IN SILICO EVALUATION OF THE INHIBITORY POWER OF AMIDINOUREA DERIVATIVES FOR PHOSPHOPANTETHEINYL TRANSFERASE IN MYCOBACTERIUM TUBERCULOSIS

Kouakou Jean-Louis Kouakou1, Konan Auriol Kra1image, Elvice Akori Esmell*image,

Mélalie Kéïtaépouse Guego1image, Eugene Megnassan1,2,3image

1Laboratory of Fundamental and Applied Physics, University of Abobo-Adjamé (Now Nangui Abrogoua), Côte d’Ivoire.

2Laboratory of Constitution and Reaction of Matter, University of Cocody (Now Felix Houphouët Boigny), Côte d’Ivoire.

3International Centre for Theoretical Physics (ICTP), Strada Costiera, Trieste, Italy.

 

Abstract

Background: Tuberculosis (TB) is an infectious, contagious, and fatal disease, according to the 2025 Global Tuberculosis Report of the World Health Organization (WHO). Phosphopantetheinyl transferase (PptT) is an enzyme that acts as a catalyst in the biosynthesis of cellular lipids in Mycobacterium tuberculosis. It is considered potential target for the development of anti-tuberculosis drugs.

Methods: We used molecular docking to establish a quantitative structure-activity relationship (QSAR) model for 25 amidinourea derivatives inhibitors of PptT. The binding free energies, a key physical descriptor, enabled us to elucidate the thermodynamic factors governing the interaction between the enzyme and the inhibitor. This approach, which combines structure-based molecular docking with quantitative modeling, identified specific, critical chemical substituents. This, in turn, led to the development of new derivatives that exhibit potent antitubercular activity. 

Results: The resulting QSAR model that accounts for 87% of the variation in biological activity, based on the binding free energies (BFE/PBSA) of the analogues used. Subsequently, an intuitive fragment-replacement approach, based on per-residue interaction energy, which indicates stabilizing interactions, was implemented to identify new analogues. Moreover, 200-ns molecular dynamics simulations  have evaluated with satisfaction, the temporal evolution of various properties of the bound inhibitors, such as root mean square deviation (RMSD) from the initial conformation, radius of gyration (rGyr), number of intramolecular hydrogen bonds (intraHB), molecular surface area (molSA), solvent-accessible surface area (SASA), and polar surface area (PSA). The BFE/GBSA of the three best analogs, calculated over the last 50 nanoseconds of the trajectory, align with BFE/PBSA values. Finally, favorable values for the pharmacokinetic descriptors of the best analogues are obtained using the QikProp, pkCSM, and mycoCSM programs.

Conclusion: This study enabled us to identify a new amidinourea analogue AU_1-3-7-9,  image which is 10 times more active.

Keywords: Mycobacterium tuberculosis, Phosphopantetheinyl transferase, molecular docking, amidinourea, molecular dynamics, QSAR.

 

 

INTRODUCTION

 

Tuberculosis (TB) is a chronic and contagious infectious disease caused by the Mycobacterium tuberculosis complex (Koch's bacillus)1. It generally affects the lungs; hence, the clinical form known as active pulmonary tuberculosis (APT), but it can also affect other parts of the body, in which case it is referred to as extra-pulmonary tuberculosis2. TB remains a persistent global health challenge. According to the 2025 Global Tuberculosis Report, the World Health Organization (WHO) estimated approximately 10.7 million cases and 1.23 million deaths in 20243. While significant progress has been made in diagnostic techniques and healthcare accessibility, the emergence of multidrug-resistant tuberculosis (MDR-TB) continues to threaten international recovery efforts. In 2024 alone, an estimated 410,000 individuals contracted MDR-TB, highlighting a critical need for the discovery of novel therapeutic agents with distinct mechanisms of action3,4. To address this threat, recent research has pivoted toward newly validated biological targets. One such promising target is the protein phosphopantetheinyl transferase (PptT), which plays an essential role in the biosynthesis of virulence-critical lipids required for the survival of Mycobacterium tuberculosis5. Inhibiting PptT offers a strategic pathway to disrupt the pathogen's metabolic integrity. In this study, we explore the in silico inhibition of PptT using a series of amidinourea derivatives. Our objective is to establish a univariate QSAR model that correlates the binding free energies (BFE), derived from molecular docking simulations, with the observed experimental activities (the biological response). By utilizing BFE as a primary physical descriptor, we aim to elucidate the thermodynamic factors driving the formation of the enzyme-inhibitor complex. Furthermore, this dual approach, integrating structure-based docking with quantitative modeling, allows us to pinpoint specific chemical substituent at key molecular positions, providing a predictive framework for designing novel derivatives with enhanced anti-tuberculosis potency.

