High-throughput screening and meta-analysis for lead compounds in antimalarial drug discovery

Abstract

Background

The discovery of novel antimalarial drugs against Plasmodium falciparum has become globally urgent due to the consistent increase in mortality, morbidity, and drug resistance in endemic areas.

Methods

Using an in-house library, novel antimalarial agents were identified through in vitro high-throughput screening (HTS) and meta-analysis. Hit compounds were selected from the primary HTS at 10 µM and confirmed in a dose-dependent manner to determine their IC₅₀ values. The identified hit molecules were further selected based on the following criteria: novelty, antimalarial activity (IC₅₀), pharmacokinetic properties (C_max and T_1/2), mechanism of action, and safety (in vitro and in vivo) (CC₅₀, SI, LD₅₀, and MTD). In vitro and in vivo antimalarial activity against drug-sensitive and resistant strains (3D7, NF54 and K1, Dd2, Dd2-R539T (+), and CamWT-C580Y (+)) and the rodent Plasmodium berghei parasite-infected animal model, respectively, were subsequently used to validate hit compounds.

Results

Based on the top 3% threshold, 256 compounds were selected for dose‒response curve analysis from the HTS. Among them, 110 compounds without published research related to Plasmodium and 157 compounds with IC₅₀ values < 1 µM were identified. Further analysis confirmed 69 compounds with median lethal doses, maximum tolerated doses or treated doses greater than 20 mg/kg, 48 compounds with FDA approval, 29 compounds characterized by C_max > IC₁₀₀ and T_1/2 > 6 h, and 38 compounds with a potential mechanism in Plasmodium. Next, 19 candidates were further evaluated for in vitro inhibition of drug-resistant parasites and inhibition in a mouse model of P. berghei parasites. Notably, three potent inhibitors were identified, exhibiting 95.9% and 81.4% suppression via oral delivery at a dose of 50 mg/kg ONX-0914 and methotrexate, respectively, and 96.4% suppression via intraperitoneal delivery at a dose of 20 mg/kg of an antimony compound. In addition, strong in vitro antimalarial activity was demonstrated against CQ- and ART-sensitive and resistant strains (IC₅₀ < 500 nM).

Conclusions

Combining HTS and meta-analysis provides a robust method for screening antimalarial candidate compounds and identifying new hits with in vivo activity as candidates to treat drug-resistant malarial strains.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12936-025-05587-0.

Background

Malaria, which is caused by Plasmodium parasites, remains a major global health burden, with an estimated 263 million cases and 597,000 deaths annually [1]. Along with increasing incidence, high rates of antimalarial drug resistance have been reported in Africa, the Americas, and Southeast Asia. Specifically, therapeutic efficacy studies (TESs) have shown treatment failure with both single or combination regimens, including with newer drugs such as dihydroartemisinin-piperaquine and artesunate-amodiaquine, primarily due to genetic mutations associated with drug resistance [2–5]. To achieve the goal of malaria eradication, there is an urgent need to expand antimalarial drug discovery efforts. However, drug discovery and development is a long, costly, and high-risk process, typically taking over a decade and costing an estimated over $1–2 billion for a drug before approval for human treatment [6, 7].

Among the advancements in technology that have accelerated drug discovery, high-throughput screening (HTS) has emerged as a powerful method for screening millions of compounds in pharmaceutical libraries. Over the past decade, HTS has significantly contributed to the antimalarial development pipeline, leading to the identification of new chemotypes that target the disease-causing asexual stages of Plasmodium [8–10]. The pharmaceutical industry has developed HTS based on two primary approaches: phenotypic (whole-cell) screening and target-based screening. Phenotypic screening evaluates changes in parasites upon exposure to antimalarial compounds, whereas target-based screening assesses the effects of compounds on purified target proteins [11]. Recent analyses have shown that phenotypic approaches are more successful for small molecule inhibitors, largely due to advancements in high-resolution optical microscopy and improvements in image analysis software [12–15]. In phenotypic HTS, parasite-infected red blood cells (RBCs) are stained with nucleic acid-conjugated fluorescence dyes, followed by the detection and classification of parasites at different developmental stages. Compared with the conventional SYBR Green I assay, which is often used as a reference method, phenotypic HTS has demonstrated enhanced accuracy in detecting antimalarial activity [11–14].

Following the HTS stage, an enormous workload remains for subsequent steps, including ‘hit’ confirmation, lead optimization, and preclinical evaluations before a candidate drug can proceed to clinical trials and receive approval from the U.S. Food and Drug Administration (FDA) [6, 16]. Drug discovery is an inherently challenging process, and the development of antimalarials presents additional difficulties due to several key factors, including the need for high tolerance and safety across diverse endemic populations, ease of administration, short-term duration, availability of combination regimens to address noncompliance and prevent resistance development, and the requirement for low-cost production to ensure accessibility [17]. To address these challenges, candidate compounds identified through HTS undergo lead optimization and efficacy and safety evaluations in animal models. However, only a few potential compounds progress to clinical trials. The failure rate in clinical drug development remains high, and approximately 90% of drug candidates fail during development, with failure rates of approximately 17% in lead optimization, 7% in preclinical testing, and 66% in Phase I clinical trials from 2010 to 2017 [18]. To achieve a reduced rate of failure, bioinformatic and virtual screening approaches have been conducted to prioritize the most promising lead candidates. Nevertheless, virtual screening results must be validated through in vivo or in vitro models to ensure their reliability. Another effective bioinformatics approach involves systematically collecting and analysing parameters from previous trials (in vitro, in vivo, and clinical trial studies) to facilitate evidence-based hit selection, thereby improving efficiency and saving time.

To overcome the bottleneck in lead discovery, this study employs an integrated approach that combines HTS and meta-analysis. This strategy enables the identification of compounds with strong in vitro antimalarial activities while also considering optimal parameters such as novelty, potency, mechanism of action, cytotoxicity, safety, and pharmacokinetics, as reported in previous studies on Plasmodium spp. Applying this approach before conducting in vivo testing was intended to reduce the high attrition rate inherent in the early phase of drug discovery, thereby increasing the chances of success and increasing cost-effectiveness.

Methods

Ethics statement

All experiments were conducted following relevant guidelines and regulations, and all protocols involving human samples were approved by the Institutional Ethical Committee of Kangwon National University Hospital (IRB No. KNUH-B-2021-06-034). The animal study was reviewed and approved by the Institutional Animal Care and Use Committee of Kangwon National University and was conducted in accordance with the Ethical Guidelines for Animal Experiments of Kangwon National University (KIACUC-22-133/KNU/KMRL/HET, dated July 1, 2022).

