Microbiome Indoles Dock at the TYR61–GLU67 Hotspot of Giardia lamblia FBPA: Evidence from Docking, Rescoring, and Contact Mapping

Abstract

Giardiasis, caused by the protozoan parasite Giardia lamblia, remains a prevalent intestinal infection worldwide and a growing concern due to increasing resistance to nitroimidazole drugs. This study proposes an alternative therapeutic strategy by targeting fructose-1,6-bisphosphate aldolase (FBPA), a key glycolytic enzyme of the parasite, through structure-based virtual screening. A curated library of microbiome-derived metabolites was computationally evaluated and compared with clinically used antigiardial drugs. Several indole-based compounds exhibited favorable binding affinities and stable interactions within the catalytic pocket of FBPA. These findings suggest that microbiome metabolites could serve as promising scaffolds for the rational design of new antiparasitic agents. Overall, the study highlights the potential of integrating metabolic and computational approaches to identify next-generation therapeutics against giardiasis.

1. Introduction

Giardiasis is one of the most widespread protozoan intestinal infections, affecting both developed and developing countries. The World Health Organization classifies Giardia lamblia (syn. G. intestinalis, G. duodenalis) as a neglected intestinal protozoan that causes significant gastrointestinal morbidity, particularly in children and immunocompromised individuals. The infection is transmitted via the fecal–oral route through contaminated water or food, leading to acute or chronic diarrheal disease, nutrient malabsorption, and post-infectious irritable bowel syndrome. The global prevalence of giardiasis is estimated at over 200 million symptomatic cases per year, representing a persistent public health burden in low-resource settings [1,2,3].

Current frontline therapy relies on nitroimidazoles such as metronidazole and tinidazole, with albendazole as an alternative. However, their variable efficacy, adverse reactions, and increasing clinical resistance underscore the urgent need for new chemotypes [1,2,3]. Nitroimidazoles act primarily via redox-mediated DNA and protein damage, a mechanism that can also lead to host toxicity. Thus, exploring alternative metabolic targets in G. lamblia is a rational approach for drug discovery.

Among potential enzymatic targets, fructose-1,6-bisphosphate aldolase (FBPA) has emerged as a key node within the parasite’s glycolytic pathway. This enzyme catalyzes the reversible cleavage of fructose-1,6-bisphosphate into dihydroxyacetone phosphate and glyceraldehyde-3-phosphate, providing a central metabolic link essential for energy homeostasis. Giardia FBPA belongs to the class II, Zn²⁺-dependent aldolases, structurally and mechanistically distinct from class I aldolases found in humans. This divergence confers selectivity, making FBPA an attractive antiparasitic target. The structure of G. lamblia FBPA has been resolved at 2.30 Å (PDB ID: 3OHI), co-crystallized with a synthetic inhibitor, providing an ideal framework for structure-based drug design [4].

In recent years, microbiome-derived metabolites have gained attention as novel molecular scaffolds for therapeutic discovery. Compounds such as indoles and short-chain fatty acids (SCFAs), which are bacterial products of amino acid and carbohydrate metabolism, influence epithelial barrier function, immune regulation, and host–pathogen interactions [5,6,7,8]. Indoles, derived from tryptophan metabolism, possess aromatic surfaces capable of hydrogen bonding and π–π stacking, making them chemically suitable for docking-based studies and biologically relevant as ligands of the aryl hydrocarbon receptor (AhR) and pregnane X receptor (PXR). Conversely, SCFAs are small aliphatic molecules that modulate host physiology through G-protein-coupled receptors (GPCRs) and histone deacetylase (HDAC) inhibition. The exploration of such metabolites as antiparasitic candidates bridges microbial ecology and medicinal chemistry, suggesting dual-action molecules with both antiparasitic and host-beneficial potential.

Structure-based molecular docking offers a rapid and cost-effective strategy to prioritize candidate ligands. Nevertheless, its predictive power can be limited by intrinsic biases related to molecular size, lipophilicity, and scoring-function variability [9,10,11,12]. To mitigate these biases, consensus scoring approaches and physicochemical normalization metrics such as ligand efficiency (LE) and lipophilic ligand efficiency (LLE) have become standard tools in rational drug design [9,10,11,12]. Among available docking programs, AutoDock Vina and its derivative Vinardo have been extensively validated for pose prediction accuracy and virtual-screening robustness [13,14]. Based on this rationale, the present study explores microbiome-derived metabolites—particularly indole derivatives—as potential inhibitors of G. lamblia FBPA. We hypothesize that these compounds interact within a conserved aromatic–polar hotspot centered on TYR61–GLU67 and display favorable LE/LLE profiles relative to more lipophilic clinical drugs. This work aims to identify promising indole-based scaffolds through structure-based virtual screening, providing a rational and reproducible computational framework for future biochemical and cellular validation.

2. Materials and Methods

2.1. Computational Environment and Code Availability

All analyses were performed on Google Colab (August 2025), running a Linux x86_64 environment with Python 3.10. The following packages and tools were used: RDKit 2023.09, Meeko 0.5.x, AutoDock Vina 1.2.x (Vinardo-supported), Open Babel 3.1.1, and Biopython 1.81. Vinardo scoring was accessed through the Vina 1.2.x engine using the flag –scoring vinardo [13,14,15].

All analyses and scripts are available in the accompanying Google Colab notebooks, which reproduce the complete computational workflow from ligand preparation to post-processing. These notebooks follow FAIR data principles (Findable, Accessible, Interoperable, and Reusable) and are released under a Creative Commons license [16,17].

