Introduction

Lipoic acid (ALA), a natural antioxidant also known as 1,2-dithiolane-3-pentanoic acid, is essential in mitochondrial enzymatic activities1,2,3. This organosulfur compound, found in plants and animals, was initially considered an acetate replacement and discovered by Reed in 19514. Its first medical use was in 1959 to treat Amanita phalloides poisoning5. In the Krebs cycle, ALA acts as a cofactor in enzyme complexes, forming covalent bonds with proteins, enhancing its therapeutic potential6. Its chiral nature results in two optical isomers, R-lipoic acid and S-lipoic acid, which further expand its applications5,6.

ALA may increase cellular antioxidant potential indirectly by facilitating the absorption or synthesis of endogenous low-molecular-weight antioxidants, including ubiquinone, glutathione (GSH), and ascorbic acid. This property is further increased by reducing oxidized forms of ubiquinone, GSH, and vitamin C, as well as promoting the uptake of cysteine and cystine from the plasma, which are later converted into cysteine7,8,9,10. Apart from this, ALA can enhance intracellular levels of GSH in various cell types and tissues, while dihydrolipoic acid (DHLA) supports the conversion of cystine to cysteine, an amino acid crucial for GSH synthesis11,12,13,14.

It is believed that supplementing with antioxidants may help reduce the impacts of infections, including inflammation of the kidneys15, liver16 and heart in rats17. Since oxidative stress is recognized as an important factor in the development of various viral infections, as evidenced in previous studies]. The therapeutic importance of antioxidants against viral infections has been extensively validated through various in vitro and in vivo investigations, including their protective effect against hepatotoxicity19,20, and in rat spleen tissues21.

This acid has been investigated not only for its antioxidant properties but also for anti-inflammatory activity, which may be more relevant in the context of viral infections, including the Omicron variant of SARS-CoV-2. Omicron, known for its high potential for mutation and immune escape, can induce a significant increase in oxidative stress and inflammation in the body. ALA may help mitigate these effects by reducing the production of ROS, and by promoting the regeneration of endogenous antioxidants such as glutathione22,23,24,25. In addition, ALA may modulate the immune response, potentially helping to control the exacerbated inflammation associated with infection by the Omicron variant while improving cellular antioxidant status, thus offering auxiliary therapeutic support in the management of COVID-1926,27,28,29.

The use of ALA has been associated with a reduced increase in the Sequential Organ Failure Assessment (SOFA) score and a lower 30-day all-cause mortality compared to placebo30. Despite the mortality rate being twice as high in the placebo group as in the ALA group, the statistical difference observed was marginal, likely due to the limited sample size23. Further studies with larger patient cohorts are required to confirm the potential role of ALA in the treatment of critically ill patients with COVID-191,23.

This study intends to examine the parameters related to the absorption, distribution, metabolism and excretion (ADME) of ALA, using ADME-based analyses, metabolic studies and structural binding analyses. In addition, a molecular docking analysis was performed to explore how ALA interacts with the Omicron variant of SARS-CoV-2.

Materials and methods

Absorption, distribution, metabolism and excretion (ADME) study

Calculation of molecular Absorption, Distribution, Metabolism, and Excretion (ADME) parameters can be achieved using Swiss ADME software, which takes the SMILES format “C1CSS[C@@H]1CCCCC(= O)O”31,32,33,34,35. This method evaluates physicochemical properties, lipophilicity, water solubility, pharmacokinetics, and synthetic accessibility of the studied ligands36.

Computational assessment of acute toxicity using QSAR models with stoptox

To evaluate the acute toxicity of the compound, the STopTox tool was employed, utilizing Machine Learning (ML) and Quantitative Structure-Activity Relationship (QSAR) models. Predictions were performed for six toxicity endpoints: oral, dermal, inhalation, skin irritation/corrosion, ocular damage, and dermal sensitization. The model’s reliability was ensured by referencing validated experimental data35.

The chemical structure of the compound (PubChem CID: 6112) was retrieved from the PubChem database, and its SMILES notation was extracted. This SMILES code was then input into the STopTox application for computational analysis, enabling toxicity predictions based on molecular descriptors and QSAR modeling.

Prediction of toxicity of chemicals

ProTox 3.0 predicts toxicity by integrating molecular similarity, fragment-based propensities, frequently occurring features, and machine learning approaches. The platform employs CLUSTER cross-validation, which is based on fragment similarity, to enhance predictive accuracy. Built on 61 predictive models, ProTox 3.0 assesses multiple toxicity endpoints, including acute toxicity, organ toxicity, toxicological outcomes, molecular initiating events, metabolism, adverse outcome pathways (Tox21), and toxicity targets37. This comprehensive framework enables a robust evaluation of potential toxic effects based on molecular structure and known toxicological data.