Compound library preparation

An in-house compound library of 9,547 small molecules, including FDA-approved small molecules from Institut Pasteur Korea, was used for screening. Stock solutions of these compounds were prepared in 100% dimethyl sulfoxide (DMSO) (Sigma‒Aldrich), and chloroquine diphosphate salt (CQ) (Sigma‒Aldrich) was dissolved in distilled water (DDW), stored at − 20 °C and diluted as needed. Library compounds were diluted in 5 µL of phosphate-buffered saline (PBS) and transferred using Hummingwell (CyBio^®) into 384-well glass plates.

In vitro culture of Plasmodium falciparum asexual stages and synchronization

Plasmodium falciparum parasites, including the CQ-sensitive strains (3D7 and NF54), CQ-resistant strains (K1 and Dd2), artemisinin (ART)-sensitive strain (Cam WT), ART-resistant strain (CamWT-C580Y (+)) and CQ and ART-resistant strain (Dd2-R539T (+)) (BEI Resources) [19], were cultured in vitro in O⁺ human RBCs in RPMI 1640 (Sigma‒Aldrich) supplemented with 100 µM hypoxanthine, 12.5 µg/ml gentamicin, 0.5% (wt/vol) Albumax™ I (Thermo Fisher) and 2 g/L sodium bicarbonate (WELGENE Inc.) at 37 °C in 1% O₂ and 5% CO₂ in N₂ as described previously [20].

The maintained parasites were double synchronized at the ring stage via 5% sorbitol (wt/vol) (Sigma‒Aldrich) treatment, as previously described [21], and further cultivated through one complete cycle to test antimalaria drug sensitivity.

High-throughput screening

Malaria drug sensitivity assay optimization

The compounds were arrayed at a single final drug concentration of 10 µM or in a dose-dependent manner with concentrations ranging from 10 µM to 20 nM (1 in 2 serial dilutions) in a final DMSO (Sigma‒Aldrich) concentration of 1% per well of 50 µL sample. Plasmodium falciparum (strain 3D7) cultures were dispensed in drug-treated 384-well plates with 1% schizont-stage parasites at 2% haematocrit and incubated for 72 h in a malaria culture chamber with mixed gas at 37 °C.

Image-based antimalarial drug screening

After the antimalarial drug sensitivity assay plate was diluted to 0.02% haematocrit in PhenolPlate™ 384-well ULA-coated microplates (PerkinElmer, Inc.), the plate was stained with a solution for RBCs and nucleic acid stain comprising 1 µg/mL wheat agglutinin–Alexa Fluor™ 488 conjugate (Thermo Fisher Scientific) and 0.625 µg/mL Hoechst 33,342, trihydrochloride, and trihydrate (Thermo Fisher Scientific) in 4% paraformaldehyde (Thermo Fisher Scientific) to stain and fix the culture. The incubation conditions for the image screening plates were 20 min at room temperature (RT) for image acquisition. Nine microscopy image fields from each well were obtained via Operetta^® CLS™ (Perkin Elmer, Inc.) via a 40 × water immersion lens, and the final resolution of the acquired images was 0.299 µm pixel size, 16 bits per pixel, and 1080 × 1080 pixels. The acquired antimalarial drug screening images were transferred to Columbus™ (Perkin Elmer, Inc.) version 2.9 software via a central database server for image analysis.

SYBR green I-based flow cytometry analysis for growth inhibition and drug sensitivity

The SYBR Green I assay was performed as described in a previous study with modifications [22]. Lysis buffer was prepared with 20 mM Tris pH 7.5 (Sigma‒Aldrich), 5 mM EDTA (Sigma‒Aldrich), 0.008% (wt/vol) saponin (Sigma‒Aldrich), and 0.08% (vol/vol) Triton X-100 (Sigma‒Aldrich) containing 1 × final concentration of SYBR Green I nucleic acid gel stain from 10,000 × stock in dimethyl sulfoxide (DMSO, Thermo Fisher Scientific). To lyse 50 µL of culture, 15 µL of lysis buffer was added to each well of malaria drug sensitivity assay plates via a multipipette. The plates were incubated at room temperature (RT) for 1 h, after which the fluorescence intensity was measured at 485 nm excitation and 535 nm emission with a SpectraMax M5 multimode microplate reader (Molecular Devices).

The fluorescence-activated cell sorting (FACS) method was modified on the basis of a previously reported procedure [23]. After incubation, the cultures were washed twice with phosphate-buffered saline (PBS), fixed with 0.05% glutaraldehyde (Sigma‒Aldrich) for 10 min, washed twice before being stained with SYBR Green I (Thermo Fisher) at 0.2 × dilution for 10 min, and washed twice with PBS. The samples were analysed with an Accuri C6 flow cytometer (Accuri Cytometers, Inc.). A total of 200,000 cells were recorded, and parasitaemia was determined by gating the population. The percent inhibition was calculated via the following formula: % growth inhibition = 100-(100 × (intervention well parasitaemia—uninfected well parasitaemia)/(infected well parasitaemia—uninfected well parasitaemia)). The IC₅₀ value was defined as (log(inhibitor) vs. response-variable slope (four parameters)) [24]. In each case, at least two independent assays were conducted, each in triplicate, and the data are shown as the mean ± standard deviation.

Meta-analysis

Information on the top compounds of the library was collected from various online databases, including PubChem, Drug.com, Google, and the FDA, before being stored in Excel. A multifactor compound was evaluated for novel antimalarial activity (IC₅₀, pharmacokinetics [T_1/2 and C_max], delivery, structure, and potent mechanism), safety (in vitro cytotoxicity [CC₅₀] and in vivo lethal dose [LD₅₀]), maximum tolerated dose (MTD), and FDA approval. The selection process combined a basic weighting system- applied consistently across all compounds for key parameters such as IC₅₀, CC₅₀, SI, C_max, and T_1/2 –with subjective expert judgment based on nolvety, mechanism of action, and potential for development (cost, FDA approval, and potent bioavailbility). This hybrid approach was evaluated using guidelines for antimalarial agents from the WHO, Medicines for Malaria Venture (MMV), and previous studies [25–30].

