors and patterns related to biological activity. The model was subsequently validated through in vitro testing, demonstrating its robustness and potential to optimize the discovery of novel antileishmanial agents.

Experimental

Chemical database

Our initial chemical database was elaborated from nine articles and one patent published up to the year 2020, documenting a total of 73 benzophenones and xanthones with reported antileishmanial activity. The selection criteria focused on the inclusion of robust and reliable quantitative data, with emphasis on half-maximal inhibitory concentrations (IC₅₀ values) obtained against the promastigote form of Leishmania spp. (Table S1) in Supplementary Information (SI) section.

Chemical data curation

We utilized ChemBioDraw Ultra 13.0¹⁷ (PerkinElmer, Inc., Cambridge, USA) and MarvinSketch 20.5 (ChemAxon Ltd., Budapest, Hungary)¹⁸ to create two-dimensional chemical structures and export them as .mol files. KNIME 5.1 (KNIME AG, Zurich, Switzerland).¹⁹ an open-source platform for data integration, processing, analysis, and exploration, was employed to curate and analyze the chemical and molecular data through specific workflows tailored to different substances. This platform facilitated the development of all workflows essential for executing the analysis steps of predictive models using chemoinformatic nodes.²⁰,²¹,²²,²³

A comprehensive molecular optimization process was conducted to identify the most suitable approaches or combinations of methods for the dataset, with multiple modeling strategies executed and compared within KNIME using the OpenBabel 4.1.0 node (Open Babel Development Team, Pittsburgh, PA, USA).²⁴ This process involved critical steps such as excluding structures with missing experimental values, adding explicit hydrogens, performing 3D cleanup, aromatization, interatomic distance analysis, removal of inorganic components, 3D structural optimization, and determining lower-energy conformers. These procedures were meticulously aligned with established data curation protocols to ensure the reliability and accuracy of the molecular models (Figure S1, SI section).²¹,²⁵ Finally, all optimized structures were consolidated into a single .sdf file, providing a comprehensive dataset for subsequent analyses.

Molecular descriptors

PaDEL-Software 2.2.1 (Singapore)²⁶ was employed to calculate a comprehensive range of 0D, 1D, 2D, and 3D molecular descriptors.²⁷ The 0D descriptors include atom and bond counts, as well as the sum or average of atomic properties. In contrast, the one-, two-, and three-dimensional features encompass molecular formulae, fragment counts, molecular connectivity, and geometric spatial quantum information. The resulting data were saved in .csv format (Table S2 at Zenodo)²⁸,²⁹ and integrated into a workflow designed for descriptor classification and predictive model development. For the final classification of molecular descriptors, the “BestFirst” search algorithm, combined with the “CfsSubsetEval” evaluator, was applied using the “AttributeSelectedClassifier 3.7” node of the Weka 3.8.0 software (University of Waikato, Hamilton, New Zealand).³⁰ integrated within the KNIME platform. This methodology was specifically designed to identify the most relevant descriptors by assessing their predictive power and inter-correlation, ensuring the selection of those descriptors with a significant relationship to IC₅₀ values. This step was crucial in refining the dataset to focus on descriptors that contributed most effectively to the predictive accuracy of the model.

Prediction models

Prediction models were developed using the selected molecular descriptors (X variables) and IC50 values (Y variables) to assess the antileishmanial activity of benzophenone and xanthone derivatives. The evaluation incorporated tree-based classifiers, including Decision Tree Classifiers (J48), Random Forest (RF), and Artificial Neural Networks (ANN).³¹,³²,³³ This comparison utilized a dataset of metabolites characterized by molecular descriptors, ensuring robust and accurate predictions. The dataset was partitioned using the “Partitioning” node in the KNIME platform, dividing the data into training and testing subsets. Specifically, 80% of the data was allocated for model training, while the remaining 20% served as an external test set. This process was carefully designed to validate the models’ predictive performance and ensure their reliability for practical applications. The workflows used for these predictions, alongside with the datasets and supplementary files are available in the Zenodo repository.²⁹ Researchers can download the data and trained predictive models, as well as replicate the workflows designed in the KNIME platform.³⁴