Prediction of small molecule protein targets

The predictive analysis identified the most probable macromolecular targets for the bioactive compound by utilizing 2D and 3D similarity assessments. This approach leveraged a comprehensive database of 37,000 compounds, encompassing interactions with over 3,000 proteins from Homo sapiens, Mus musculus, and Rattus norvegicus38,39,40,41. The predictions were based on structural similarity, allowing for the identification of potential biological targets and interaction patterns.

Epoxidation

Epoxides are highly reactive metabolites, typically generated by cytochrome P450 through oxidation of unsaturated or multiple bonds. The specific molecular region where epoxidation occurs is referred to as the epoxidation site (SOE). A novel computational approach systematically analyzes and quantifies epoxidation reactions, achieving 94.9% accuracy in predictive modeling and 78.6% accuracy in distinguishing epoxidized from non-epoxidized compounds42. This method enhances the understanding of metabolic transformations and potential bioactivation pathways of chemical compounds.

Quinonation

Quinones, including key intermediates such as quinone imine moieties, quinone methides, and imino-methides, are highly reactive electrophilic Michael acceptors. These compounds account for over 40% of known bioactive metabolites and are primarily formed by the enzymatic action of cytochrome P450 and peroxidases. This novel predictive method is the first to comprehensively model quinone formation, capturing both single-step and multi-step processes. At the atomic level, it accurately identifies quinone formation sites with an AUC of 97.6% and predicts quinone-forming molecules with an AUC of 88.2%43. This approach enhances the understanding of quinone bioactivation and its potential toxicological implications.

Reactivity

Approximately 40% of drug candidates fail due to safety concerns arising from interactions between electrophilic drugs or their metabolites and nucleophilic biological macromolecules, such as DNA and proteins. These interactions can lead to toxicity and adverse effects, limiting the success of drug development. To address this issue, a deep convolutional neural network was developed to predict both the sites of reactivity (SOR) and the overall molecular reactivity. Cross-validation results demonstrated high predictive accuracy, with an AUC of 89.8% for SOR in DNA and 94.4% for SOR in proteins. Additionally, the model effectively distinguished reactive from non-reactive molecules, achieving AUC values of 78.7% for DNA and 79.8% for proteins44,45.

Phase 1

Phase I enzymes metabolize over 90% of FDA-approved drugs, catalyzing diverse reactions that often generate structurally unpredictable metabolites. To enhance the identification of metabolic transformations, we developed a system that labels both metabolic sites and reaction types, classifying them into five primary categories: stable oxidations, unstable oxidations, dehydrogenation, hydrolysis, and reduction. This classification enables the accurate recognition of 21 distinct Phase I reactions, covering 92.3% of the reactions recorded in our laboratory database. Using this labeling system, we trained a neural network on 20,736 human Phase I metabolic reactions, achieving a cross-validation AUC accuracy of 97.1%46. This approach improves the predictive modeling of drug metabolism and enhances the understanding of metabolic pathways.

N-dealkylation

Metabolic studies often overlook aldehyde analysis, despite their potential toxicity. While aldehydes are generally converted into carboxylic acids and alcohols through detoxification pathways, some remain highly reactive and evade metabolism, leading to adduct formation with DNA and proteins and triggering adverse effects. To address this, a predictive model was developed to identify N-dealkylation sites in metabolized substrates. The model demonstrated 97% accuracy in the first two validation cases and achieved an AUC of 94% in the receiver operating characteristic (ROC) curve analysis47. This approach enhances the understanding of aldehyde reactivity and its implications in drug metabolism.

UGT conjugation

Uridine diphosphate glucuronosyltransferases (UGTs) metabolize approximately 15% of FDA-approved drugs, playing a crucial role in drug clearance and detoxification. Rapid identification of UGT metabolism sites is essential for optimizing lead compounds in drug development. The XenoSite UGT model predicts UGT-mediated metabolic sites in drug-like molecules with high accuracy. In a training dataset of 2,839 UGT substrates, the model achieved 86% accuracy in Top-1 predictions and 97% accuracy in Top-2 predictions48. This computational approach enhances drug metabolism studies by providing fast and reliable insights into UGT-mediated transformations.

ProtParam

The ProtParam tool was used to calculate various physical and chemical parameters for a given protein. The protein sequence was obtained from the PDB (PDB ID: 7TLY) in FASTA format. The computed parameters include molecular weight, theoretical pI, amino acid composition, atomic composition, extinction coefficient, estimated half-life, instability index, aliphatic index, and grand average hydropathy (GRAVY)49.

SinalP - 6.0

Signal peptides (SPs) are short amino acid sequences that regulate the secretion and translocation of proteins across cellular membranes in all living organisms. To predict SPs, we utilized the SignalP 6.0 model, a machine learning-based tool capable of detecting all five known types of SPs. This model was applied to sequence data and is designed to work effectively with metagenomic data. The tool’s ability to identify various SP types offers a comprehensive approach for analyzing protein sequences in diverse biological contexts50.