Haemolytic assay

The haemolytic activity of the compound was evaluated according to previously established methods [31, 32]. Fresh RBCs were washed three times with PBS, adjusted to a 2% haematocrit, and then added to a 96-well plate, after which 100 µl of the compound in PBS (at different concentrations prepared by serial dilution) was added. PBS alone served as the baseline control, whereas 0.4% Triton X-100 in PBS was used as the positive control. After incubation at 37 °C for 3 h, the samples were centrifuged, and the supernatant was collected. Haemolytic activity was quantified via a spectrophotometer, with the absorbance (abs) measured at 415 nm. The experiment was conducted in three independent assays. The percent haemolysis was calculated via the following formula: % haemolysis = (abs of sample-abs of blank sample)/(abs of positive control) × 100.

Cytotoxicity in a human cell line

Cytotoxicity was assessed against human embryonic kidney 293 T (HEK293T) cells via a Cell Counting Kit-8 (CCK-8) (Dojindo), as described previously [33, 34]. The HEK293T cell line was obtained and maintained according to American Type Culture Collection (ATCC) guidelines [35]. A 90 µL suspension of approximately 5000 cells was dispensed into each well of a 96-well plate, followed by incubation for 24 h in a humidified incubator at 37 °C with 5% CO₂. The cells were then treated with 10 µL of serially diluted compound solutions and incubated for 48 h. To assess cell viability, 10 µL of CCK-8 solution was added to each well, followed by incubation for an additional 1–4 h. Absorption was measured at 450 nm using a SpectraMax M5 plate reader (Molecular Devices). The experiment was performed in three independent assays. Cytotoxicity was calculated via the following formula: % cytotoxicity = (abs of sample-abs of blank sample)/(abs of positive) × 100. The half maximal cytotoxic concentration value (CC₅₀) was defined via log-linear interpolation of the cytotoxicity curves [24].

In vivo assay of antimalarial activity

Animals

For the in vivo experiments, female mice (BALB/c) with an average weight of 25 ± 2 g and 6–8 weeks of age were purchased from Orient (Orient Bio) and maintained in an animal facility in polypropylene cages, with 5 animals per cage, at 25 ± 2 °C with a light:dark photoperiod of 12:12 h and 70% relative humidity. The animals were fed a standard laboratory diet and provided water ad libitum. The mice were acclimatized to their experimental environment for 4 d before the experiment started.

Murine model of malaria

The assessment of in vivo efficacy of the Plasmodium berghei (strain ANKA) parasite was modified on the basis of previously described methods [28]. Female BALB/c mice, with 5 mice in each group, were inoculated with 10⁶ P. berghei-infected RBCs taken from passaged mice (excluding those in the negative control group). The compounds were dissolved in vehicles (10% DMSO, 3% ethanol, 7% Tween-80, or 80% distilled water [DDW]), vortexed well before injection, and administered at 48, 72, 96, and 120 h after infection at doses of 50 mg/kg per oral (PO) or 20 mg/kg intraperitoneal (IP) administration. The remaining groups each received the vehicle (200 µl), 20 mg/kg CQ dissolved in DDW, or no treatment (positive control) or were not infected (negative control). The parasitaemia of the mice was assessed by obtaining blood via the tail vein, followed by Giemsa staining and observation under a 100 × objective. At least 1000 RBCs were observed via microscopy to determine parasitaemia. Twenty-four hours after the last treatment, the percent suppression at 4 d was determined via the following formula: % suppression at 4 days = 100–100* parasitaemia (intervention)/parasitaemia (positive). Parameters such as adverse events, mean survival days, average survival rates, clinical parameters, and weights were also evaluated.

For the dose range test, ONX-0914 and antimony were tested at 4 different doses (50, 25, 10, and 5 mg/kg) and (25, 20, 10, and 5 mg/kg), respectively, via the IP routes of administration. The follow-up panel was the same as that described above. The ED₅₀ was calculated as (log[dose] vs. response (4-d suppression)-variable slope [four parameters]) [24].

Histopathology

The organs (brain, heart, lung, liver, and kidney) of at least three mice from each group were collected after sacrifice, washed in PBS, fixed in 10% formalin buffer, and the changes in colour and size were evaluated [36]. The size of the organs was defined as the height × width (mm). For histological examination, the organs were cut at an appropriate thickness and embedded in paraffin via standard procedures. The paraffin-embedded tissues were then sectioned (3 µm thick) and stained with haematoxylin and eosin. Histological examination was performed via a light microscope (BX53, Olympus). In the histological examination, organ lesions, if any, were recorded and graded as minimal (1 +), mild (2 +), moderate (3 +), or severe (4 +).

Statistical analysis

The data were analysed via GraphPad Prism (GraphPad Software) and Microsoft Excel 2018 (Microsoft). For comparisons of experimentally measured values across groups, Student’s t tests and Mann–Whitney U tests were used. Spearman analysis was used to examine the correlation of nonparametric data; a p < 0.05 indicated a significant difference.

Results

Development of a phenotypic HTS platform for screening

To identify new antimalarial agents from an in-house library of 9547 compounds, an HTS platform was established for primary confirmation and hit confirmation (Figs. 1, 2, Supplementary Fig. 1). HTS of P. falciparum 3D7 by SYBR Green I with a single dose of 10 µM identified 256 hits compounds with > 80% growth inhibition. The Z’-factor of 0.55 confirmed the robustness and validity of the method (Fig. 2a). The IC₅₀ values of the 256 hit compounds from the primary screen were investigated via SYBR Green I. Among them, 50 priority compounds were further evaluated at different doses via both imaging and SYBR Green I assays as previous study [37]. The accuracy of the image-based assay was validated by its strong correlation (r = 0.97, p < 0.001) with that of the SYBR Green I fluorescence assay in determining the DRC and IC₅₀ for CQ and ART (Fig. 2b-c).

Fig. 1.

Progression cascade of drug compound selection. The 256 top compounds from HTS with an inhibition rate of more than 98% at 10 µM were selected from 9547 compounds. Further selection combined basic weighting and subjective judgment. A basic weighting involved major parameters incorporating antimalarial activity (dose–response curve (DRC), 50% inhibitory concentration (IC₅₀), 50% cytotoxic concentration (CC₅₀), selective index (SI), in vivo toxicity, pharmacokinetic (PK) and minor factors including cost and function relevance. Subjective expert judgment was based on novelty, chemical structure, delivery, mechanism, and U.S Food and Drug Administration (FDA) approval. Nineteen compounds were included in the confirmation round. Ultimately, 3 hits were confirmed

Fig. 2.