Model validation

The “CSV Reader” node was utilized to import the file containing biological activity information for each compound along with their respective descriptors, forming the basis for developing the prediction models. The “X-Partitioner” node was employed to partition the data and perform internal validation by selecting 10 random groups. Subsequently, the “X-Aggregator” and “Data to Report” nodes were used to combine the validation results and save the outcomes, respectively.³⁵ The effectiveness of our classification model was evaluated using a suite of standard performance metrics. Accuracy, defined as the proportion of correct predictions, reflects the overall efficacy of the model. Sensitivity (True Positive Rate, TPR) assesses the ability of the model to correctly identify positive instances. Precision, which measures the ratio of true positives among all positive predictions, evaluates the capability of the model to minimize false-positive classifications. Cohen’s Kappa provides a statistical measure of agreement between the predicted and actual classifications, accounting for chance agreement. The F-measure, a composite metric combining precision and recall, offers a balanced assessment of the performance of the model, specifically addressing the trade-off between precision and sensitivity. High values across these metrics, accuracy, sensitivity, precision, and F-measure, indicate a robust and reliable model, demonstrating strong predictive performance. Collectively, these metrics provide a comprehensive evaluation of the predictive capabilities of the model on the dataset.²⁶,³⁵,³⁶

Virtual screening

Virtual screening was performed using the most robust algorithms in the predictive models. This process aimed to assess the ability of the model to predict the antileishmanial activity of 14 newly synthesized benzophenones. These compounds were specifically synthesized for this study and were not included in the training or test datasets used to build the predictive models, ensuring an unbiased evaluation.³⁷ Further details on the molecular structures and properties of the 14 benzophenones can be found at Zenodo (Table S3).²⁹

Antileishmanial evaluation

The antileishmanial activity of previously obtained benzophenone derivatives (1-14) was evaluated.⁹,¹⁶ Promastigotes of Leishmania amazonensis (MHOM/ BR/71973/M2269) were cultured in Schneider’s Drosophila medium (Sigma-Aldrich, Burlington, Massachusetts, USA) supplemented with 10.0% (v/v) heat-inactivated fetal bovine serum and 1.0% penicillin (10,000 UI mL^-1)/streptomycin (10.0 mg mL^-1) (Sigma-Aldrich, Burlington, Massachusetts, USA) at 25 °C. For the in vitro assay, 1 × 10⁶ cells mL^-1 were plated 24-well plates, and the compounds were tested at concentrations of 0.1, 1.0, 5.0, 10.0, 20.0, and 40.0 μg mL^-1. After 72 h of incubation at 25 °C, 100 μL of resazurin (7-hydroxy-3//-phenoxazine-3-one-10-oxide, Sigma-Aldrich, Burlington, Massachusetts, USA) was added to each well. Plates were read 6 h later, at 570 and 600 nm using UV-Vis spectrophotometer (Asys Anthos, Biochrom LTD, Cambridge, United Kingdom). Negative controls consisted of L. amazonensis in blank culture media, while amphotericin B and pentamidine (both Sigma-Aldrich, Burlington, Massachusetts, USA) were used as positive controls.

Cytotoxicity assay

Cytotoxicity was evaluated as described by Folquitto et al.³⁸ Murine peritoneal macrophages were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Sigma-Aldrich, Burlington, Massachusetts, USA) at 37 °C with 5% CO₂ in 96-well plates, at a density of 8 × 10⁶ cells per well. Compounds were evaluated at concentrations of 500; 250; 125; 62.5; 31.25; 15.625; 7.8125; 3.90 μg mL^-1, followed by incubation for 72 h. Afterward, 10 μL of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma-Aldrich, Burlington, Massachusetts, USA) was added to each well, and the plates were incubated for an additional 4 h. Cells were treated with dimethyl sulfoxide (DMSO, Sigma-Aldrich, Burlington, Massachusetts, USA), and absorbance was measured at 570 nm using a UV-Vis spectrophotometer (Asys Anthos, Biochrom LTD, Cambridge, United Kingdom) to determine the 50% cytotoxicity concentrations (CC50) for the derivatives and control groups.

Results and Discussion

Chemical database

The dataset for building the prediction models was compiled from literature data.¹²,³⁸,³⁹,⁴⁰,⁴¹,⁴²,⁴³,⁴⁴,⁴⁵ Each compound (instance) was assigned with an identification number (ID), and its respective IC₅₀ value as Y-variables (Table S1, SI section). A total of 73 compounds were selected based on their chemical diversity, specifically focusing on benzophenone and xanthone scaffolds, and guided by the concept of the applicability domain, to ensure precise predictions.⁴⁶ The reported antileishmanial properties of these compounds served as the foundation for developing the prediction models in this study. To construct a classification prediction model within the QSAR framework, the median IC50 value was calculated and used as a threshold for categorization. This classification method was derived from statistical quartile values calculated from the IC₅₀ data in the activity column. Based on these quartile thresholds, the compounds were grouped into three activity categories: A (active) IC₅₀ < 30 μM, M (moderate) IC₅₀ > 30 μM and < 40 μM, and I (inactive) IC₅₀ > 40 μM.