2.2. Target Structure (FBPA)

The target enzyme was the Giardia lamblia fructose-1,6-bisphosphate aldolase (FBPA), a class II Zn²⁺-dependent enzyme (PDB ID: 3OHI). This structure was selected after evaluating all available FBPA entries in the Protein Data Bank (PDB) according to criteria of completeness, biological relevance, and crystallographic quality. The 3OHI entry, solved at 2.30 Å resolution (R_free = 0.228), represents the only experimentally determined FBPA structure from G. lamblia. It is co-crystallized with a competitive inhibitor (3-hydroxy-2-pyridone) bound within the catalytic pocket, providing an experimentally validated active-site geometry that includes the Zn²⁺ cofactor and key residues (TYR61, GLU67, HIS109, ASP131). The structure maintains full chain integrity and correct metal coordination, making it suitable for structure-based virtual screening. No higher-resolution FBPA structures from Giardia are currently available, and 3OHI has been referenced in prior studies as a reliable docking template [4].

Receptor Preparation

The receptor file (PDB/mmCIF) was standardized following AutoDock conventions: non-essential water molecules and heteroatoms were removed, Zn²⁺ ions retained, and biological chains selected. Alternate atom locations were resolved, hydrogens added at physiological pH (7.4), and AutoDock atom types and Gasteiger charges assigned. The receptor was exported in PDBQT format using Meeko (mk_prepare_receptor.py) or AutoDockTools scripts. Structure manipulation and I/O operations were handled using Bio.PDB from Biopython [18].

2.3. Ligand Panel and Preparation

A total of 15 ligands were selected, including microbiome-derived indoles (indole, indole-3-acetic acid, tryptamine), short-chain fatty acids (SCFAs) (acetic, propionic, butyric, isobutyric, valeric, isovaleric), other microbial metabolites (lactic acid, p-cresol), and clinical controls (nitazoxanide, albendazole, metronidazole, tinidazole).

Ligands were standardized and embedded in 3D using RDKit (ETKDGv3) and energy-minimized via MMFF94 (Universal Force Field fallback). Final PDBQT files were generated using Meeko (mk_prepare_ligand.py) with Gasteiger charges and AutoDock4 atom types.

Ligand protonation states were assigned at physiological pH (7.4 ± 0.1) using RDKit’s Chem.AddHs combined with Open Babel pKa estimation to determine the dominant microstate. Each structure was visually inspected for chemical plausibility and consistency. This protocol follows established AutoDock-based ligand preparation practices [13,19].

2.4. Docking Box Definition

The grid box was centered on the centroid of the co-crystallized ligand in PDB 3OHI and encompassed all residues within 10 Å of the catalytic Zn²⁺ ion, including TYR61, GLU67, HIS109, and ASP131. The box dimensions were adjusted to 22–26 Å per axis to include the catalytic site and maintain native geometry [20,21].

2.5. Docking Protocol

Docking was carried out using AutoDock Vina 1.2.x, with multithreading enabled [13,15].

▪

Exhaustiveness: 16;

▪

Number of poses: 20;

▪

Energy range: 3 kcal·mol⁻¹;

▪

Replicates: 3 random seeds (− seed 1, 2, 3);

▪

Output: best binding free energy (ΔG) and all docking poses.

Each ligand was docked in triplicate to assess stochastic reproducibility. Mean and standard deviation (SD) of ΔG were calculated. Docking results were considered stable when ΔΔG < 0.3 kcal·mol⁻¹, following best practices for reproducible docking workflows [13,15].

2.6. Vinardo Rescoring and Internal Validation

A second pass was performed using the Vinardo scoring function, maintaining identical grid parameters and random seeds to enable consensus scoring [14,15].

For validation, the co-crystallized ligand was re-docked into FBPA, and the root-mean-square deviation (RMSD) between predicted and experimental poses was calculated. A RMSD ≤ 2.0 Å was considered acceptable. RMSD values were computed in RDKit after substructure alignment.

2.7. Structural/Contact Analysis

Residue–ligand contacts were defined as heavy-atom pairs within 4.0 Å. For each ligand, contacts were aggregated by residue ID and compiled into ligand × residue matrices. Frequency heatmaps and contact distributions were generated using NumPy (v1.26) and Matplotlib (v3.8), with visualization via Py3Dmol (v2.0) and NGLView (v3.0). These analyses identified key residues mediating hydrogen bonding and hydrophobic interactions within the catalytic pocket.

2.8. Physicochemical and Quality Metrics

Ligand physicochemical properties were computed in RDKit, including molecular weight (Mw), LogP (Crippen), hydrogen-bond donors (HBD), hydrogen-bond acceptors (HBA), and topological polar surface area (TPSA).

Ligand efficiency (LE) was calculated as −ΔG/N_heavy, and lipophilic ligand efficiency (LLE) as pK_d − LogP, with pK_d = −ΔG/(R·T·ln 10), R = 1.987 × 10⁻³ kcal·mol⁻¹·K⁻¹, T = 298.15 K. The rationale follows established best practices in medicinal chemistry [9,10,11,12].

Statistical Analysis

Group differences were assessed using the Kruskal–Wallis test, followed by Dunn’s or Mann–Whitney U post hoc tests with false discovery rate (FDR) correction when appropriate. Correlations (ΔG vs. Mw) were evaluated using Pearson and Spearman coefficients. Statistical procedures were implemented in SciPy and Statsmodels.

2.9. Reproducibility Checklist

All seeds, parameters, CSV output files (per-replicate ΔG, consensus scores, contact matrices), and figures (heatmaps, 3D models, and statistical plots) are archived within the public Google Colab notebook. Each notebook can be executed with a single click to reproduce the workflow, ensuring full transparency and reproducibility according to FAIR principles [16,17].