 

METHODS

 

Dataset selection and preparation

The chemical dataset for this study consisted of 25 amidinourea derivatives with known experimental biological activitiesimageagainst PptT of M. tuberculosis, as reported by Ottavi et al.5. The spatial geometry of the structures of these ligands was modeled and energy-minimized within the Discovery Studio® (BIOVIA) environment6. For the biological target, the crystal structure of phosphopantetheinyl transferase (PptT) in complex with the native ligand Au8918 which occupies the active site as previously described23,24, was retrieved from the Protein Data Bank (PDB ID: 6CT5)7. The protein was prepared using the “Prepare Protein” protocol of Discovery Studio®, which involved adding missing hydrogen atoms, assigning proper ionization states to amino acid residues at pH 7.0, and removing crystallographic water molecules.

Molecular Docking using CDOCKER

Molecular docking simulations were performed using the CDOCKER algorithm, a grid-based method grounded in the CHARMm force field6. Water molecules are removed because the docking takes place within the gas. The docking sphere (radius = 10 Å) was centered on the coordinates of the co-crystallized inhibitor within the PptT binding pocket. For each ligand, a series of random conformations were generated and refined through grid-based simulated annealing followed by a final energy minimization. The top-ranked pose for each derivative, characterized by the lowest CDOCKER energy, was selected for subsequent thermodynamic analysis.

Poses selection

For each of the 25 amidinourea derivatives, the CDOCKER algorithm generated a pool of 10 initial docking poses through high-temperature molecular dynamics and subsequent simulated annealing. To select the most energetically relevant orientation for each ligand, poses were initially ranked based on their CDOCKER Energy, which represents the sum of the internal ligand strain and the protein-ligand interaction energy. Following this energy scoring, the top-ranked poses were visually inspected for interaction consistency to ensure their scaffold aligned closely with that of the co-crystallized ligand and that they maintained key interactions observed in the crystal structure, particularly the hydrogen bonding network with the catalytic residue Glu157.

Each protein-ligand complex containing the single best-scoring pose for each ligand that satisfied these structural constraints was then subjected to the “Calculate binding energy” protocol for final binding free energy (∆Gbind) calculation including the loss of conformational entropy of the bound ligand at 298.15K.

Binding Free Energy Calculations with PBSA

To achieve a more accurate physical representation of the binding process, the binding free energies ∆Gbind (Eq. 1) calculated for the best poses obtained from docking, using the “Calculate binding energy” protocol; optimal coverage of the conformational space was ensured, and the Poisson-Boltzmann/surface area (PBSA) method—an implicit solvation model—was applied. The total binding free energy was determined according to the following equation:

           ∆Gbind = ∆Ebind + ∆Gsolv - T∆Sconf          Eq. 1

          ∆Ebind = Ecomplex - Ereceptor - Eligand            Eq. 2

Where:

Furthermore, to mini mize systematic errors and highlight the impact of structural modifications, the relative binding free energy ∆∆Gbind (Eq. 3) was calculated for each derivative Li compared to the most active compound used as a reference Lref:

   ∆∆Gbind = ∆Gbind(Li) - ∆Gbind Lref

                =∆∆Ebind + ∆∆Gsolv - T∆∆Sconf            Eq. 3

Univariate QSAR model development and statistical validation: The QSAR model was constructed and rigorously validated using QSARINS software version 2.2.4-2019, which specifically implements the validation protocols required by the OECD (Organization for Economic Cooperation and Development)8-14.

A univariate linear regression was established using the relative binding free energy ∆∆Gbind as the descriptor to predict the experimental biological activity image as the response. The dataset (n=25) was split into a training set (n=20) and a validation set (n=5). To ensure the model's robustness, internal consistency, and external predictivity, the following statistical metrics were calculated:

Per-residue interaction energy decomposition

To identify the individual amino acids driving the molecular recognition process, per-residue interaction energy decomposition was performed. This analysis calculated the non-bonded interaction energy (Eint) between each active conformation of the ligands and the individual residues within the PptT binding pocket.