Molecular descriptors, attribute selection, and performance criteria

For the 73 optimized molecules, a comprehensive set of 1.876 two-dimensional (2D) and three-dimensional (3D) descriptors was calculated using the PaDEL-Descriptor software. This process focused on capturing the topological, geometric, and structural characteristics of the compounds. Subsequently, the CfsSubsetEval algorithm, in combination with the in combination with the BestFirst classifier, was employed to select a total of 11 molecular descriptors. This selection methodology evaluates the value of a subset of attributes (molecular descriptors) by assessing both the individual predictive ability of each descriptor and the degree of redundancy among them, ensuring the selection of the most relevant and non-redundant features for the model.⁴⁶,⁴⁷ Finally, the selected subset of attributes was refined to include the following six descriptors: MATS3i, GATS5c, VE3_Dzv, maxsCH3, R_TpiPCTPC, and E2i (Table S4, SI section). These descriptors were chosen based on their relevance to the predictive and classification models, representing key physicochemical and structural properties such as charges, first ionization potential, the ratio of total conventional bond order (up to order 10), van der Waals interactions, and the maximum atom-type E-State.⁴⁷,⁴⁸ This selection ensures the incorporation of descriptors that are highly relevant to the predictive accuracy and interpretability of the models.

For a prediction model to be considered satisfactory, it should present accuracy and external validation values greater than 70%.⁴⁷,⁴⁹,⁵⁰ The performance of three machine learning models (ANN, J48 and RF) was evaluated for predicting the antileishmanial activity of benzophenone and xanthone derivatives using metrics such as accuracy, kappa coefficient, sensitivity, specificity, precision, recall, and F-measure. ANN achieved the highest overall accuracy (86.2%) and external validation accuracy (80.0%), demonstrating strong generalizability and reliability in identifying active compounds. RF exhibited exceptional robustness, achieving 100% training accuracy and balanced performance during external validation (80.0% accuracy, 0.625 kappa), with high adaptability to complex datasets. J48 performed well on training data (84.5% accuracy, 0.717 kappa), but showed moderate generalizability (73.3% external validation accuracy), suggesting potential for refinement. Overall, ANN and RF emerged as complementary approaches, with ANN excelling in accuracy and recall, while RF demonstrated superior robustness, making them reliable tools for predicting antileishmanial activity in diverse chemical datasets. (Figure S2 (SI section) and Table 1). These findings suggest the existence of common chemical characteristics among the benzophenone and xanthone derivatives isolated from the plants, which might be important for their antileishmanial activity.

Table 1
Performance obtained from the ANN (artificial neural network), Random Forest (RF) and J48 classifiers. It includes accuracy, Cohen’s kappa, sensitivity, specificity, precision, recall, and F-measure

Algorithm	Metric	Accuracy / %	Kappa	Sensitive	Specificity	Precision	Recall	F-measure
ANN	accuracy	86.2	0.745	0.90	0.821	0.844	0.90	0.87
	external validation	80.0	0.625	0.857	0.75	0.75	0.857	0.80
	internal validation	74.1	0.508	0.767	0.786	0.793	0.767	0.78
J48	accuracy	84.5	0.717	0.80	0.893	0.889	0.80	0.842
	external validation	73.3	0.524	0.571	0.875	0.80	0.571	0.667
	internal validation	72.4	0.503	0.70	0.786	0.778	0.70	0.737
RF	accuracy	100	1	1	1	1	1	1
	external validation	80.0	0.625	0.714	0.875	0.833	0.714	0.769
	internal validation	72.4	0.478	0.867	0.643	0.722	0.60	0.788