2.10. Ethics Statement

This study relied exclusively on previously published, de-identified data from open repositories, including protein structures from the Protein Data Bank (PDB) (e.g., PDB 3OHI) and compound information (SMILES, physicochemical descriptors) from public resources such as PubChem and ChEMBL. No new biological materials were collected, and no identifiable human or animal data were used. Under the policy of the Universidad Nacional Jorge Basadre Grohmann Institutional Review Board, research conducted solely on de-identified, publicly available datasets is exempt from further ethics review. Accordingly, IRB approval and informed consent were not required for this in silico study.

3. Results

Affinity window and recovery of controls.

Docking to FBPA (3OHI) produced a ~2.29 kcal/mol affinity window (from −5.20 to −2.91 kcal/mol) (Figure 1). The positive controls concentrated at the top of the ranking—nitazoxanide (−5.20) and albendazole (−5.17)—which supports the ability of the protocol to prioritize compounds with higher affinity for this target. Metronidazole and tinidazole showed intermediate values (−4.36 and −4.04 kcal/mol), consistent with expectations for this scoring function (

Table 1. Docking affinities (ΔG, kcal/mol) of microbiome-derived metabolites and positive controls against Giardia lamblia fructose-1,6-bisphosphate aldolase (PDB: 3OHI).

No.	Ligand	Group	Target	Affinity (kcal/mol)
0	Nitazoxanide	Control	3OHI	−5.199
1	Albendazole	Control	3OHI	−5.165
2	Indole-3-acetic acid (IAA)	Indole	3OHI	−5.147
3	Tryptamine	Indole	3OHI	−4.930
4	Metronidazole	Control	3OHI	−4.363
5	Indole	Indole	3OHI	−4.278
6	Isovaleric acid	SCFA	3OHI	−4.209
7	Valeric acid	SCFA	3OHI	−4.111
8	Tinidazole	Control	3OHI	−4.035
9	p-Cresol	Aromatic	3OHI	−4.021
10	Butyric acid	SCFA	3OHI	−3.620
11	Isobutyric acid	SCFA	3OHI	−3.533
12	Lactic acid	Microbiome	3OHI	−3.501
13	Propionic acid	SCFA	3OHI	−3.225
14	Acetic acid	SCFA	3OHI	−2.909

).

Microbiome metabolites among the best-scoring compounds.

Among microbiome-related molecules, indole-3-acetic acid (IAA) achieved −5.15 kcal/mol, essentially indistinguishable (≤0.05 kcal/mol) from the best control within the typical precision of Vina. Tryptamine (−4.93) and indole (−4.28) followed. In contrast, the short-chain fatty acids (SCFAs) clustered at weaker affinities (≈ −4.21 to −2.91), with isovalerate and valerate as the top SCFAs (−4.21 and −4.11, respectively).

Chemical pattern.

Overall, indoles outperformed SCFAs by approximately 0.9–2.2 kcal/mol across the series, suggesting that aromatic scaffolds capable of π-stacking and/or hydrogen bonding are better accommodated at the active site than short aliphatic chains.

Bars show the best Vina score (ΔG, kcal/mol) per ligand (more negative = stronger predicted binding). The panel comprises microbiome metabolites (SCFAs, indoles and related aromatics) and clinical antigiardial controls. The affinity window spans roughly −5.20 to −2.91 kcal/mol (n = 15). Positive controls (nitazoxanide, albendazole) rank at the top, while indole-3-acetic acid (IAA) and tryptamine are the best microbiome hits, approaching the controls. SCFAs cluster at weaker affinities toward the right of the plot.

Docking to FBPA (3OHI) produced a ~2.3 kcal/mol spread in predicted binding free energies, with nitazoxanide (−5.20) and albendazole (−5.17) recovered at the top of the ranking, validating the protocol. Among microbiome-related molecules, IAA (−5.15) and tryptamine (−4.93) were the strongest, followed by indole (−4.28). In contrast, SCFAs—including isovalerate and valerate—grouped at weaker values (≈ −4.2 to −2.9 kcal/mol), consistent with the need for aromatic interactions at the FBPA active site (Figure 2).

Boxplots summarize the best AutoDock Vina score (ΔG, kcal/mol; more negative = stronger predicted affinity) per chemical class. Sample sizes: Controls (n = 4), Indole (n = 3), SCFA (n = 6), Aromatic (n = 1), Microbiome (lactic acid) (n = 1). Medians: Indole −4.93, Control −4.76, SCFA −3.58, Aromatic −4.02, Microbiome −3.50 kcal/mol. Controls and indoles concentrate at more negative values than SCFAs, indicating a clear class-dependent trend. (Exact statistics are provided in the group-comparison table.)

Non-parametric statistical methods (Kruskal–Wallis and Spearman correlation) were employed to handle the small sample size and non-normal data distribution. These tests provided exploratory insight into the relationships between docking scores, molecular size, and efficiency metrics.

Grouping ligands by chemistry reveals a pronounced separation in predicted affinities (

Table 2. Per-ligand docking and developability metrics for microbiome metabolites and positive controls against Giardia lamblia FBPA (PDB: 3OHI).