Phenotypic high-throughput screening (HTS) assay development and primary screen-based SYBR Green I assay. a Primary screen of HTS with P. falciparum 3D7 using the SYBR Green I assay with a single dose of 10 µM. Red circle = positive control (artemisinin); black circle = DMSO negative control; blue circle = compound test. A total of 256 hit compounds were selected from 9547 compounds with Z value at 0.71. b Correlation coefficients for IC₅₀ values between the two methods. Fifty priority compounds from the primary screen had IC₅₀ values determined via two methods. The image-based assay had a very strong positive correlation with the SYBR green I assay, with r = 0.97 (p < 0.001). c Dose–response curves of chloroquine (CQ) and artemisinin (ART) were determined via SYBR Green I and image-based assays

Integration of bioinformatics and meta-analysis for hit selection

The top 3% of the compounds (256) from HTS were further analysed based on the desired profile, including novelty, antimalarial activity, safety, chemical structure, and mechanism of action (Figs. 1, 2, 3). The information on the compounds was collected from previous studies. Among these, 110 compounds demonstrated novelty within the malaria research field and were selected for further studies (Fig. 3a). IC₅₀ values were used to prioritize compounds with potent antimalarial activity, and 157 compounds presented IC₅₀ values under 1 µM, a widely accepted cut-off for potent in vitro activity (Fig. 3b). The selectivity index (SI), which is calculated based on the half-maximal cytotoxic concentration (CC₅₀), was used to assess the safety profile of the compounds. A total of 25 compounds had SI values above 50, indicating potential antiplasmodial activity with acceptable cytotoxicity as previously studied [30] (Fig. 3c). The safety profile was also evaluated using in vivo toxicity data, including the LD₅₀ value, maximum tolerated dose (MTD), and prior FDA approval status (Fig. 3d–e). A total of 69 compounds had documented evidence of in vivo toxicity, whereas 48 were previously FDA-approved. Pharmacokinetic profiling was reference-based, prioritizing compounds with C_max values at least twofold greater than the IC₅₀, a half-life (T_1/2) of more than 6 h, and favourable bioavailability in prior studies. This analysis identified 25 priority compounds (Fig. 3f). Structural clustering was performed, and compounds with mechanisms of action relevant to malaria treatment were selected (Fig. 3g–h). Overall, 19 compounds were selected for further investigation (Fig. 4). Of these, 6 compounds had IC₅₀ values under 50 nM, and 3 compounds had prior FDA approval. These compounds target diverse biological pathways, including the proteasome system, folic acid metabolism, metabolism pathway, apicoplast, DNA-dependent RNA polymerase, kinase enzymes, ion channels, hemozoin inhibition, and immune modulation [38–49]. Additionally, eight compounds previously used for oral delivery for other diseases were repurposed for malaria treatment.

Fig. 3.

Selected compounds based on criteria (novelty, antiplasmodial activity, safety, structure, and mechanism). a Distribution of compounds with confirmed antiplasmodial activity (Yes = validated, No = invalidated or less information). The novelty of compounds in malaria was identified by checking resource information; 110 compounds were invalidated or had few evidence with regard to Plasmodium. b Classification of IC₅₀ values. The top 3% of the compounds (256) were classified based on IC₅₀ values and 157 compounds with IC₅₀ < 1 µM were selected. c Selection index (SI) distribution of compounds. The SI value is defined by the IC₅₀ of P. falciparum / CC₅₀ of human cell lines. Twenty-five compounds with a SI ≥ 50, highlighting their potent and selective activity against P. falciparum. d Classification of in vivo toxicity. Low in vivo toxicity is determined based on LD₅₀, maximum tolerated dose (MTD) or treated dose exceeding 20 mg/kg. e Candidate compounds approved by the FDA. FDA approval is evidence for safety (48 compounds). f Classification of pharmacokinetics (PKs). Twenty-nine compounds with longer T_1/2 and C _max > IC₁₀₀ (IC₁₀ ⁓ 2 IC₅₀) were selected because of their potential antiplasmodial drug activity. g Distribution of compound clusters based on structural similarity. The numbers indicate the count of compounds within each cluster. Only compounds showing relative potent antiplasmodial activity in each cluster were included. h Mechanisms of action of the compounds. Compounds with novel mechanisms related to malaria were a priority for enrollment

Fig. 4.

Key biological and physical properties of the candidate compounds for further screening. Based on the combined multifactor characteristics (novelty, safety profile, potent antimalarial activity, pharmacokinetic profile, and mechanism-related support of antimalarial activity) of the 256 prioritized compounds, 19 potential compounds were defined as leads for the next round. The mechanisms mentioned in the figure represent known or proposed modes of action based on previous research. ^*Validated (Mechanisms are supported by experimental evidence in Plasmodium). ^**Inferred (Mechanisms are inferred from their activity in mammalian systems, not validated in Plasmodium)

Hit confirmation via in vitro assays

To further evaluate in vitro antiplasmodial potency, the selected hit compounds were tested against both CQ-sensitive (P. falciparum strains 3D7 and NF54) and CQ-resistant (P. falciparum strains K1 and Dd2), ART-resistant ( P. falciparum strains Dd2-R539T (+) and CamWT-C580Y (+)) via a growth inhibition assay (Fig. 4, Tables 1, 2). Among the hits, several compounds, including ONX-0914, clofoctol, antimony, WS3, suloctidil, methotrexate, tilorone dihydrochloride, algestone, guanabenz acetate, and 7,8-dihydroflavone, exhibited potent inhibitory activity against P. falciparum 3D7 strain with IC₅₀ values under 100 nM. Notably, among these compounds, ONX-0914, antimony, WS3, suloctidil, methotrexate, visomytin, cepharatine, and rifampentine also strongly inhibited CQ-resistant strains (K1 and Dd2) in the low nanomolar range (< 100 nM). Resistance index below 0.8 revealed the potency against CQ- and ART-resistant strains of ONX-0914, sulotidil and visomytin.

Table 1.