The scrambling technique, which involves randomizing the Y-axis (IC₅₀ values) in the training set, was performed to evaluate the robustness and potential overfitting of the models (Table S5, SI section). When trained with scrambled dataset, the performance of all models decreased significantly across all metrics, including accuracy, sensitivity, specificity, precision, recall, and F-measure (Table S6, SI section). For the ANN model, accuracy dropped to 74.1%, and metrics like precision (64.0%) and F-measure (71.1%) were markedly lower compared to the original dataset, indicating reduced predictive reliability. RF exhibited a similar trend, with accuracy falling to 32.8% and precision at 26.1%, reflecting diminished robustness. The J48 model performed the worst, with accuracy as low as 34.5% and no ability to distinguish active compounds (sensitivity and precision at 0%). These results collectively demonstrate that the models, particularly ANN and RF, rely heavily on the meaningful relationships between descriptors and IC50 values to achieve high predictive performance. The significant performance degradation with scrambled data confirms that the original models were well-fitted and not overfitted, as they were able to extract relevant patterns from the properly structured dataset.

Virtual screening

The virtual screening conducted in this study allowed the rapid evaluation of the antileishmanial potential of 14 newly synthesized benzophenones. Using computational models such as J48, ANN, and RF, the screening provided predictions that were compared with experimental in vitro results (Table 2), offering insights into the activity of the compounds.

Table 2
Comparison in vitro antileishmanial activity for synthesized benzophenone derivatives for external validation and their respective predictions outputs

Compound	IC50 / μM	CC50 / μM	Activity	ANN	RF	J48
CC1=CC=C(C=C1)C(=O)C1=CC=C(C=C1)[N+]([O-])=O (1)	> 100	-	I	I	I	I
OC(=O)C1=CC=C(C=C1)C(=N/NC1=NC(=CS1)C1=CC=CC=C1)\ C1=CC=CC=C1 (2)	> 100	-	I	I	I	I
COC1=CC=C(C=C1)C(=O)C1=CC=C(C)C=C1 (3)	> 100	-	I	I	I	I
CC1=CC=C(C=C1)C(=O)C1=CC=C(Cl)C=C1 (4)	> 100	-	I	I	I	I
COC1=CC=C(C=C1)C(=N\NC(N)=S)\C1=CC=C(C)C=C1 (5)	10.20	188.4	A	I	I	I
COC1=CC=C(\C(=N\NC(N)=O)C2=CC=CC=C2)C(O)=C1 (6)	> 100	273.3	I	I	I	I
CC1=CC=C(C=C1)C(=N/NC(N)=S)\C1=CC=C(Cl)C=C1 (7)	14.3	27.8	A	I	I	A
N(N=C(C1 = CC = CC=C1)C1=CC = CC = C1)C1 = NC(=CS1) C1=CC=CC=C1 (8)	> 100	-	I	I	I	I
N\C(NN = C(C1=CC = CC = C1)C1 = CC = CC = C1) = N\C( = S) NC1=CC=CC=C1 (9)	65.4	130.6	I	I	I	I
NC(=S)N\N=C(\C1=CC=CC=C1)C1=CC=C(C=C1)C(O)=O (10)	> 100	-	I	I	I	I
CC1=CC=C(C=C1)C(=N/NC(N)=S)\C1=CC=CC=C1 (11)	31.0	32.5	M	I	A	A
CC1=CC=C(C=C1)C(=N/NC1=NC(=CS1)C1=CC=CC=C1)\ C1=CC=C(C=C1)[N+]([O-])=O (12)	96.6	-	I	I	I	A
NC(=O)NN=C(C1=CC=CC=C1)C1=CC=CC=C1 (13)	> 100	61.1	I	I	I	I
CC1=CC=C(C=C1)C(=O)C1=CC=CC=C1 (14)	> 100	-	I	I	I	I
Pentamidine	19.5		A	-	-	-
Amphotericin B	5.1		A	-	-	-

• A: active (IC50 > 30 μM), M: moderate (IC50 > 30 and < 40 μM), I: inactive (IC50 > 40 μM). IC50: half-maximal inhibitory concentrations; CC50: 50% cytotoxicity concentrations; ANN: artificial neural networks; RF: Random Forest.

Overall, virtual screening efficiently identified the most promising candidates for experimental validation, underscoring its value as a powerful complementary approach to streamline and accelerate drug discovery efforts against leishmaniasis.