No.	Ligand	Group	Target	Affinity (kcal/mol)	pKd (Approx.)	LE (kcal/mol per Heavy Atom)	LLE (approx.)	MolWt (g/mol)	Heavy Atoms	HBD	HBA	TPSA (Å2)	RotB	LogP
0	Nitazoxanide	Control	3OHI	−5.199	3.811	0.2266	1.1549	336.0625	23	3	6	113.44	6	2.6563
1	Albendazole	Control	3OHI	−5.165	3.786	0.2247	−0.5200	329.0931	24	2	4	65.15	5	4.3060
2	Indole-3-acetic acid (IAA)	Indole	3OHI	−5.147	3.773	0.3959	1.9782	175.0633	13	2	2	53.09	2	1.7950
3	Tryptamine	Indole	3OHI	−4.930	3.611	0.4108	1.9449	160.1000	12	2	1	41.81	3	1.6691
4	Metronidazole	Control	3OHI	−4.363	3.193	0.3564	3.7092	173.0437	15	2	5	101.42	5	−0.5180
5	Indole	Indole	3OHI	−4.278	3.138	0.4753	0.9628	117.875	9	1	1	41.79	1	2.1400
6	Isovaleric acid	SCFA	3OHI	−4.209	3.085	0.5813	1.9684	102.0680	7	1	2	37.30	4	1.1171
7	Valeric acid	SCFA	3OHI	−4.111	3.016	0.5162	1.7525	102.0680	7	1	2	37.30	4	1.1749
8	Tinidazole	Control	3OHI	−4.035	2.959	0.2690	1.1525	249.0242	18	2	5	81.19	8	0.8566
9	p-Cresol	Aromatic	3OHI	−4.021	2.947	0.5026	2.4375	108.0575	8	1	1	20.23	2	2.2192
10	Butyric acid	SCFA	3OHI	−3.620	2.654	0.6033	1.7827	88.0504	6	1	2	37.30	3	0.8700
11	Isobutyric acid	SCFA	3OHI	−3.533	2.589	0.5889	1.8629	88.0504	6	1	2	37.30	3	0.7270
12	Lactic acid	Microbiome	3OHI	−3.501	2.567	0.5855	3.1742	90.0316	6	2	3	57.53	3	−0.5840
13	Propionic acid	SCFA	3OHI	−3.225	2.366	0.6460	1.7665	74.0368	5	1	2	37.30	3	0.8900
14	Acetic acid	SCFA	3OHI	−2.909	2.133	0.7273	2.0416	60.0211	4	1	2	37.30	3	1.0900

). Indoles (median −4.93 kcal/mol) and positive controls (median −4.76 kcal/mol) cluster at the more favorable end of the distribution, whereas SCFAs show weaker binding (median −3.58 kcal/mol). The singletons (p-cresol, lactic acid) fall near the SCFA/indole boundary. This pattern is consistent with the barplot ranking (Figure 1) and supports a binding mode that benefits from aromatic scaffolds capable of π and hydrogen-bond interactions at the FBPA active site.

Best AutoDock Vina score ΔG (kcal/mol; more negative = stronger predicted binding), approximate pKd back-calculated from ΔG (ΔG = RT·lnKd, 298 K), LE (ligand efficiency, kcal/mol per heavy atom), LLE (lipophilic ligand efficiency ≈ pKd − LogP), and standard physicochemical descriptors (MolWt, HeavyAtoms, HBD/HBA, TPSA, RotB, LogP). Values are computational predictions.

The per-ligand table (Table 2) summarizes docking and developability features across the 14 compounds (Figure 3). Predicted binding free energies span −5.199 to −2.909 kcal/mol (median −4.111), with nitazoxanide and albendazole recovered at the top, followed closely by indole-3-acetic acid (IAA). Notably, LLE penalizes lipophilicity: albendazole shows LLE −0.52 (LogP 4.31), whereas IAA and tryptamine retain favorable LLE ≈ 2 at substantially lower LogP (≈1.7–1.8). Descriptor ranges (MolWt 60–336 g·mol⁻¹, TPSA 15.8–113 Ų, RotB 0–8) suggest that several microbiome-derived aromatics fall within classic drug-like envelopes, whereas short-chain fatty acids—while chemically tractable—consistently populate the weaker end of the ΔG distribution.

Each point is a ligand colored by chemical class; higher pKd indicates stronger predicted potency (back-calculated from Vina ΔG at 298 K), LogP is RDKit cLogP (lipophilicity). Labeled points highlight the top region. The plot reveals a moderate positive trend between potency and lipophilicity, with two notable efficient zones (Figure 3):

(i)

Indoles—indole-3-acetic acid (IAA) (pKd ≈ 3.8, LogP ≈ 2.7) and tryptamine (≈3.6, 1.6)—show high potency at lower LogP;

(ii)

Controls—nitazoxanide (≈3.85, 2.7) achieves strong potency with moderate LogP, while albendazole (≈3.8, 4.3) is potent but lipophilicity-driven. Metronidazole sits at low LogP (~−0.5) with mid potency (~3.2). SCFAs cluster at lower potency (pKd ≈ 2.1–3.1) and modest LogP (≈0.5–1.2).

Boxplots summarize the best AutoDock Vina scores (more negative = stronger predicted binding) by chemical class (Figure 4). Whiskers represent 1.5 × IQR. Sample sizes: Control (n = 4), Indole (n = 3), SCFA (n = 6), Aromatic (n = 1), Microbiome (n = 1). Medians (kcal/mol): Indole −4.93, Control −4.76, SCFA −3.58, Aromatic −4.02, Microbiome −3.50. Controls and indoles cluster at more favorable ΔG than SCFAs, in line with the ranking in Figure 1. (Formal statistics are reported in the group-comparison table.)

When ligands are grouped by chemistry, indoles and clinical controls concentrate at significantly more negative ΔG than SCFAs, with median differences of ~1.4 kcal/mol, a trend that is consistent with size-normalized efficiency patterns observed across ligands (Figure 5). The singletons (p-cresol, lactic acid) fall near the SCFA/indole boundary. This class-level separation supports a binding mode that favors aromatic scaffolds capable of π and hydrogen-bond interactions at the FBPA active site.