In vitro antimalarial activity against a panel of resistant and sensitive strains of Plasmodium

Compounds	IC50 (nM)
	3D7	NF54	K1	Dd2	Dd2-R539T ( +)	CamWT-C580Y ( +)
ONX-0914	22.4 ± 4.9	6.9 ± 0.1	6.6 ± 0.1	5.0 ± 0.1	13.2 ± 2.0	16.3 ± 7.5
Clofoctol	63.6 ± 1.3	16,330 ± 2.0	0.3 ± 0.0	45.3 ± 0.0	5104 ± 1.5	200,000 ± 4.6
Antimony	43.4 ± 3.9	ND	1.3 ± 0.2	46.2 ± 5.5	735.2 ± 178.1	77.6 ± 17.6
WS3	78.8 ± 1.0	ND	3.3 ± 0.2	0.6 ± 0.0	ND	ND
Suloctidil	807.0 ± 25.6	72.6 ± 0.2	5.1 ± 0.0	62.5 ± 0.7	38.8 ± 0.1	137.2 ± 1.2
Rifampicin	ND	900.2 ± 2.1	33.0 ± 0.9	573.1 ± 26.0	563.7 ± 4.1	1148 ± 5.1
Methotrexate	17.8 ± 1.0	51.5 ± 1.4	51.0 ± 0.6	38.5 ± 0.2	32.3 ± 2.1	8.8 ± 0.0
Visomytin	266.9 ± 1.0	109.6 ± 1.1	52.5 ± 0.7	51.0 ± 1.2	16.9 ± 1.0	74.7 ± 3.1
ART	19.5 ± 0.7	11.2 ± 1.1	0.6 ± 0.1	8.9 ± 0.4	9.2 ± 0.4	7.6 ± 0.1
CQ	22.4 ± 1.9	11.7 ± 1.9	91.2 ± 4.9	76.8 ± 2.3	68.7 ± 0.1	45.6 ± 1.5
Cepharanthine	664.4 ± 3.4	188.3 ± 2.4	91.4 ± 4.9	135.6 ± 17.7	85.4 ± 1.1	488.9 ± 4.4
Rifapentine	237.5 ± 30.9	240.1 ± 3.0	98.4 ± 1.1	1979.5 ± 64.3	65.7 ± 1.1	185.8 ± 2.3
Ivermectin	119.8 ± 1.5	4.7 ± 0.1	116.1 ± 10.9	55.5 ± 0.1	21.1 ± 1.0	ND
Selamectin	553.4 ± 15.2	698.4 ± 3.0	310.7 ± 21.4	179.2 ± 4.5	678.1 ± 1.1	ND
Palmatine	2244.4 ± 14.1	619.6 ± 2.2	402.4 ± 17.9	457.5 ± 1.5	17,610 ± 8.5	494.3 ± 4.3
Tilorone	1540 ± 57	647.0 ± 4.1	499.6 ± 0.4	555.6 ± 1.4	32,850 ± 9.6	ND
Algestone	141.9 ± 6.5	ND	621.3 ± 167.1	616.7 ± 52.3	4.5 ± 0.4	ND
Guanabenz acetate	39.8 ± 1.3	ND	930.6 ± 10.8	1309.0 ± 188.1	ND	ND
RN-486	1068.1 ± 2.0	ND	940.0 ± 49.8	611.5 ± 11.7	541.3 ± 2.4	ND
7,8-DHF	98.3 ± 1.4	ND	1010.9 ± 18.5	1727.5 ± 147.8	482.3 ± 1.4	7868.0 ± 4.1
Demethylzeylasteral	977.9 ± 2.3	585.2 ± 2.4	1204.0 ± 21.2	3642.5 ± 16.3	933.0 ± 3.4	3107.0 ± 2.1
NVS-ZP7-4	706.8 ± 5.2	1180.0 ± 1.2	1241.5 ± 31.8	4577.0 ± 264.5	4747.0 ± 11.1	4723.0 ± 3.1
Phentolamine	1787.5 ± 227.0	36,620.0 ± 2.2	188,850.0 ± 12,515.8	7815.5 ± 3054.0	7549 ± 9.2	8232.0 ± 3.1
Duloxetine	233.3 ± 13.0	706.1 ± 4.2	ND	ND	0.5 ± 0.1	1499.0 ± 4.1

The data are presented as mean ± SD with three independent assays, each performed in triplicate. The IC₅₀ values were determined by fitting the data to Log vs. response-variable slope equation in GraphPad Prism. K1 and Dd2 are CQ-resistant strains. Dd2R539T (+) (MRA1255) and CamWT-C580Y (+) (MRA1251) are ART-resistant strains with K13 mutation

Table 2.

Resistance index of compounds against P. falciparum

Compounds	Resistance indexa
	K1	Dd2	Dd2-R539T ( +)	CamWT-C580Y ( +)
ONX-0914	0.3	0.2	0.6	0.7
Clofoctol	0.0	0.7	80.3	3144.7
Antimony	0.0	1.1	16.9	1.8
WS3	0.0	0.0	ND	ND
Suloctidil	0.0	0.1	0.0	0.2
Rifampicin	0.0	0.6	0.6	1.3
Methotrexate	2.9	2.2	1.8	0.5
Visomytin	0.2	0.2	0.1	0.3
ART	0.0	0.5	0.5	0.4
CQ	4.1	3.4	3.1	2.0
Cepharanthine	0.1	0.2	0.1	0.7
Rifapentine	0.4	8.3	0.3	0.8
Ivermectin	1.0	0.5	0.2	ND
Selamectin	0.6	0.3	1.2	ND
Palmatine	0.2	0.2	7.8	0.2
Tilorone	0.3	0.4	21.3	ND
Algestone	4.4	4.3	0.0	ND
Guanabenz acetate	23.4	32.9	ND	ND
RN-486	0.9	0.6	0.5	ND
7.8-DHF	10.3	17.6	4.9	80.0
Demethylzeylasteral	1.2	3.7	1.0	3.2
NVS-ZP7-4	1.8	6.5	6.7	6.7
Phentolamine	105.7	4.4	4.2	4.6
Duloxetine	ND	ND	0.0	6.4

^aResistance index (RI) is the ratio of IC₅₀ for the resistant versus the sensitive strain (3D7 or NF54). The IC₅₀ values were determined by fitting the data to Log vs. response-variable slope equation in Graph Pad Prism. K1 and Dd2 are CQ-resistant strains. Dd2-R539T (+) (MRA1255) and CamWT-C580Y (+) (MRA1251) are ART-resistant strains with K13 mutation. ND (not defined)

The safety profile of these candidates was evaluated with respect to their haemolytic activity in human erythrocytes and human cell lines (Fig. 5). Although selamectin and suloctidil presented higher haemolysis rates than CQ (> 18% at 50 µM), the remaining compounds presented lower haemolysis rates (under 15%), suggesting selective parasite toxicity without significant host cell damage (Fig. 5a). The compounds that exhibited promising inhibition against Plasmodium were further assessed for in vitro cytotoxicity in HEK293T cells (Fig. 5b). The potential compounds (rifapentine, methotrexate, ONX-0914 and antimony) presented CC₅₀ values greater than 100 µM and IC₅₀ values in nanomolar (IC₅₀ < 55 nM), indicating low cytotoxicity to human cells and highlighting their selective activity against P. falciparum.