QSAR and “in vitro” assays

The in vitro assays identified compounds 5 and 7 as active, while compound 11 exhibited moderate activity. All other compounds were classified as inactive (Table 2). Structurally, the active and moderate compounds (5, 7, and 11) feature a thiosemicarbazone moiety, whereas the inactive compounds possess a carbonyl group in the same position. Additionally, the presence of a methyl group at position 4 is unique to these three derivatives, further distinguishing them from the inactive compounds. Compared to the reference drugs, the IC50 values of compounds 5 and 7 were lower than those of pentamidine but higher than amphotericin B. Notably, compound 5 showed a higher selectivity index compared to the controls, indicating potentially reduced cytotoxicity. In contrast, compound 10, which includes a thiosemicarbazone moiety but lacks the methyl group at position 4 (replaced by a carboxyl group), demonstrated no significant activity (IC₅₀ > 100 μM). These observations suggest that both the thiosemicarbazone and methyl groups play a critical role in enhancing the antileishmanial activity of benzophenone and xanthone derivatives.

Compounds 5 and 7 also differ in the electronegative substituents at position 11, featuring a methoxy (-OMe) and chloro (-Cl) group, respectively. These substituents appear to enhance the antileishmanial activity against L. amazonensis promastigotes. Regarding cytotoxicity, compound 5, with a chloro group, exhibited lower cytotoxicity and higher selectivity compared to compound 7. These results suggest that, in addition to the primary structural features (thiosemicarbazone and 4-methyl groups), substitutions at position 11 with electronegative groups, such as chloro or methoxy, further contribute to the antileishmanial activity. The contrasting in vitro and predicted activities (Table 2) emphasize the utility of computational tools in refining compound classification. Among the tested models, the ANN method (Figure S2, SI section) demonstrated superior performance, achieving an accuracy of 86% for classifying the activity of synthesized derivatives during external validation (Table 1). This highlights the potential of ANN-based predictions in supporting the design and screening of new benzophenone derivatives for antileishmanial applications.

Conclusions

In this study, two promising benzophenone derivatives were identified as potential antileishmanial agents, demonstrating significant biological activity coupled with favorable selectivity indices. Moreover, the ANN prediction model showed high accuracy for predicting and classifying novel derivatives within the benzophenone and xanthone chemical space, reinforcing its value in virtual screening and drug design. Structural features, particularly the thiosemicarbazone group, the 4-methyl substitution, and the presence of chloro or methoxy groups at position 11, were identified as key contributors to antileishmanial activity, aligning with the computationally selected descriptors. These findings provide valuable insights for the design and development of novel antileishmanial agents, offering a foundation for further studies in drug discovery.

Supplementary Information

Supplementary information (Tables S1-S5, Figures S1-S2), is available free of charge at http://jbcs.sbq.org.br as PDFfile.

Data Availability Statement

Data supporting the findings of this study (Database Datasets (training, test, and scrambling), descriptor values, external test sets, and outputs from Random Forest, ANN, and J48 models, as well as SMILES of the compounds investigated) are available at link DOI https://doi.org/10.5281/zenodo.14657181.

Acknowledgments

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)-Finance Code 001. The authors acknowledge the following funding agencies for the fellowships and financial support for this research project: Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) 408115/2023-8, 316204/2021-8, and 406837/2021-0. Fundação de Amparo à Pesquisa do Estado de Minas Gerais (FAPEMIG) APQ-05218-23, APQ-00544-23, APQ-03701-17 (Programa Pesquisa para o SUS, PPSUS), and BPD-00760-22.