Scatter of ligand efficiency (LE, kcal·mol⁻¹·heavy-atom⁻¹; higher = better per-atom efficiency) versus molecular weight (Mw). Representative compounds are labeled. A strong inverse trend is evident: as Mw increases, LE decreases. SCFAs (e.g., acetic acid) show high LE due to their very small size, yet this does not translate into absolute potency (weaker ΔG). Indoles (IAA, tryptamine) retain moderate LE at mid-range sizes with competitive ΔG, whereas controls (nitazoxanide, albendazole) are heavier with lower LE but still rank near the top by ΔG.

The LE–Mw plane highlights a clear decreasing relationship: smaller molecules maximize per-atom efficiency, but not necessarily overall potency. SCFAs cluster at the highest LE (e.g., acetic acid ≈0.72) consistent with minimal size, though their ΔG remains modest. In contrast, IAA and tryptamine combine intermediate Mw with LE ~0.36–0.41 and favorable ΔG, indicating a more balanced efficiency–potency profile. Albendazole and nitazoxanide fall into the low-LE (~0.22–0.27) region due to larger size, yet retain strong ranking by ΔG. Overall, the plot stresses that LE can over-reward very small fragments; thus, prioritization should use LE alongside LLE and pKd (see Figure 4 and Table 2 and

Table 3. Consensus rescoring of the top ligands against Giardia lamblia FBPA (PDB: 3OHI): AutoDock Vina vs. Vinardo.

No.	Ligand	Vina (kcal/mol)	Vinardo (kcal/mol)
1	Nitazoxanide	−4.433	−3.175
2	Albendazole	−5.198	−3.524
3	Indole-3-acetic acid (IAA)	−5.107	−4.212
4	Tryptamine	−4.912	−4.340
5	Metronidazole	−4.203	−3.022
6	Indole	−4.271	−3.613
7	Isovaleric acid	−3.302	−2.989
8	Valeric acid	−3.343	−3.968
9	Tinidazole	−3.997	−3.175
10	p-Cresol	−4.022	−3.671

) to mitigate size bias.

Rescoring with Vinardo broadly preserved the high-affinity tier identified by Vina, while causing moderate rank shifts typical of cross-scorer comparisons. Notably, indole-3-acetic acid (IAA) and tryptamine remain among the top entries under both functions. Tryptamine becomes the top hit with Vinardo (−4.34 kcal/mol), and IAA stays in the top-2 (−4.21). Among the controls, nitazoxanide and albendazole—top-ranked by Vina—remain competitive but drop several positions under Vinardo (−3.18 and −3.52, respectively), indicating potency is not exclusively lipophilicity-driven. Within SCFAs, valeric acid improves relative position with Vinardo (−3.97), while isovalerate stays in the lower tier across both. Overall, the upper quartile overlaps across scorers, supporting the robustness of the primary hits, with expected local re-ordering between chemically similar compounds (

Table 4. Docking-score stability across random seeds for top ligands (AutoDock Vina, target: Giardia lamblia FBPA, PDB 3OHI).

No.	Ligand	Mean ΔG (kcal/mol)	SD (kcal/mol)	n (Replicates)
1	Albendazole	−5.159	0.00497	3
2	Indole-3-acetic acid (IAA)	−5.138	0.00680	3
3	Tryptamine	−4.922	0.01034	3
4	Nitazoxanide	−4.584	0.43520	3
5	Metronidazole	−4.270	0.05361	3

).

Replicate docking indicates excellent numerical stability for albendazole, IAA, and tryptamine (SD ≤ 0.011 kcal/mol; CV ≤ 0.21%), implying that sampling noise is negligible relative to the typical accuracy of Vina (≈0.3–0.5 kcal/mol). Metronidazole shows modest variability (SD 0.054; CV ≈ 1.3%). In contrast, nitazoxanide exhibits substantial dispersion (SD 0.44; CV ≈ 9.5%), compatible with multiple low-energy poses and/or insufficient convergence for this chemotype. The identity of the top tier (albendazole/IAA/tryptamine) is therefore stable across seeds, while nitazoxanide should be inspected further (pose clustering and re-dock with higher exhaustiveness) (

Table 5. Residue-level contact profiles for the top-ranked ligands against Giardia lamblia FBPA (PDB 3OHI).

Index	Ligand	Group	ΔG (kcal/mol)	No. of Contacts	Residues Contacted (Chain: Residue)
0	Nitazoxanide	Control	−5.199	345	MET28:B, SER57:A, MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:B, GLU67:A, LEU97:A
1	Albendazole	Control	−5.165	373	ASP58:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:B, CYS66:A, GLU67:A, ALA68:B
2	Indole-3-acetic acid (IAA)	Indole	−5.147	235	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:B, CYS66:A, GLU67:A
3	Tryptamine	Indole	−4.930	209	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:B, GLU67:A
4	Metronidazole	Control	−4.363	232	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:A, CYS66:A, GLU67:A
5	Indole	Indole	−4.278	173	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:A, GLU67:A
6	Isovaleric acid	SCFA	−4.209	137	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:A
7	Valeric acid	SCFA	−4.111	122	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:A, LEU65:B
8	Tinidazole	Control	−4.035	140	ILE31:A, MET59:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:A, GLU67:A, ALA68:B
9	p-Cresol	Aromatic	−4.021	148	MET59:A, ILE60:A, TYR61:A, LEU62:A, LYS63:A, LYS64:A, LEU65:A, GLU67:A

).

For each ligand, the table reports the Vina score (ΔG), the number of heavy-atom contacts within 4.0 Å (n_contacts), and the set of protein residues contacted by the best pose (chain ID shown as A/B). A conserved pocket comprising MET59–ALA68 dominates the interactions of high-scoring ligands.