NetNGlyc − 1.0

To identify N-linked glycosylation acceptor sites in protein sequences, we used artificial neural networks trained with the surrounding sequence context. The consensus sequence Asn-Xaa-Ser/Thr (where Xaa is not Pro) was used as a basis for prediction, although not all such sequences undergo modification. In cross-validation, the model was able to identify 86% of glycosylated sequences and 61% of non-glycosylated sequences, with an overall accuracy of 76%51.

Recipient treatment

The protein target of interest, chosen based on a literature review, underwent molecular docking analysis. The target protein (PDB ID: 7TLY) and its corresponding ligand were retrieved from the Protein Data Bank, a comprehensive repository of protein 3D structures. Each entry in the PDB contains detailed information such as atomic coordinates, polymer sequences, and associated metadata. The protein inhibitors and water molecules were removed from the receptor structure using DISCOVERY STUDIO 2021 CLIENT software.

Binder treatment

For in silico analysis, ALA and Ebselen were selected for molecular docking studies. The ligand was modeled in 3D using ACD/ChemSketch software, and the 2D model was obtained from ChemSpider (CSID: 841 and CSID: 3082). The compounds were docked using a “flexible ligand with rigid protein” approach through the Autodock VINA system in the PyRx software52. After the docking process, the most stable binding conformations were evaluated with the Discovery software.

Grid and fitting calculation

The grid calculation was performed with 100 conformations using the AutoDock Vina system in the PyRx software. For the ligand-protein docking, the grid was configured with dimensions of 126 × 126 × 126 Å on the X, Y and Z axes and a spacing of 0.375 Å. The grid centre was set at coordinates 231.737, 187.05, and 260.473 Å for the 7TLY protein. The binding site was identified using ligands that had been co-crystallized with the protein in previous studies and are accessible in the Protein Data Bank. The interaction energy between these ligands and the amino acids of the 7TLY protein was computed using Discovery Studio software with RMSD < 2, which evaluates binding energy by considering components such as van der Waals interactions, electrostatic forces, and hydrogen bonds.

Results

Absorption, distribution, metabolism and excretion

ALA is composed of 12 heavy atoms and has unique chemical characteristics, including the ability to accept two hydrogen bonds and donate one hydrogen bond. According to Table 1, it is possible to observe a total polar surface area (TPSA) of 87.90 Ų. This suggests that ALA has a relatively high degree of polarity, which is typically associated with good solubility in aqueous environments, enhancing its absorption in the gastrointestinal tract. ALA demonstrates a high absorption capacity of the compound.

Table 1 SwissADME numerical results: physicochemical properties, lipophilicity, pharmacokinetics, drug similarity, and medicinal chemistry.
Full size table

According to the data in Table 1, ALA showed high absorption in the gastrointestinal tract, but showed low transcutaneous absorption, with a rate of -6.37 cm/s, which indicates limited dermal penetration. This low transcutaneous absorption may suggest that ALA is not effective for topical applications and may require alternative methods of administration to achieve therapeutic effects. In addition, ALA failed to cross the blood-brain barrier (BBB) and was not identified as a P-glycoprotein (P-gp) substrate. This lack of BBB penetration further limits its potential use in central nervous system-targeted therapies. The compound also did not exhibit inhibitory activity on cytochrome P450 enzymes such as CYP1A2, CYP2C19, CYP2C9, CYP2D6, and CYP3A4. This suggests that ALA is unlikely to interfere with the metabolism of other drugs, making it a relatively safe compound for concurrent use with other treatments.

In addition, ALA met the Lipinski criteria, without violations, and showed compatibility with the parameters of Ghose, Veber, Egan, and Muegge. This indicates that ALA has favorable drug-like properties, which could facilitate its development as a potential pharmaceutical compound. The ALA bioavailability score was recorded at 0.56, which indicates a moderate potential for bioavailability, with a flag at the disulfide site in Brenk. This bioavailability score suggests that while ALA has some potential for absorption, there may be challenges related to its solubility or metabolic stability that need to be addressed for optimal therapeutic efficacy.

Prediction of toxicity parameters

The in silico toxicity prediction assays for ALA indicated a predicted LD50 of 502 mg/kg and a Predicted Toxicity Class of 4. This suggests that ALA has a moderate potential for toxicity, with a lethal dose expected at a relatively high amount compared to highly toxic substances. The analysis achieved an average similarity of 100% and a prediction accuracy of 100%. Such high prediction accuracy implies that the in silico models used are highly reliable and provide a confident estimation of the compound’s toxicity profile. However, the predictions identified activity only in nephrotoxicity (organ toxicity) with a probability of 0.55 and BBB permeability (toxicity endpoints) with a probability of 0.87 (Table 2). The higher probability associated with BBB permeability suggests that ALA may have a potential risk of central nervous system toxicity, though further studies would be needed to confirm this.