The total interaction energy for each residue (i) was determined by the sum of its van der Waals (Evdw) and electrostatic (Eelec) components:

                 Eint(i) = Evdw(i) + Eelect(i)                   Eq. 4

This quantitative assessment allows for the identification of “hot-spot” residues that contribute most significantly to the complex stability. By pinpointing these critical residues, we can determine which specific molecular interactions such as hydrogen bonding, π-stacking, or hydrophobic contacts must be preserved or optimized in the design of next-generation amidinourea derivatives to enhance PptT affinity.

Molecular Dynamics Simulations (MDS)

The structural stability of the most active ligand and of new analogues, built from the native ligand, was evaluated by 200 ns molecular dynamics (MD) simulations, using the OPLS_2005 force field available in the Desmond module of Schrodinger 2018-4 software15,16, as described recently17.

MM-GBSA binding energy

After molecular dynamics, the average of the changes in binding free energy during complex formation was calculated over the all 200 nanoseconds and the last 50 nanoseconds of the trajectory using the molecular mechanics/generalized born surface (MM-GBSA) method18. The theory and method of MM-GBSA have been described in detail previously19.

The change in binding free energy for each analogue ∆GMM-GBSA was computed as:

        ∆Gbind = ∆EMM + ∆Gsolv + ∆GSA      (Eq. 5)

Where ∆EMM represents molecular mechanics energy changes, ∆Gsolv accounts for solvation effects, and ∆GSA represents surface area contributions.

Pharmacokinetic profile evaluation, and predictive description of the synthesis.

We employed machine learning to evaluate pharmacokinetic profiles. Tools such as pkCSM and mycoCSM, along with Schrödinger’s QikProp program, were used to predict pharmacokinetic properties20,21.

 

RESULTS

 

Dataset Partitioning

To ensure the robustness and external predictive power of the QSAR model, the 25 ligands were divided into training and validation subsets using a standard 80/20 ratio. This division was performed using the “Generate Training and Test Data” protocol within Discovery Studio®6, where twenty (20) ligands (Table 1) were assigned to the training set to build the QSAR model, and five (05) ligands (Table 2) were reserved as an external validation set to independently verify its predictive accuracy. 

Validation of the Molecular Docking Protocol

To confirm the reliability of the CDOCKER algorithm and the CHARMm force field in the context of the PptT binding pocket, a self-docking (redocking) study was performed. The co-crystallized ligand was extracted from the 6CT5 PDB structure and redocked into the active site using the parameters defined in the methodology. 

The predicted binding pose demonstrated a good level of structural alignment with the experimental crystal orientation, yielding a Root Mean Square Deviation (RMSD) of 1.6 Å as shown in Figure 1. This value is well within the widely accepted 2.0 Å threshold for successful docking validation15. 

Molecular docking

The poses of ligands that have the same conformation after rigid docking as that of the endogenous ligand of the active site are selected (Figure 2). The different energies mentioned below are determined for these poses in complex with the enzyme. 

Statistical Quality and Robustness of the QSAR Model

The QSAR model (Eq. 5), developed using the relative binding free energy (∆∆Gbind) as the sole thermodynamic descriptor, demonstrated excellent statistical quality that complies with OECD rules. The resulting regression equation is:

      image = -0.1417*∆∆Gbind + 6.5117       Eq.6

The model was constructed using a training set of 20 ligands. The experimental activities covered a broad biological range, with image values spanning from 210 nM to 51,400 nM. We chose 5 g as a reference because it possesses greater biological activity. Furthermore, the activities are expressed in nanomolar (nM) units to facilitate better comparison, as our goal is to identify molecules with higher activity—effectively shifting from the micromolar to the nanomolar range.

The ratio   imageshows that the model is good for estimating the activities of new molecules.

Internal robustness and fitting

The model showed a strong correlation with a coefficient of determination of 0.87 and a high Fisher F-test value of 125.79, indicating a highly significant linear relationship. Its internal stability was confirmed through rigorous cross-validation, yielding a Leave-One-Out value image of 0.85 and a Leave-Many-Out image value of 0.84. Furthermore, there is a high degree of proximity between the error metrics—specifically between the root-mean-square error RMSE of 0.19 and the root-mean-square error of cross-validation RMSECV of 0.21, as well as the corresponding mean absolute errors (MAE) of 0.17 and MAECV 0.18

External predictivity

The predictive power of the model was successfully verified using an external validation set (n=5). The external squared correlation coefficient image was 0.87, matching the training set's R2. 