Contact mapping converged on a compact hotspot spanning MET59–ALA68 on chain A (with occasional B-chain symmetry equivalents), consistently contacted by all top entries (controls and indoles). The most recurrent residues were MET59, ILE60, TYR61, LEU62, LYS63, LYS64, LEU65, GLU67 and ALA68. Within this motif, TYR61 provides an aromatic platform, while LYS63/LYS64 and GLU67 likely mediate polar/ionic or H-bond interactions; MET59/ILE60/LEU62 build a hydrophobic wall that accommodates aromatic scaffolds.

▪

Indole class (IAA, tryptamine, indole): rich contacts with TYR61–LEU62–LYS63 and GLU67/ALA68, consistent with π-stacking/edge-to-face against TYR61 and H-bonding near GLU67.

▪

Controls: albendazole shows extensive packing (n_contacts = 373) and additional reaches to ASP58/CYS66, while nitazoxanide engages the canonical pocket but also touches residues outside the core (SER57, MET28, GLU71, LEU97), explaining its higher pose dispersion in seed-stability tests.

▪

SCFAs (isovaleric/valeric acids): fewer, more superficial contacts (n_contacts ≈ 120–140) largely limited to MET59–LEU62, in line with their weaker ΔG.

Together, the shared MET59–ALA68 hotspot explains why indoles (aromatic, H-bond capable) outperform SCFAs (aliphatic) and supports the binding mode proposed in the 3D panels (Figure 6).

Heat map of heavy-atom contacts within 4.0 Å between each ligand (rows) and pocket residues (columns; chain ID shown). Color encodes the number of contacts (scale at right). A conserved hotspot centered on TYR61:A–LEU62:A–LYS63:A–LYS64:A dominates the interactions across high-scoring ligands; MET59:A/ILE60:A contribute upstream, whereas GLU67:A/ALA68:A appear as distal, lower-intensity contacts. Albendazole and IAA display the broadest and most intense footprints, while SCFAs show sparser contact patterns largely confined to MET59–LEU62. Occasional peripheral contacts (e.g., ASP58:A, SER57:A, MET28:B) are ligand-specific.

Contact profiling consolidates the binding mode inferred from individual poses: a shared pocket spanning MET59–ALA68 is repeatedly engaged, with TYR61/LEU62/LYS63/LYS64 forming the core contact quartet. The strongest densities are observed for albendazole (peak at LYS64:A) and indole-3-acetic acid (IAA), consistent with their high ranks. Tryptamine follows a similar footprint but with reduced intensity. SCFAs (isovalerate/valerate) produce narrow, weaker patterns mostly at MET59–LEU62, matching their poorer ΔG. These residue-level trends explain the class separation seen in Figure 1, Figure 2, Figure 3 and Figure 4 and support an aromatic-friendly microenvironment around TYR61–LYS63/LYS64.

Detailed examination of residue–ligand contact patterns revealed that TYR61 and GLU67 form the core of the catalytic pocket, engaging in hydrogen-bond and π–π stacking interactions with multiple top-ranked ligands. HIS109 and ASP131 provide additional stabilization through polar contacts, while the indole ring system enhances hydrophobic packing within the catalytic groove.

These high-resolution contact patterns highlight the catalytic importance of the TYR61–GLU67 hotspot and its role in mediating hydrogen-bond and aromatic interactions that stabilize ligand binding (

Table 6. Recurrence of pocket residues across the top-N docked ligands (FBPA, PDB 3OHI).

Residue (Chain:ID)	Frequency in Top-N Ligands
TYR61:A	10
ILE60:A	10
MET59:A	10
LYS63:A	10
LYS64:A	10
LEU65:A	10
LEU62:A	10
GLU67:A	8
ALA68:B	6
LEU65:B	5
LYS64:B	5
CYS66:A	3
ASP58:A	1
ILE31:A	1
SER57:A	1

). Future studies will include molecular dynamics (MD) refinement and molecular mechanics/generalized Born surface area (MM/GBSA) per-residue energy decomposition to further quantify these stabilizing contributions [22,23].

Residue recurrence consolidates the heat-map and per-ligand tables: a highly conserved interaction nucleus comprises MET59–TYR61–LEU62–LYS63–LYS64–LEU65 (all 10/10), with GLU67 contributing in most cases (8/10). This MET59–ALA68 region defines a compact recognition motif that preferentially accommodates aromatic ligands (indoles, controls), explaining their superior ΔG relative to SCFAs. The presence of B-chain counterparts (e.g., LEU65:B, LYS64:B) indicates pocket symmetry but lower recurrence, consistent with primarily chain-A–centered binding (Figure 7).

Box-and-scatter plot of the best AutoDock Vina score per ligand (more negative = stronger predicted binding) across chemical classes. A Kruskal–Wallis test across the five groups yielded p = 0.0543 (trend-level). Group medians (kcal/mol): Indole −4.93, Control −4.76, SCFA −3.58, Aromatic −4.02 (singleton), Microbiome −3.50 (singleton). Dots show individual ligands.

Across groups, indoles and controls concentrate at more favorable ΔG than SCFAs, consistent with Figure 1 and Figure 2. The overall Kruskal–Wallis p = 0.0543 indicates a near-significant trend given the small sample size and the presence of singletons in Aromatic and Microbiome, which limit power. Median differences (indole vs. SCFA ≈ 1.35 kcal/mol; control vs. SCFA ≈ 1.18 kcal/mol) support a class effect in the expected direction (Figure 8). We therefore interpret the class separation as biologically meaningful but statistically underpowered in this dataset; pairwise non-parametric tests (e.g., Dunn/MWU with FDR) were considered where applicable.