Fig. 5.

Haemolysis and cytotoxicity of the test compounds in human cells. a Haemolysis of the compounds at a single concentration (50 µM). Twenty-one compounds demonstrated a haemolysis rate lower than 20% at 50 µM. b Cytotoxicity (CC₅₀) and antiplasmodial activity (IC₅₀) of promising compounds against human cells (HEK293T) and P. falciparum. IC₅₀ values (µM) and CC₅₀ values (µM) are plotted on separate log-transformed axes to highlight therapeutic windows. Red dashed line indicated selective index (SI) = 100 (CC₅₀ = 100 × IC₅₀). CC₅₀ of Methotrexate and Rifapentine was > 300 µM. Among them, rifapentine, methotrexate, antimony, and ONX-0914 exhibited low cytotoxicity to HEK293T cells (CC₅₀ > 100 µM) and selective antiplasmodial activity against P. falciparum (IC₅₀ < 55 nM), yielding SI values > 100. The data are shown as the mean ± standard deviation from three independent assays. Error bars indicate the standard deviation (SD)

In vivo antimalarial activity in Plasmodium berghei ANKA-infected mice

Among the 19 selected compounds, 16 hit compounds demonstrated significant in vivo antiplasmodial efficacy by suppressing P. berghei infection in a mouse model (Fig. 6). To evaluate their therapeutic potential, an initial oral treatment was conducted with a single dose (20–50 mg/kg) to identify compounds with potent oral antimalarial efficacy. This was followed by a single dose of intraperitoneal (IP) (20 mg/kg) for candidates with low oral bioavailability (Fig. 6a). In this assessment, chloroquine (CQ) was used as a positive control, whereas rifampicin was compared with rifapentine due to its structural and functional similarities. The standard dosing regimen for all hits was the administration of a single dose of 50 mg/kg, except for CQ (20 mg/kg) and antimony (25 mg/kg).

Fig. 6.

In vivo assay via oral (PO) and intraperitoneal injection (IP) delivery. a Flowchart of the in vivo assay. b Microscopy image of a 10% Giemsa-stained sample at the 6th day postinfection after treatment with 20 mg/kg via IP delivery. Bars = 10 µm. c Mean survival day and survival rate after oral treatment at 10 days postinfection. d Suppression rate after four days of oral treatment. Suppression after 4 d was calculated via the following equation: (% suppression = 100–100 × (parasitaemia of intervention/parasitaemia positive). e Mean survival day and survival rate via IP delivery after 10 days postinfection. f Survival of mice in the group treated via IP delivery. g Parasitaemia of the groups treated via IP delivery. At least 1000 RBCs were observed via microscopy to determine parasitaemia. h Suppression rate after four days of IP treatment. i E₅₀ of ONX-0914 and antimony via the IP delivery routes. Data are shown as the mean ± standard deviation. Error bars indicate the standard deviation (SD) from the mean. The dashed line refers to the group with greater suppression. Each group included 5 mice. Bars = 10 µm

Oral treatment with 5 compounds (rifapentine, 7,8-DHF, WS3, methotrexate, ONX-0914, and algestone) resulted in a 100% survival rate at 10 days postinfection. Additionally, antimony and palmatine resulted in prolonged mean survival days exceeding 9.5, with survival rates above 80% (Fig. 6c). The mean survival days and survival rate were strongly correlated with parasite suppression efficiency. ONX-0914, methotrexate (MTX), WS3, rifapentine, duloxetine, and visomytin all exhibited > 65% suppression of P. berhei (Fig. 6d). Specifically, ONX-0914 demonstrated the highest suppression rate (95.93 ± 0.75%), followed by methotrexate (81.38 ± 0.96%), WS3 (77.13 ± 0.56%), rifapentine (70.46 ± 25.28%), and duloxetine (69.38 ± 21.21%). These findings suggest that ONX-0914 is particularly promising for oral antimalarial treatment.

Owing to the low oral bioavailability of some compounds, in vivo antimalarial activity was further assessed via IP administration (Fig. 6b–e–f–g–h). Notably, ONX-0914, antimony, and rifapentine resulted in 100% survival at 10 days postinfection, with strong suppression of P. berghei (> 90%) after four days of treatment (Fig. 6e–f). By 8 days postinfection, these compounds also maintained parasitaemia levels below 5% (Fig. 6b–g–h). Further pharmacodynamic assessment revealed that ONX-0914 and antimony had the strongest in vivo efficacy, with EC₅₀ values of 6.52 mg/kg and 10.02 mg/kg, respectively (Fig. 6i). Notably, the EC₅₀ in this study was calculated based on parasite suppression starting 48 h postinfection, and ONX-0914 may have an even lower EC₅₀ if treatment is initiated within 2–4 h postinfection, as suggested in previous studies [28].

Parasitaemia suppression by the selected hits significantly improved clinical symptoms and organ pathology. Notably, weight loss was not observed in mice treated with ONX-0914, rifapentine, MTX, duloxetine, or antimony after 14 days postinfection (p < 0.01) (Fig. 7a). Additionally, these compounds effectively reduced malaria-associated symptoms, including quiet, ruffled fur, and anaemia (Fig. 7b). In terms of organ pathology, hepatomegaly and splenomegaly were common in the control group but were significantly reduced in the rifapentine, MTX, duloxetine, and antimony treatment groups (p < 0.01). Compared with infected control mice, mice treated with ONX-0914, MTX, antimony, and rifapentine presented smaller livers and spleens (p < 0.01) with no visible discolouration or damage (Fig. 7c–d–e).

Fig. 7.