References

• ¹ Bekhit, A. A.; El-Agroudy, E.; Helmy, A.; Ibrahim, T. M.; Shavandi, A.; Bekhit, A. E.-D. A.; Eur. J. Med. Chem. 2018, 108, 229. [Crossref]
  » Crossref
• ² World Health Organization (WHO); http://www.who.int/en/ news-room/fact-sheets/detail/leishmaniasis, accessed in January 2025.
  » http://www.who.int/en/ news-room/fact-sheets/detail/leishmaniasis
• ³ Wyllie, S.; Thomas, M.; Patterson, S.; Crouch, S.; De Rycker, M.; Lowe, R.; Gresham, S.; Urbaniak, M. D.; Otto, T. D.; Stojanovski, L.; Simeons, F. R. C.; Manthri, S.; MacLean, L. M.; Zuccotto, F.; Homeyer, N.; Pflaumer, H.; Boesche, M.; Sastry, L.; Connolly, P.; Albrecht, S.; Berriman, M.; Drewes, G.; Gray, D. W.; Ghidelli-Disse, S.; Dixon, S.; Fiandor, J. M.; Wyatt, P. G.; Ferguson, M. A. J.; Fairlamb, A. H.; Miles, T. J.; Read, K. D.; Gilbert, I. H.; Nature 2018, 560, 192. [Crossref]
  » Crossref
• ⁴ Maia, C.; Conceição, C.; Pereira, A.; Rocha, R.; Ortuño, M.; Muñozid, C.; Jumakanova, Z.; Pérez-Cutillas, P.; Özbel, Y.; Töz, S.; Baneth, G.; Monge-Maillo, B.; Gasimov, E.; Van der Stede, Y.; Torres, G.; Gossner, C. M.; Berriatua, E.; PLoS Negl. Trop. Dis. 2023, 17, e0011497. [Crossref]
  » Crossref
• ⁵ Upadhyay, A.; Chandrakar, P.; Gupta, S.; Parmar, N.; Singh, S. K.; Rashid, M.; Kushwaha, P.; Wahajuddin, M.; Sashidhara, K. V.; Kar, S.; J. Med. Chem. 2019, 62, 5655. [Crossref]
  » Crossref
• ⁶ Manzano, J. I.; Cueto-Díaz, E. J.; Olías-Molero, A. I.; Perea, A.; Herraiz, T.; Torrado, J. T.; Alunda, J. M.; Gamarro, F.; Dardonville, C.; J. Med. Chem. 2019, 62, 10664. [Crossref]
  » Crossref
• ⁷ de Almeida, L.; Alves, K. F.; Maciel-Rezende, C. M.; Jesus, L. O. P.; Pires, F. R.; Viegas Jr., C.; Izidoro, M. A.; Júdice, W. A. S.; dos Santos, M. H.; Marques, M. J.; Biomed. Pharmacother. 2015, 75, 93. [Crossref]
  » Crossref
• ⁸ Niazi, S. K.; Mariam, Z.; Int. J. Mol. Sci. 2023, 24, 11488. [Crossref]
  » Crossref
• ⁹ Nilar; Nguyen, L. H. D.; Venkatraman, G.; Sim, K. Y.; Harrison, L. J. ; Phytochemistry 2005, 66, 1718. [Crossref]
  » Crossref
• ¹⁰ Seo, E. K.; Kim, N. C.; Wani, M. C.; Wall, M. E.; Navarro, H. A.; Burgess, J. P.; Kawanishi, K.; Kardono, L. B. S.; Riswan, S.; Rose, W. C.; Fairchild, C. R.; Farnsworth, N. R.; Kinghorn, A. D.; J. Nat. Prod. 2002, 65, 299. [Crossref]
  » Crossref
• ¹¹ Xue, Q.; Chen, Y.; Yin, H.; Teng, H.; Qin, R.; Liu, H.; Li, Q.; Mei, Z.; Yang, G. ; Bioorg. Chem. 2020, 104, 104339. [Crossref]
  » Crossref
• ¹² Pereira, I. O.; Marques, M. J.; Pavan, A. L. R.; Codonho, B. S.; Barbiéri, C. L.; Beijo, L. A.; Doriguetto, A. C.; D’Martin, E.C.; dos Santos, M. H.; Phytomedicine 2010, 17, 339. [Crossref]
  » Crossref
• ¹³ Chantarasriwong, O.; Milcarek, A. T.; Morales, T. H.; Settle, A. S.; Rezende Jr., C. O.; Althufairi, B. D.; Theodoraki, M. A.; Alpaugh, M. L.; Theodorakis, E. A.; Eur. J. Med. Chem. 2019, 168, 405. [Crossref]
  » Crossref
• ¹⁴ Pontius, A.; Krick, A.; Kehraus, S.; Brun, R.; König, G. M.; J. Nat. Prod. 2008, 71, 1579. [Crossref]
  » Crossref