Scatter of ΔG (kcal/mol; more negative = stronger predicted binding) vs. molecular weight (Mw) with a least-squares fit (orange). We observe a strong inverse correlation (r = −0.77; p = 6.99 × 10⁻⁴; n = 15), i.e., heavier ligands tend to achieve more favorable Vina scores. While expected—more atoms can form more contacts—this size dependency motivates reporting LE/LLE and potency–lipophilicity analyses (Figure 3 and Figure 5; Table 2) to avoid over-prioritizing large, lipophilic molecules.

Docking scores scaled negatively with Mw (Figure 9; r = −0.77, p = 0.000699), accounting for roughly R² ≈ 0.59 of the variance. The high-Mw controls (albendazole, nitazoxanide) contribute to the left-tail of ΔG, whereas indoles deliver competitive ΔG at mid-range Mw, and SCFAs cluster at low Mw with weaker ΔG. Because this trend reflects a known size bias of empirical scorers, we complemented the analysis with LE (size-normalized efficiency) and LLE (penalizing lipophilicity), which favor IAA/tryptamine over purely lipophilicity-driven solutions. As a robustness check, we recommend also reporting Spearman’s ρ or a partial correlation ΔG~Mw|LogP to decouple size from lipophilicity.

The FBPA receptor is shown as a grey ribbon; ligands are orange sticks (Figure 10). The callouts report the best AutoDock Vina ΔG (kcal/mol; more negative = stronger predicted binding). Panels (left→right, top then bottom): Nitazoxanide, Albendazole, Indole-3-acetic acid (IAA), Tryptamine, Metronidazole, Indole.

All poses converge in the same cavity defined by MET59–TYR61–LEU62–LYS63–LYS64–LEU65–GLU67–ALA68. Indoles orient the aromatic ring parallel to TYR61 and place donors/acceptors toward GLU67/ALA68, whereas albendazole and nitazoxanide extend deeper toward CYS66/ALA68. Metronidazole sits more superficially and indole makes fewer contacts, consistent with their weaker ΔG.

▪

The same hotspot (MET59–ALA68) is engaged by all top ligands, matching the contact heat map and frequency analysis.

▪

Binding depth and ring orientation correlate with ΔG: IAA ≈ albendazole > tryptamine > metronidazole/indole.

▪

The pocket appears aromatic-friendly (π interactions with TYR61) with a polar edge (GLU67) that the indoles exploit efficiently.

Minor differences in ligand ranking between AutoDock Vina and Vinardo were observed, reflecting variations in force-field parameterization. However, the overall rank correlation was strong (Spearman’s ρ = 0.82), indicating consistency between both scoring schemes [12,24].

A positive correlation between molecular size and docking score was observed, which likely reflects the energetic scaling associated with increased ligand complexity rather than true differences in binding affinity. This observation is presented as a descriptive trend rather than a direct predictor of inhibitory potency.

2D depictions (RDKit style) for the compounds discussed in the text. Labels report the best AutoDock Vina ΔG (kcal/mol; more negative = stronger predicted binding). Atom colors: C black, N blue, O red, S yellow, Cl green. From top-left to bottom-right: nitazoxanide, albendazole, indole-3-acetic acid (IAA), tryptamine, metronidazole, indole, isovaleric acid, valeric acid, tinidazole.

▪

Indole family (IAA, tryptamine, indole): common indole ring enabling π-stacking with TYR61 and H-bonding via side-chain donors/acceptors (Fig. panels). IAA’s carboxylate side chain explains its ΔG close to the best controls.

▪

Clinical controls: albendazole (benzimidazole carbamate) and nitazoxanide (thiazolide) are the top scorers; both provide extended aromatic surfaces plus heteroatoms for polar contacts near GLU67/ALA68. Metronidazole/tinidazole (nitroimidazoles) are smaller and bind more shallowly, in line with intermediate ΔG.

▪

SCFAs (isovaleric/valeric): short aliphatic acids lack aromatic surface; they give fewer hydrophobic/π contacts and weaker ΔG.

Across replicate-seed docking runs, most ligands showed consistent binding energies (ΔΔG < 0.3 kcal/mol), confirming numerical reproducibility. Nitazoxanide displayed slightly higher variation, likely reflecting its conformational adaptability and multiple binding orientations within the fructose-1,6-bisphosphate aldolase (FBPA) active site [15,25].

4. Discussion

Our docking campaign against the fructose-1,6-bisphosphate aldolase (FBPA) of Giardia lamblia (PDB ID: 3OHI) identified microbiome-derived indoles, particularly indole-3-acetic acid (IAA) and tryptamine, as top-ranked ligands approaching the performance of clinical antiparasitic benchmarks such as albendazole. In contrast, short-chain fatty acids (SCFAs) displayed consistently weaker affinities. All high-ranking ligands converged on a compact binding hotspot defined by MET59–TYR61–LEU62–LYS63–LYS64–LEU65–GLU67–ALA68, where TYR61 provides an aromatic stacking platform and GLU67/ALA68 form a polar edge—consistent with crystallographic descriptions of the FBPA active site [4].

Current giardiasis treatment relies primarily on nitroimidazoles (metronidazole, tinidazole), with albendazole as a validated alternative [1,2,3]. Our control compounds reproduced this order in silico, confirming the reliability of the docking workflow. The fact that simple indoles—IAA and tryptamine—rank near these clinical references suggests a dual-mechanism hypothesis: (i) direct enzymatic inhibition of FBPA, and (ii) potential host-modulatory effects previously described for these metabolites, such as activation of the aryl hydrocarbon receptor (AhR) and pregnane X receptor (PXR) pathways [5,8].