Effects of ONX-0914 on the clinical symptoms, weights, and organs of treated mice. a Weights of the mice at 10 DPI. Compared with no treatment, seven compounds improved body weight. b Clinical symptoms. The normalized expression of symptoms was calculated as the percentage of mice with symptoms relative to the total number of mice in each group. c Characteristics (colour and size) of organs (brain, heart, lung, liver, kidney, and spleen) after sacrificing the mice at 10 days. Organs were collected from 3 mice per group. d Size of the liver. e Size of the spleen. The size of the organ was defined as height × width (mm). Bar = 5 mm. Compared with untreated mice, ONX-0914-, antimony-, methotrexate-, rifapentine-, and WS3-treated mice presented smaller livers and spleens (*p < 0.01, **p < 0.001, and ***p < 0.0001)

Histological examination confirmed reduced organ lesions in ONX-0914-treated P. berghei (Table 3, Supplementary Fig. 2). Malaria pigment deposition, a hallmark of P. berghei infection, was abundant in the blood vessels of the brain, heart, lung, liver, spleen, and kidneys of control mice. However, pigment accumulation was absent in ONX-0914- and CQ-treated mice and significantly lower in rifapentine-treated hearts and MTX-treated kidneys.

Table 3.

Histopathological findings observed in the organs of PbANKA-infected mice treated with antimalarial drug compounds

Organ and Histopathology/Test materials	Negative	Positive	CQ	ONX-0914	Antimony	Duloxetine	7,8-DHF	Algeston	Rifampicin	Clofoctol	Rifapentine	WS3	Visomytin	Methotrexate
Brain	NSL		NSL	NSL
Stasis in capillaries		P			P	P	P	P	P	P	P	P	P	P
Bleeding		P			P	P	P		P	P			P	P
Pigments in blood		4 +			2 +	3 +	3 +	4 +	3 +	3 +	4 +	3 +	3 +	2 +
Precursor cells in blood					3 +	2 +		2 +	2 +		3 +	2 +	1 +	1 +
Heart	NSL
Epicarditis, Rt. Ventricle						2 +						ND	ND	ND
Mineralization, epicardium		2 +	2 +	2 +				2 +			1 +	ND	ND	ND
Pigments in blood		4 +			3 +	2 +	2 +	3 +	2 +	2 +	3 +	ND	ND	ND
Lung	NSL		NSL	NSL
Pigments in alveolar walls		4 +			2 +	2 +	2 +	2 +	1 +	2 +	3 +	2 +	2 +	2 +
Pigments in blood		4 +			2 +	1 +	1 +	3 +	2 +	2 +	3 +	2 +	1 +	1 +
Precursor cells in blood		4 +			2 +	2 +	1 +	2 +	2 +	3 +	2 +	1 +	1 +	1 +
Thrombosis										2 +		2 +	2 +
Hemopoietic precursor cells, perivascular, multifocal		1 +								1 +	1 +			2 +
Liver	NSL		NSL	NSL
Pigments, Kupffer cells		4 +			3 +	3 +	3 +	4 +	3 +	3 +	4 +	4 +	4 +	3 +
Hematopoietic progenitor cells, sinusoids		4 +			3 +	3 +	3 +	3 +	3 +	3 +	3 +	3 +	3 +	3 +
Hematopoietic cell aggregates, multifocal		4 +			3 +	2 +	3 +	3 +	2 +	3 +	4 +	3 +	3 +	3 +
Pigments, blood		4 +			3 +	2 +	1 +	3 +	3 +	2 +	4 +	3 +	3 +	3 +
Spleen	NSL
Pigments, red pulps		4 +			3 +	3 +	3 +	3 +	2 +	2 +	3 +	2 +	2 +	2 +
Pigments, sinusoids		4 +			3 +	3 +		3 +	2 +	2 +	3 +	3 +	3 +	3 +
Apoptotic cells, white pulp		4 +	2 +	1 +	1 +	2 +	3 +		2 +	1 +	2 +	3 +	3 +	3 +
Hemorrhage & fibrins, multifocal						3 +	3 +	4 +	3 +	2 +
Extramedullary hematopoiesis		4 +			4 +				3 +	3 +	4 +	4 +	4 +	4 +
Kidney	NSL		NSL	NSL
Degeneration/necrosis, tubules		3 +			3 +	3 +	2 +	3 +	3 +	3 +	2 +	1 +	1 +	1 +
Pigments in blood		4 +			3 +	1 +	1 +	2 +	1 +	1 +	4 +	1 +	1 +	1 +
Pigments in glomerular tufts					1 +		1 +	1 +	1 +	1 +	1 +	1 +	1 +	1 +

NSL, No specific lesion; ND, Not examined; p, present

Criteria for evaluation: 1 + , minimal; 2 + , mild; 3 + , moderate; 4 + , severe

Antimony, Methotrexate reduced lesions caused by Plasmodium in collected organs compared to infected mice without treatment (positive). No specific lesions related to Plasmodium or toxicity to organs were detected in mice that received ONX-0914 treatment

Discussion

The increasing prevalence of antimalarial drug resistance poses a major challenge to malaria eradication, highlighting the urgent need for novel therapeutic agents [12, 14, 37, 50, 51]. However, drug discovery is a complex, multistage process that is resource-intensive, time-consuming, and costly. In this study, an image-based HTS platform with an automated image analysis algorithm combined with meta-analysis-driven selection was developed to increase the efficiency of antimalarial drug discovery.

HTS is a powerful drug discovery technique that integrates robotics, liquid handling systems, data acquisition, and computational analysis to rapidly screen large compound libraries [52]. Compared with traditional assays such as the [³H]-hypoxanthine or lactate dehydrogenase enzyme assay, the SYBR Green I assay is used for HTS and is highly sensitive, fast, cost-effective and safe. Nevertheless, the main disadvantage of this method is the overestimation of parasite viability due to the presence of DNA from dead parasites, interference from autofluorescent compounds, and the inability to assess parasite morphology [22, 53, 54]. Therefore, image-based HTS was developed and offers significant advantages over traditional fluorescence plate reader assays, particularly in assessing drug efficacy against drug-resistant P. falciparum strains and other protozoan parasites, including Leishmania donovani and Trypanosoma brucei as well as bacterial and viral infections [15, 55–58]. This study demonstrated that the SYBR Green I assay, used for both primary screening and confirmation screening, saves time, with a Z’-factor of 0.55 indicating the validity of the method. The image-based analysis was repeated with 50 promising hits to increase the accuracy of parasite detection and antimalarial drug sensitivity assays, outperforming fluorescence-based approaches. However, a major limitation of the image-based method is the difficulty in classifying late-stage (trophozoite and schizont) parasites in cases of multiple infections, which requires high-magnification imaging and advanced segmentation algorithms [14]. In the future, addressing the cost constraints of automated instrumentation and improving the efficiency of HTS relative to target-based screening approaches will be critical for optimizing its utility in antimalarial drug discovery.