Table 2 Toxicity prediction categorized by classification, target, prediction outcome, and probability.
Full size table

ALA demonstrated no acute inhalation toxicity, with a confidence level of 52% (Fig. 1A), and exhibited acute dermal toxicity with a confidence level of 77% (Fig. 1C). The moderate confidence in dermal toxicity indicates that ALA may cause skin irritation or other dermal effects, warranting precautions during handling. Notably, it showed acute oral toxicity with a confidence level of 80% (Fig. 1B) and eye irritation and corrosion with a confidence level of 70% (Fig. 1D). These findings highlight potential safety concerns for oral and ocular exposure, which should be considered in product development. The compound was not associated with skin sensitization, with a confidence level of 70% (Fig. 1E), and tested negative for skin irritation and corrosion, with a confidence level of 60% (Fig. 1F). This suggests that ALA may be safer in terms of skin sensitization and irritation, though some precautionary measures may still be necessary.

Small molecule metabolism

As shown in Fig. 2, ALA showed a slight reaction in the double bond of oxygen during the epoxidation process (Fig. 2A), indicating that ALA undergoes minor modifications during this process, which could influence its reactivity and metabolism in biological systems. However, there were no significant reactions in quinonation and N-dealkylation, as shown in Fig. 2B and C, respectively. This suggests that ALA is less likely to undergo these types of modifications, potentially preserving its structure and bioactivity. On the other hand, in the conjugation with UGT, there was a strong reaction with the hydroxyl group (Fig. 2D), highlighting that ALA may form conjugates that could enhance its solubility and facilitate its elimination from the body.

In interactions with GSH, ALA showed medium reactivity with sulfur and in two carbons adjacent to sulfur. This indicates that ALA may engage in detoxification processes, potentially reducing its toxicity by forming glutathione conjugates. Regarding protein, reactivity was medium with sulfur, oxygen-bound carbon, and hydroxyl groups. These interactions suggest that ALA might influence protein function, possibly by modifying key residues involved in enzymatic activity or binding. For cyanide, the reactions were few and weak, while in DNA, moderate carbon reactivity was observed (Fig. 2E). The weak reactivity with cyanide suggests that ALA is not likely to form harmful cyanide conjugates, while the moderate reactivity with DNA could indicate a potential for interaction with genetic material, requiring further investigation into its genotoxicity.

In Phase 1 reactions, ALA reacted with sulfur and with the single bond between sulfur atoms, in addition to other interactions with carbons during stable oxidation. These stable oxidation products might play a role in the metabolic activation of ALA, which could affect its pharmacological properties. However, for unstable oxidation, dehydrogenation, hydrolysis, and reductions, reactions were limited (Fig. 2F). This suggests that ALA may be relatively stable under certain metabolic conditions, which could be beneficial for its sustained activity or prolonged therapeutic effects.

Target prediction

ALA showed an affinity with targets of the Homo sapiens species, predominantly with enzymes, registering a 22% interaction rate, indicating a significant tendency to bind with these macromolecules. This strong affinity for enzymes suggests that ALA may play a role in regulating key metabolic and enzymatic pathways in human cells. Among the other classes of proteins, the compound showed a 12% affinity for family A G-protein-coupled receptors (GPCRs), which are recognized to play vital roles in the transduction of cellular signals. This interaction with GPCRs could indicate that ALA may modulate cellular signaling pathways, potentially affecting various physiological processes. In addition, ALA showed a 10% affinity for oxidoreductases, which are enzymes involved in oxidation-reduction reactions essential for various metabolic processes. These interactions suggest that ALA could influence redox homeostasis, possibly contributing to antioxidant or pro-oxidant effects.

The in silico analysis revealed that the most likely interactions of ALA occurred with acetylcholinesterase, a hydrolase crucial in the breakdown of the neurotransmitter acetylcholine, and with cyclooxygenase-2 (COX-2), an oxidoreductase involved in inflammation and prostaglandin production. These interactions highlight the potential of ALA in modulating neurochemical signaling and inflammatory responses, making it a promising candidate for therapeutic applications related to neurodegenerative diseases and inflammation. In addition, ALA showed a remarkable interaction with the enzyme SUMO-activating, which plays an important role in post-translational modification of proteins by regulating various cellular functions. This suggests that ALA may impact cellular regulation through modification of protein functions, further broadening its potential biological activity. However, the remaining interactions presented a significantly lower probability, suggesting a more restricted specificity of ALA in terms of its molecular targets (Fig. 3). This limited specificity may indicate that ALA is more selective in its interactions, potentially reducing the risk of off-target effects.