• ¹⁵ Kelly, J. X.; Ignatushchenko, M. V.; Bouwer, H. G.; Peyton, D. H.; Hinrichs, D. J.; Winter, R. W.; Riscoe, M.; Mol. Biochem. Parasitol. 2003, 126, 43. [Crossref]
  » Crossref
• ¹⁶ Lo, Y. C.; Rensi, S. E.; Torng, W.; Altman, R. B. ; Drug Discovery Today 2018, 23, 1538. [Crossref]
  » Crossref
• ¹⁷ ChemBioDraw Ultra, 13.0; PerkinElmer, Inc., Cambridge, MA, USA, 2013. [Link] accessed in January 2025
  » Link
• ¹⁸ MarvinSketch, 20.5 and 4.0.0; ChemAxon Ltd., Budapest, Hungary, 2020. [Link] [Link] accessed in January 2025
  » Link» Link
• ¹⁹ KNIME Analytics Platform, version 5.1; KNIME AG, Zurich, Switzerland, 2009. [Link] accessed in January 2025
  » Link
• ²⁰ Classification and Multivariate Analysis for Complex Data Structures; Fichet, B.; Piccolo, D.; Verde, R.; Vichi, M., eds.; Springer: Berlin, 2011. [Crossref]
  » Crossref
• ²¹ Khalid, B.; Ghorab, H.; Benkhemissa, A.; J. Ind.Chem. Soc. 2022, 99, 100672. [Crossref]
  » Crossref
• ²² Witten, I. H.; Ian, H.; Frank, E.; Hall, M. A.; Mark, A.; Data Mining: Practical Machine Learning Tools and Techniques, 3^rd ed.; Morgan Kaufmann: Burlington, US, 2010.
• ²³ Salzberg, S. L.; Inf. Control 1964, 7, 55. [Crossref]
  » Crossref
• ²⁴ Open Babel, version 4.1.0; Open Babel Development Team, Pittsburgh, PA, USA, 2023; O’Boyle, N. M.; Banck, M.; James, C. A.; Morley, C.; Vandermeersch, T.; Hutchison, G.; J. Cheminf. 2011, 3, 33. [Crossref]
  » Crossref
• ²⁵ Melagraki, G.; Afantitis, A. ; Chemom. Intell. Lab. Syst. 2013, 123, 9. [Crossref]
  » Crossref
• ²⁶ PaDEL-Descriptor, version 2.2.121; Singapore, 2011 [Link]; Yap, C. W.; J. Comput. Chem. 2010, 32, 1466. [Crossref]
  » Link» Crossref
• ²⁷ Bunkhumpornpat, C.; Sinapiromsaran, K.; Lursinsap, C. In Advances in Knowledge Discovery and Data Mining, 1^st ed.; Springer: Berlin, Heidelberg, 2009, p. 5476. [Crossref]
  » Crossref
• ²⁸ Santos, W. T.; Katchborian-Neto, A.; Viana, G. S.; Ferreira, M. S.; Martins, L. C.; Vale, T. C.; Murgu, M.; Dias, D. F.; Soares, M. G.; Chagas-Paula, D. A.; Paula, A. C. C.; ACS Chem. Neurosci. 2024, 15, 3168. [Crossref]
  » Crossref
• ²⁹ Farias, B. F.; Ferreira, M. S.; Miranda, D. O.; Nunes, T. R.; Pereira, N. F.; Espuri, P. F.; Januario, J. P.; Colombo, F. A.; Marques, M. J.; Zanin, J. L. B.; Soares, M. G.; de Souza, T. B.; Carvalho, D. T.; Chagas-Paula, D.; Dias, D. F.; Zenodo 2025. [Crossref]
  » Crossref
• ³⁰ Witten, I. H.; Frank, E.; Hall, M. A.; Pal, C. J.; Weka, 3.8.0; University of Waikato, Hamilton, New Zealand, 2016. [Crossref]
  » Crossref
• ³¹ Zhan, X.; Yu, S.; Li, Y.; Zhou, Z.; Cao, H.; Tang, G.; Landscape Urban Plann. 2024, 242, 104950. [Crossref]
  » Crossref
• ³² Pham, B. T.; Nguyen, M. D.; Nguyen-Thoi, T.; Ho, L. S.; Koopialipoor, M.; Kim Quoc, N.; Armaghani, D. J.; van Le, H.; Transp. Geotech. 2021, 27, 100508. [Crossref]
  » Crossref
• ³³ Saad, G.; Khadour, A.; Kanafani, Q.; Egypt. J. Radiol. Nucl. Med. 2016, 47, 1803. [Crossref]
  » Crossref
• ³⁴ Villarroel Ordenes, F.; Silipo, R.; J. Bus. Res. 2021, 137, 393. [Crossref]
  » Crossref