Hit identification in drug discovery remains a major bottleneck owing to low success rates and time-consuming validation [6, 16–18]. To address this problem, various computational approaches, such as virtual screening, chemical diversity libraries, target deconvolution, and toxicology assessments, have been used to streamline hit selection [10, 59, 60]. However, these strategies alone are insufficient, as additional parameters, including pharmacokinetics, safety profiles, and antimalarial potency, must be considered. Meta-analysis screening was integrated using multiple selection criteria (novelty, antimalarial efficacy, safety, and FDA approval) based on WHO and MMV guidelines for antimalarial agents [25–29]. This approach enabled the prioritization of 19 hits from 256 candidates, ensuring a balanced profile of efficacy and safety for further evaluation. Notably, a limitation of meta-analysis screening is the lack of comprehensive compound information, which can result in the exclusion of promising candidates due to incomplete datasets. Addressing this challenge requires further experimental validation of compounds with limited prior data and the development of artificial intelligence-driven models utilizing big data to improve predictive accuracy in drug discovery.

To address the challenge of drug resistance, the potent selected hits were further evaluated against CQ- and ART-resistant P. falciparum strains. Among the 19 hits, ONX-0914, suloctidil, and MTX significantly inhibited both the CQ- and ART-resistant strains (IC₅₀ < 500 nM). One of the proteasome inhibitors, ONX-0914, exhibited similar activity to previously studied proteasome-targeting compounds, reinforcing its potential as a novel antimalarial agent [61, 62]. Similarly, MTX, a known antifolate drug, has demonstrated strong activity against CQ-resistant P. falciparum, which is consistent with previous reports [63]. In addition, suloctidil, a vasodilator that inhibits the sodium‒potassium pump, has shown promising activity against multidrug-resistant strains, which aligns with recent findings on its potential as an antimalarial agent [64]. Further investigations into their mechanisms of action, efficacy in in vivo drug-resistant models, and synergistic potential will be essential for advancing these compounds towards clinical development.

A key limitation in antimalarial drug development is the poor translation of in vitro potency to in vivo efficacy, especially for orally administered compounds. Previous studies have focused primarily on the intravenous administration of ONX-0914 in cerebral malaria models, as recommended by the manufacturer [65]. MTX has a well-established safety profile for short-term, low-dose regimens and has been proposed as a potential combination therapy for rheumatoid arthritis [39, 63]. Likewise, MTX has exhibited strong activity against drug-resistant Plasmodium vivax and P. falciparum [38]. Rifapentine, a novel antituberculosis drug, exhibited greater antimalarial activity than rifampicin, supporting its potential for malaria treatment. This work provides strong evidence for the oral efficacy of ONX-0914, MTX, and rifapentine, as demonstrated by prolonged mean survival days and survival rate, parasite suppression, clinical symptom improvement, and histopathology analysis. In our study, suppression and EC₅₀ were determined starting at 48 h postinfection, but growth inhibition could be even greater if treatment was initiated earlier (2–4 h postinfection), as reported in previous studies [28]. Although MTX, WS3, rifapentine, and duloxetine demonstrated high bioavailability in other disease models, suloctidil and antimony exhibited poor oral bioavailability, which may be attributed to pharmacokinetic limitations [41]. In contrast to its modest oral activity, antimony, an antileishmanial agent, demonstrated high in vivo efficacy via IP administration in this study. Optimizing dose regimens, administration timing, and drug-like properties will be crucial for improving oral bioavailability and therapeutic outcomes in future research.

Conclusions

HTS is a powerful and widely applied technique in drug discovery, including malaria research. In this study, automated HTS-based SYBR Green I and image assays were utilized to facilitate high-content analysis, with the aim of improving the efficiency and consistency of antiplasmodial activity assessment. Integration of automated HTS with meta-analysis established a structured approach to support the identification of potential novel antimalarial compounds. Using this platform, 19 potential compounds with activity against blood-stage P. falciparum parasites were identified from an in-house drug library, from which three promising candidates (ONX-0914, antimony, and MTX) were prioritized for further investigation. These compounds will be evaluated insight their safety profiles, synergistic potential in combination with artemisinin, pharmacokinetic properties, structural optimization, and dosing regimen to enhance therapeutic efficacy and minimize toxicity. However, several limitations of this avenue should be acknowledged. The initial screening focuses on blood-stage parasites, which does not fully capture all Plasmodium life stages. Additionally, some compounds lacked comprehensive pharmacological and toxicological profile, and this approach, while efficient, may not fully predict in vivo efficacy and safety. Besides, the high cost of HTS equipment, along with technical challenges such as assay variability, compound solubility and the need for extensive training, also pose constraints. Despite these hurdles, the integration of HTS and meta-analysis provides a systematic approach to support early-phase antimalarial drug discovery.

Abbreviations

ART

Artemisinin

Abs

Absorbance

BSS

Bliss synergy score

Conc.

Concentration

CQ

Chloroquine

DRC

Dose-response curve

DMSO

Dimethyl sulfoxide

DRM

Dose-response matrix

FAC

Fluorescence-activated cell

FDA

Food and Drug Administration

FIC

Fractional inhibitory concentration

HCS

High-content screening

HEPES

Hydroxyethyl piperazine ethane sulfonic acid

HTS

High-throughput screening

IP

Intraperitoneal

IC₅₀

Half-maximal inhibitory concentration

iRBC

Infected red blood cell

LD₅₀

Lethal dose 50

NSL

No specific lesion

P

Present

PBS

Phosphate-buffered saline

PO

Per oral

RBC

Red blood cell

RT

Room temperature

SI

Select index

TES

Therapeutic efficacy studies

uRBC

Uninfected red blood cell

Author contributions

NVT, ETH, and JHN were involved in conceiving and designing the experiments and writing the manuscript. NVT, SWN, JHP, TKN, NST, TTHC, CV, BS, WSP, WJC, JWN, SHN, and JHH conducted the laboratory work and data analysis. SHN provided reagents/materials. ETH confirmed the authenticity of all the data. All the authors approved the content and agreed to be accountable for the content of the work.

Funding

This study was supported by a National Research Foundation of Korea (NRF) grant (NRF-2021R1A2C2008235 & RS-2025-00517986 to E.T.H.) funded by the Korean government (MSIP), and by a grant issued for the Basic Science Research Program (NRF-R1A4A1031574 to E.T.H.), funded by the Ministry of Science, ICT and Future Planning.

Availability of data and materials

The datasets analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.