ALA exhibited a significant affinity with several targets in the species Mus musculus, especially the interaction with enzymes, where it registered a 26% interaction rate, indicating a strong tendency to bind with these macromolecules (Fig. 4). This further supports the idea that ALA’s primary mode of action may involve the modulation of enzymatic activity, which could influence metabolic and regulatory processes. In addition to enzymes, ALA also showed affinity with other classes of proteins, including 18% with G protein-coupled receptors of family A (GPCRs) and 14% with proteases, evidencing its versatility in interacting with different types of proteins. The broad reactivity with various protein classes points to ALA’s potential as a multitarget compound, capable of influencing diverse cellular processes. The highest probability of interaction was observed with Cyclooxygenase-2 (Oxidoreductase) and Acetylcholinesterase (Hydrolase), suggesting that these enzymes may be prime targets of ALA. This reinforces the importance of these two enzymes in ALA’s potential therapeutic effects. In contrast, interactions with other targets were detected, but with a relatively lower probability, which still underscores the potential of ALA to interact with a variety of proteins. These lower-probability interactions suggest that ALA may have additional effects that could be explored in further studies.

ALA exhibited a significant affinity with several targets in the species Rattus norvegicus, with emphasis on its interaction with enzymes, where it registered a rate of 30.2% (Fig. 5). The high affinity for enzymes in Rattus norvegicus suggests a similar mode of action in this species, further supporting the biological relevance of ALA’s interactions. In addition, the compound showed a considerable affinity with G protein-coupled receptors of family A (GPCRs), reaching 16.3%, and with voltage-gated ion channels, with 14%. These interactions point to ALA’s potential involvement in ion channel regulation and cellular signaling, which may affect various physiological processes. Although it presented a higher probability of interaction with the target Cyclooxygenase-2 (by homology) (Family A G protein-coupled receptor), interactions with other targets were observed in smaller proportions. This suggests that the primary therapeutic targets of ALA may be similar across different species, though further exploration of these interactions is necessary to fully understand its pharmacological potential.

Protein sequence analysis

The 7TLY protein (SARS-CoV-2 S B.1.1.529 Omicron variant) has a molecular weight of 178,519.52 Da and a theoretical pI of 6.68. It contains 145 negatively charged residues and 140 positively charged residues. The extinction coefficient at 280 nm is 202,695, with an absorbance of 1.135 at 0.1% (1 g/L), assuming cysteine residues form cystine, or 1.122 if reduced. The estimated half-life is 0.8 h in mammalian reticulocytes (in vitro), 10 min in yeast (in vivo), and 10 h in E. coli (in vivo). The instability index is 35.19 (stable), with an aliphatic index of 79.29 and a GRAVY of -0.215.

For chain A, the most likely signal peptide is Sec/SPI, with a probability of 0.8341, followed by Sec/SPII with a probability of 0.1651. The probabilities for other peptides (Tat/SPI, Tat/SPII, and Sec/SPIII) are extremely low, all below 0.0003 (Fig. 6A). Chain B shows similar results, with Sec/SPI as the most probable peptide (0.8342), followed by Sec/SPII (0.1649), while the probabilities for other peptides are minimal (Fig. 6B). In contrast, for chain C, the probability for Sec/SPII is considerably high at 0.9993, while the probability for Sec/SPI is very low (0.0002), and the probabilities for Tat/SPI, Tat/SPII, and Sec/SPIII are nearly zero (Fig. 6C).

In the sequence for Chain C, several N-glycosylation sites were identified. At positions 55, 57, 59, and 116, the sequences show varying potentials and jury agreement values, with a positive or negative indication for N-Glyc presence (Table 3; Fig. 7A). Similarly, Chain B reveals N-glycosylation sites, such as at positions 137, 138, 152, 158, and 210, with associated potential scores and jury agreement (Table 3; Fig. 7B). Chain C presents several positions of N-glycosylation sites, covering several positions, in which, depending on their potential, most of them presented a positive index (+) (Table 3; Fig. 7C).

Table 3 Analysis of sequence name, position, potential, jury agreement, and N-Glycosylation results.
Full size table

Docking molecular

Ebselen (affinity − 6.369 Kcal/mol) demonstrated distinct donor and acceptor regions for hydrogen bonding (H-bonds), as illustrated in Fig. 8A. The solvent-accessible surface area was observed to range between 22.5 and 25 Ų within the ligand regions (Fig. 8B). Aromatic interactions were limited, with the only notable interaction being an edge-to-edge contact near VAL A:93 (Fig. 8C). The interpolated charge was minimal, remaining close to zero (Fig. 8D). Hydrophobicity predominantly ranged from − 1 to -3 (Fig. 8E), and the compound exhibited a neutral ionizability profile (Fig. 8F). Ebselen formed π-alkyl interactions at VAL A:93, VAL B:86, and PRO B:41, as well as conventional hydrogen bonds at GLY B:42, GLN B:43, and LYS B:40. Additionally, LYS B:40 exhibited a secondary interaction type, amide-π stacked (Fig. 8G).