Docked poses revealed that indoles align their bicyclic rings parallel to TYR61 (π–π stacking) and orient hydrogen-bond donors and acceptors toward GLU67 and ALA68, explaining the superior ΔG of IAA (via its carboxylate anchor) and the intermediate affinity of tryptamine. Conversely, albendazole and nitazoxanide penetrate deeper into the cavity, engaging CYS66 and ALA68 through extended aromatic and polar interactions, consistent with their stronger binding. These observations agree with residue-level contact maps and previously reported FBPA topologies [4].

A significant negative correlation between predicted binding free energy (ΔG) and molecular weight confirmed the known size bias of empirical scoring functions [9,10,11,12]. To mitigate this, we incorporated ligand efficiency (LE) and lipophilic ligand efficiency (LLE), which normalize for molecular size and hydrophobicity [11]. Indoles—particularly IAA and tryptamine—achieved a more balanced LE/LLE profile than albendazole or nitazoxanide, which are more lipophilic. These metrics thus provide a comparative framework for rational prioritization rather than absolute potency [26,27].

The use of AutoDock Vina and Vinardo enabled consensus-based evaluation of binding energies through complementary scoring functions [13,14]. Rank concordance across scorers (Spearman’s ρ = 0.82) and reproducibility across random seeds support the robustness of our predictions. Nonetheless, docking energies are approximations of true binding free energies; therefore, molecular dynamics (MD) simulations and free-energy refinements such as molecular mechanics/generalized Born surface area (MM/GBSA) remain essential for quantitative validation [15].

The residue-level hotspot analysis identified TYR61 and GLU67 as catalytic anchors mediating hydrogen-bond and π-stacking interactions with multiple top ligands. HIS109 and ASP131 contribute additional stabilization within the catalytic pocket. These findings suggest that the MET59–ALA68 region forms a recognition motif that preferentially accommodates aromatic ligands with polar termini—the “chemical grammar” characteristic of indoles. Medicinal chemistry refinements could explore: (i) carboxylate or sulfonamide bioisosteres to strengthen GLU67 interactions; (ii) benzimidazole-based analogues of albendazole to reduce lipophilicity while preserving aromatic contacts; and (iii) prodrug strategies targeting the intestinal microenvironment [4,11].

Methodologically, the inclusion of replicate docking runs with independent random seeds increased numerical stability and reduced stochastic bias [13,15]. The moderate variability observed for nitazoxanide likely reflects its conformational flexibility, warranting further investigation through MD trajectory analysis.

This study acknowledges the limitations inherent to rigid-receptor docking, which does not explicitly account for protein flexibility or solvation dynamics. The grid was defined to encompass all catalytic residues and the Zn²⁺ cofactor, preserving the active-site geometry, yet water-mediated and induced-fit effects were not modeled. Future work will incorporate explicit-solvent MD simulations, 3D reference interaction site model (3D-RISM) or WaterMap hydration analysis, and subsequent MM/GBSA or free energy perturbation (FEP) calculations to evaluate the energetic plausibility of the proposed binding modes [28,29].

The dual-mechanism model proposed here—direct FBPA inhibition combined with potential host modulation by microbiome-derived indoles—should be regarded as a working hypothesis requiring biological validation. Indoles are known to modulate epithelial and immune responses through AhR and PXR signaling [7,8]. Biochemical inhibition assays targeting FBPA and trophozoite culture experiments will determine whether these effects coexist and contribute to the observed docking behavior.

Finally, benchmarking microbiome metabolites against clinically established antigiardial drugs (albendazole, nitazoxanide, metronidazole, tinidazole) provided a translational reference for interpreting docking scores [30]. While the present work is limited to in silico predictions, it establishes a reproducible and transparent computational framework for the discovery of antiparasitic lead compounds. Future efforts will integrate biochemical enzyme assays, trophozoite viability tests, and MD-based refinement to experimentally validate the predicted TYR61–GLU67 interactions and confirm the structural and functional relevance of the proposed binding hotspot [31].

Although this study focused on structure-based docking as an initial in silico screening phase, further validation through molecular dynamics (MD) and ADMET analyses is planned as the next stage of research. Specifically, 100 ns MD simulations under explicit solvent conditions will be conducted to evaluate the dynamic stability of the TYR61–GLU67 catalytic interactions and the persistence of ligand binding throughout the trajectory. In parallel, in silico ADMET profiling will be performed to assess absorption, distribution, metabolism, excretion, and toxicity parameters once the top indole scaffolds are chemically optimized. The methodological framework will follow recent MD validation studies [32], ensuring consistency with established best practices in structure-based drug discovery. These future analyses will provide a comprehensive dynamic and pharmacokinetic validation of the docking-derived hypotheses.

5. Conclusions

Collectively, our results identify microbiome-derived indoles, particularly indole-3-acetic acid (IAA) and tryptamine, as promising scaffolds targeting fructose-1,6-bisphosphate aldolase (FBPA) of Giardia lamblia. These ligands achieved in silico binding affinities comparable to albendazole, combined with favorable lipophilic ligand efficiency (LLE) and a consistent binding mode anchored by the TYR61–GLU67 catalytic motif.

Given their established host-beneficial properties, indoles represent a potential dual-mechanism strategy: direct enzymatic inhibition of FBPA together with host modulation through microbiome-derived signaling pathways [5,8]. These hypotheses warrant validation through molecular dynamics (MD) simulations, solvation modeling, and in vitro biochemical and trophozoite assays to confirm the stability and biological relevance of the proposed binding interactions [28,29].