• ³⁵ López, A. F. F.; Martínez, O. M. M.; Hernández, H. F. C.; J. Mol. Struct. 2021, 1225, 129142. [Crossref]
  » Crossref
• ³⁶ Pingaew, R.; Prachayasittikul, V.; Worachartcheewan, A.; Thongnum, A.; Prachayasittikul, S.; Ruchirawat, S.; Prachayasittikul, V.; Heliyon 2022, 8, e10067. [Crossref]
  » Crossref
• ³⁷ Castillo-Garit, J. A.; Barigye, S. J.; Pham-the, H.; Pérez-Doñate, V.; Torrens, F.; Pérez-Giménez, F.; SAR QSAR Environ. Res. 2021, 32, 71. [Crossref]
  » Crossref
• ³⁸ Folquitto, L. R. S.; Nogueira, P. F.; Espuri, P. F.; Gontijo, V. S.; de Souza, T. B.; Marques, M. J.; Carvalho, D. T.; Júdice, W. A. S.; Dias, D. F.; Med. Chem. Res. 2017, 26, 1149. [Crossref]
  » Crossref
• ³⁹ Dias, M. C. F.; Gularte, T. Q.; Teixeira, R. R.; Santos, J. A. N.; Pilau, E. J.; Mendes, T. A. O.; Demuner, A. J.; dos Santos, M. H.; J. Braz.. Chem.Soc. 2019, 30, 97. [Crossref]
  » Crossref
• ⁴⁰ Maciel-Rezende, C. M.; de Almeida, L.; Costa, É. D. M.; Pires, F. R.; Alves, K. F.; Viegas Jr., C. V.; Dias, D. F.; Doriguetto, A. C.; Marques, M. J.; Dos Santos, M. H. ; Bioorg. Med. Chem. 2013, 21, 3114. [Crossref]
  » Crossref
• ⁴¹ Arshia; Ahad, F.; Ghouri, N.; Kanwal; Khan, K. M.; Perveen, S.; Choudhary, M. I.; R. Soc. OpenSci. 2018, 5, 171771. [Crossref]
  » Crossref
• ⁴² Mishra, A.; Vinayagam, J.; Saha, S.; Chowdhury, S.; Roychowdhury, S.; Jaisankar, P.; Majumder, H. K.; Pharmacol. Res. Perspect. 2014, 2, e00070. [Crossref]
  » Crossref
• ⁴³ Dagnino, A. P. A.; Mesquita, C. S.; Dorneles, G. P.; Teixeira, V. D. O. N.; De Barros, F. M. C.; Vidal Ccana-Ccapatinta, G.; Fonseca, S. G.; Monteiro, M. C.; Rodrigues Jr., L. C. R.; Peres, A.; von Poser, G. L.; Romão, P. R. T.; Camb. Law J. 2018, 145, 1199. [Crossref]
  » Crossref
• ⁴⁴ Fromentin, Y.; Gaboriaud-Kolar, N.; Lenta, B. N.; Wansi, J. D.; Buisson, D.; Mouray, E.; Grellier, P.; Loiseau, P. M.; Lallemand, M. C.; Michel, S.; Eur. J. Med. Chem. 2013, 65, 284. [Crossref]
  » Crossref
• ⁴⁵ Hay, A. E.; Merza, J.; Landreau, A.; Litaudon, M.; Pagniez, F.; le Pape, P.; Richomme, P.; Fitoterapia 2008, 79, 42. [Crossref]
  » Crossref
• ⁴⁶ Baldim, J. L.; Alcantara, B. G. V. De; Domingos, O. D. S.; Soares, M. G.; Caldas, I. S.; Novaes, R. D.; Oliveira, T. B.; Lago, J. H. G.; Chagas-Paula, D. A. ; Oxid. Med. Cell Longev. 2017, 2017, ID 3789856. [Crossref]
  » Crossref
• ⁴⁷ Todeschini, R.; Consonni, V.; Handbook of Molecular Descriptors; Wiley-VCH Verlag GmbH: Weinheim, Germany, 2000. [Crossref]
  » Crossref
• ⁴⁸ Jamshidnezhad, A.; Anjomshoa, Z.; Hosseini, S. A.; Azizi, A.; Inform. Med. Unlocked 2021, 24, 100614. [Crossref]
  » Crossref
• ⁴⁹ Srivastava, S.; Int. J. Comput. Appl. 2014, 88, 26. [Crossref]
  » Crossref
• ⁵⁰ Gosav, S.; Praisler, M.; Dorohoi, D. O.; J. Mol. Struct. 2007, 834-836, 188. [Crossref]
  » Crossref

Edited by

• Editor handled this article:
  Paula Homem-de-Mello (Associate)

Publication Dates

• Publication in this collection
  21 Mar 2025
• Date of issue
  2025

History

• Received
  24 June 2024
• Accepted
  14 Feb 2025