ALA (affinity − 4.223 Kcal/mol) exhibited donor regions for hydrogen bonding, forming conventional hydrogen bonds at several binding sites (Fig. 9A). The solvent-accessible surface area ranged between 22.5 and 25 Ų for all protein-binding sites (Fig. 9B). Aromatic interactions were observed, notably an edge-to-edge contact near TYR A:95 (Fig. 9C). The interpolated charge showed minimal variation, remaining close to zero (Fig. 9D). Hydrophobicity predominantly ranged between − 1 and − 3 (Fig. 9E), while the ionizability profile was neutral (Fig. 9F). Molecular interactions included conventional hydrogen bonds at LYS B:40, GLY B:42, GLN B:43, and TYR A:95, π-alkyl interactions at PRO B:41, and π-sigma interactions at VAL B:86 (Fig. 9G).

Discussion

ALA contains an asymmetric carbon atom, resulting in two optical isomers: the dextrorotatory (R-ALA or + ALA) and the levorotatory (S-ALA or -ALA). The naturally occurring form of R-ALA can be found both in its free form and lysine-conjugated residues, playing an essential role as a cofactor. In contrast, synthetic ALA is a racemic mixture of the + ALA and -ALA (+/− ALA) isomers. The oxidized and reduced forms of ALA (ALA/DHLA) are potent antioxidants, neutralizing free radicals and reactive oxygen species (ROS), while also potentiating the activity of other endogenous antioxidants, such as vitamins C and E and glutathione53. In a balanced redox state, GSH is an intracellular antioxidant with the potential to neutralize free radicals through the thiol group of its cysteine54.

Lipoic acid (ALA) is safe but has limited ocular penetration due to its lipophilicity55,56. UNR844, a choline ester of ALA, enhances corneal absorption, reaching therapeutic levels in the aqueous humor. Once metabolized into dihydrolipoic acid (DHLA), it reduces disulfide bonds in lens proteins, potentially improving near vision56,57. ALA also shows good tolerability in humans and low acute toxicity in rats, with an oral LD50 exceeding 2000 mg/kg58,59,60. However, our assays showed that ALA exhibits acute oral toxicity and eye irritation and corrosion, likely due to its disulfide bonding and overall structure. This irritation may be attributed to its poor solubility, which could affect its bioavailability and interaction with ocular tissues.

ALA shows variations in its lipophilicity depending on the calculation method used. The iLOGP index indicates a lipophilicity of 1.81, while XLOGP3 has a value of 1.68. Other methods, such as WLOGP, MLOGP, and SILICOS-IT, provide values of 2.79, 1.57, and 2.34, respectively. In addition, the octanol/water partition coefficient (Log P o/w) was calculated at 2.04 (Table 2). These results suggest that ALA has activity in both aqueous and lipophilic environments61.

The inhibition of CYP450 enzymes reduces their catalytic activity, prolonging drug half-life and increasing plasma concentration. CYP2D6 metabolizes about 30% of medications, while CYP3A4 and CYP2C9 play crucial roles in drug metabolism, with CYP3A4 responsible for approximately 50%. Inhibiting these enzymes can interfere with drug metabolism, leading to toxic effects62,63,64,65. Beigi et al.66 attribute CYP2D6 inhibition to reversible competitive inhibition. However, in the conducted assays, ALA did not exhibit any inhibitory effects on CYP enzymes.

Gastrointestinal absorption of ALA appears to be influenced by several factors, including the presence of food. Research shows that the bioavailability of ALA can vary substantially with food intake. Concomitant administration of ALA with food reduces its absorption, suggesting the involvement of carrier proteins in the uptake of the compound. When ALA is ingested with food, a significant reduction in peak plasma concentrations (C max) is observed by about 30%, in addition to an approximate 20% decrease in total plasma ALA concentrations, compared to administration in the fasted state12.

ALA has a significant ability to cross the blood-brain barrier (BBB), as demonstrated in previous studies67. After crossing the BBB and entering the central nervous system, ALA is rapidly internalized by cells and tissues, where it is readily converted to dihydrolipoic acid (DHLA)68.

ALA does not have toxicity alerts, but there is a Brenk alert related to the disulfide site in its chemical structure. Despite this, the compound showed no similarity to lead. The synthetic accessibility of ALA was evaluated at 2.87. In addition, ALA can chelate metals like zinc, iron, and copper while also regenerating endogenous antioxidants such as glutathione and exogenous antioxidants like vitamins C and E. These actions are achieved with minimal side effects.

Among the highlighted molecular targets, acetylcholinesterase and cyclooxygenase-2 (COX-2) stand out due to their widespread application in the treatment of Alzheimer’s disease, inflammation, and pain69,70,71,72,73. Aldose reductase is crucial for addressing diabetes-related complications, while the PPAR-γ receptor plays a key role in managing type 2 diabetes and metabolic disorders74,75. Additionally, prostanoid receptors and kynurenine 3-monooxygenase have therapeutic potential for inflammation and neurodegenerative diseases76,77. Other targets, such as steroid 5-alpha-reductase, are commonly used in treating benign prostatic hyperplasia and alopecia, while thromboxane-A synthase holds significance in cardiovascular diseases78.

Furthermore, iron chelators such as heparin, deferoxamine, caffeic acid, curcumin, ALA, and phytic acid have shown potential in protecting against ferroptosis, restoring mitochondrial function, balancing iron-redox potential, and re-equilibrating Fe-RH status. This restoration of Fe-RH represents a biomarker-driven host-targeted strategy for effective clinical management and post-recovery intervention in COVID-1979.

ALA shows promise in enhancing host defenses against SARS-CoV-2 by inhibiting NF-κB signaling, reducing pro-inflammatory cytokines, and preventing oxidative depletion of tetrahydrobiopterin (BH4), a cofactor essential for nitric oxide (NO) synthesis. This restores NO bioavailability, improving endothelial function. Additionally, ALA scavenges reactive oxygen species (ROS), regenerates glutathione (GSH), and stimulates GSH synthesis by enhancing cysteine uptake and activating the Nrf2/ARE pathway. Nrf2 regulates genes involved in antioxidant defenses, anti-inflammatory responses, and mitochondrial protection, highlighting ALA’s potential as a therapeutic agent for managing oxidative stress and inflammation in COVID-1924,80,81. The combined use of ALA and insulin in diabetic patients may exhibit a synergistic effect against SARS-CoV-2. This suggests that ALA treatment holds significant potential in managing COVID-19 in individuals with diabetes, offering both therapeutic and supportive benefits82.

COVID-19 infection can lead to long-term effects, including redox imbalance, mitochondrial dysfunction, and chronic inflammation due to the Warburg effect. This may result in cytokine storms, chronic fatigue, or neurodegenerative diseases. Thus, patients who have a moderate to high disease condition have a reduction in GSH levels and, therefore, an increase in free radicals compared to patients with mild levels of COVID-19. Treatments like ALA and methylene blue have shown potential to enhance mitochondrial activity, counter the Warburg effect, and promote catabolism. Combining methylene blue, chlorine dioxide, and ALA could further mitigate COVID-19’s long-term impacts by stimulating catabolism83,84.

The Omicron variant of SARS-CoV-2 manages to evade the immunity conferred by antibodies from vaccines or previous infections due to the high number of mutations accumulated in the spike protein. To better understand this antigenic change, analyses using cryo-electron microscopy and X-ray crystal structures of the spike protein were performed, especially in the receptor-binding domain. These studies were conducted in conjunction with the monoclonal antibody S309, known for its broad neutralizing capacity against sarbecoviruses, and with the human receptor ACE285. SARS-CoV-2 enters cells by utilizing the ACE2 receptor86.

ALA is widely recognized for its potent antioxidant action, acting both endogenously and exogenously. Its active metabolite, dihydrolipoic acid, is the reduced form of ALA. This compound plays a key role in the Krebs cycle, serving as a cofactor for essential mitochondrial enzymes87. Dietary supplementation with ALA is considered safe, and its antioxidant and immunomodulatory properties have been extensively investigated88. Hydrogen bonds play a crucial role in catalysis, especially in asymmetric catalysis, as discussed by89. Organic chemists have identified that small molecules containing hydrogen bond donors, such as the functional groups OH, NH, and SH, are effective in catalyzing various reactions that involve the formation of bonds between heteroatoms and carbons, including CC and C bonds90.

Signal peptides (SPs) are short N-terminal amino acid sequences that direct proteins to the secretory pathway (Sec) in eukaryotes and facilitate translocation across the plasma membrane (internal) in prokaryotes50. Since comprehensive experimental identification of SPs is impractical, computational prediction of SPs is of significant importance in cellular biology research91. The signal peptidase complex (SPC) cleaves SPs from their non-functional precursor forms and also aids in the maturation of many viral proteins, including the precursor proteins of most flaviviruses (e.g., Zika, Dengue, and Hepatitis C viruses), HIV, and SARS-CoV92,93,94,95,96. The highest values were observed in the C-terminal region of the protein, where the probability for Sec/SPII is significantly high at 0.9993, while the probability for Sec/SPI is very low (0.0002). Additionally, the probabilities for Tat/SPI, Tat/SPII, and Sec/SPIII are nearly zero.

Ebselen is a well-known covalent inhibitor of SARS-CoV-2 3Clpro, however, the inhibitory activity of myricetin has been reported to surpass that of Ebselen, a positive control inhibitor. Ebselen exhibits inhibitory effects at a concentration of 10 µM97,98. In comparison, the Ginkgo biloba extract (GBE50) demonstrated a more potent anti-SARS-CoV-2 3CLpro effect, with an IC50 value of 17.19 µg/mL after 63 min of pre-incubation. Notably, 18 components in GBE50 were capable of covalently modifying SARS-CoV-2 3CLpro. Among these, 11 ingredients were identified as strong to moderate inhibitors in anti-SARS-CoV-2 3CLpro assays99,100,101,102.

The biological effects of Ebselen are related to its antioxidant potential and the interaction with cysteine residues in proteins through selenyl-sulfide. As a result, ebselen binds to viral proteins and makes it difficult for the virus to replicate. Among the studies with inflammation, addressing inflammation in the lungs, Ebselen being transported by blood plasma, interacts with cysteine in the serum albumin protein, concentrating in the pulmonary system and presenting an anti-inflammatory action103,104,105,106.

Ebselen formed molecular interactions with VAL A:93, VAL B:86, PRO B:41, GLY B:42, GLN B:43, and LYS B:40. Similarly, α-lipoic acid demonstrated activity at the binding sites LYS B:40, GLY B:42, GLN B:43, TYR A:95, PRO B:41, and VAL B:86. These results reveal overlapping interaction sites between Ebselen and α-lipoic acid (VAL B:86, PRO B:41, GLY B:42, GLN B:43, and LYS B:40), suggesting the potential for comparable biological effects. However, further in vitro assays are necessary to elucidate and confirm the specific activities of α-lipoic acid within these contexts.

None of the binding sites observed for ALA and ebselen involved N-glycosylation, a common post-translational modification characterized by the covalent addition of oligosaccharides to asparagine residues in polypeptide chains. N-glycosylation can be summarized in three stages: first, the formation of the lipid-linked oligosaccharide donor (LLO); second, the cotranslational transfer of the glycan to the nascent polypeptide chain, in the Asn-X-Ser/Thr pattern (where X represents any amino acid except Pro); and finally, the processing of the oligosaccharide chain Glc₃Man₉GlcNAc₂ in the endoplasmic reticulum (ER) and Golgi107,108,109,110.

Conclusion

The results suggest that ALA exhibits a favorable pharmacokinetic profile, characterized by effective gastrointestinal absorption, though with limited skin penetration and blood-brain barrier permeability. Its compliance with Lipinski’s rule of five and absence of cytochrome P450 inhibition underscore its therapeutic potential. However, the findings also note warnings from Brenk and the compound’s limited similarity to established safe compounds. Molecular docking studies demonstrate ALA’s selective interaction with various functional groups and biological targets, providing strong evidence for its antioxidant and enzyme-modulating properties. Future research should focus on exploring ALA’s clinical applications, particularly in its antioxidant effects and enzyme modulation.

Fig. 1
figure 1figure 1

In silico analysis of toxicity parameters: Acute Inhalation Toxicity (A), Acute Oral Toxicity (B), Acute Dermal Toxicity (C), Eye Irritation and Corrosion (D), Skin Sensitization (E), and Skin Irritation and Corrosion (F).

Full size image
Fig. 2
figure 2

Lipoic acid metabolic reactions (SMILES: OC(=O)CCCCC1CCSS1) - (A) Epoxidation, (B) Chinonation, (C) N-dealkylation, (D) UGT conjugation, (E) Reactivity and (F) phase 1.

Full size image
Fig. 3
figure 3figure 3

Biological structures in the species Homo sapiens that interact with lipoic acid.

Full size image
Fig. 4
figure 4

Biological structures in the species Mus musculus that interact with lipoic acid.

Full size image
Fig. 5
figure 5

Biological structures in the species Rattus norvegicus that interact with lipoic acid.

Full size image
Fig. 6
figure 6

Prediction of signal peptides and their cleavage sites across domains of life, 7TLY protein A Chain (A), 7TLY protein B Chain (B), and 7TLY protein C Chain (C).

Full size image
Fig. 7
figure 7

N-linked glycosylation sites, 7TLY protein A Chain (A), 7TLY protein B Chain (B), and 7TLY protein C Chain (C).

Full size image
Fig. 8
figure 8

In silico molecular docking analysis with the target protein, crystal structure of the SARS-CoV-2 S synthase B.1.1.529 Omicron variant (7TLY) and the Ebselen ligand, including analysis of hydrogen bonds (H-Bonds) (A), solvent-accessible surface area (SAS) (B), aromatics (C), interpolated charges (D), hydrophobicity (E), ionizability (F), and 2D binding analysis with the target protein (G).

Full size image
Fig. 9
figure 9

In silico molecular docking analysis with the target protein, crystal structure of the SARS-CoV-2 S synthase B.1.1.529 Omicron variant (7TLY) and the lipoic acid ligand, including analysis of hydrogen bonds (H-Bonds) (A), solvent-accessible surface area (SAS) (B), aromatics (C), interpolated charges (D), hydrophobicity (E), ionizability (F), and 2D binding analysis with the target protein (G).

Full size image