Y-Randomization (Y-scrambling)

To further validate the statistical significance of the model, a Y-randomization test (2000 iterations of shuffling) was performed. The resulting R2yscr was extremely low (0.05), while the R2p value of 0.84, being  significantly higher than the recognized 0.5 threshold12,16. 

Applicability domain analysis

To ensure the reliability of the model's predictions, the structural applicability domain (AD) was evaluated using the leverage approach. The warning leverage threshold (h*) was calculated as 0.30, based on the training set size (n=20) and the single thermodynamic descriptor (p=1). A comprehensive analysis of the individual leverage values (h) revealed that all 25 amidinourea derivatives, including those in the external validation set, remained strictly below this threshold (see Supporting information). Furthermore, no response outliers were identified, as all standardized residuals fell within the ±3 standard deviation limits.

Correlation Analysis of the QSAR Model

The relationship between the predicted thermodynamic descriptors and the experimental anti-tuberculosis activity is illustrated in Figure 3. The plot of experimental image versus the calculated relative binding free energy ∆∆Gbind reveals a robust linear correlation for both the training set (blue circles) and the validation set (orange circles).

Per-residue interaction energy analysis

To gain a deeper understanding of the molecular basis of PptT inhibition, the total interaction energy was decomposed into individual residue contributions for the six representative compounds spanning the activity range (Table 6), from the most potent inhibitor 5g (image=210 nM) to the least active inhibitor 4e (image=51400 nM). This analysis identifies the “hot-spot” residues that dictate the binding affinity and stability of the complexes.

Structural basis of differential biological activities

The variations in biological activity among the amidinourea derivatives can be directly attributed to the nature and strength of the interactions established within the PptT active site. The analysis of inhibitors 5g (image=210 nM), 5n (image=3800 nM), and 4e (image=51400 nM) clearly demonstrates this. 

 

The hydrophobic canal and stability

The presence of the hydrophobic canal, a critical feature identified in previous PptT studies23,24, is clearly visualized through the extensive network of Van der Waals interactions in all complexes. Substituents in the para position on benzene, particularly aromatic ones, sometimes exhibit a stabilizing cation-π interaction with lysine 75, thus increasing ligand activity. Among these, although compound 5g is the most active, its ester function makes it relatively unstable during administration in humans, despite its superior in vitro affinity.  

Key residue contributions to affinity

A comparative analysis of the quantitative energy contributions (Table 6) identifies the residues where enzyme-ligand affinity is maximized.

Strategic implications for affinity improvement

The primary challenge in developing more potent anti-tuberculosis compounds lies in the ability to strategically maintain or intensify these specific structural interactions within the active site.

Docking and predicted activities of new analogues

To discover novel and more potent inhibitors against the Mycobacterium tuberculosis PptT enzyme, an intuitive fragment substitution approach was conducted. Using the amidinourea core scaffold as a template, various chemical modifications were introduced across four positions (R1, R2, R3, and R4) utilizing a specialized structural library (Table 7). Docking of the new analogues is performed using the same parameters. Next, the free enthalpies of complex formation are calculated and their predicted inhibitory concentrations image are deduced from the correlation equation of the model, (Eq.6: image=-0.1417*∆∆Gbind+6.5117). The newly designed analogues were subjected to molecular docking simulations using the exact parameters validated in our previous protocol. Following pose generation, the binding free energies were computed, and their predicted median inhibitory concentrations image were derived directly using the validated univariate correlation equation established in this work (Equation 6): image= -0.1417*∆∆Gbind + 6.5117. The calculated thermodynamic parameters, conformational entropy penalties ∆∆Sconf, net binding free energies ∆∆Gbind, and corresponding predicted biological activities for the top four design candidates are summarized in Table 8. Compound 5g is included as the high-potency reference standard from the experimental training set.

Interaction energies per residue of new amidinourea analogues

To uncover the physical forces driving this enhanced affinity, comparative per-residue interaction energy decomposition ∆Eint was carried out for the designed analogues and the reference ligand 5g (Table 9).

Molecular dynamics

To evaluate the structural stability of the studied complexes, 200-ns molecular dynamics simulations were performed for the most active ligand from the training set (5g) and the top three analogues.

Pharmacokinetic profile evaluations

The pharmacokinetic descriptors for the best analogues as well as those for drugs currently used to treat tuberculosis or undergoing clinical trials—are calculated using the QikProp and pkCSM programs.

MM-GBSA binding free energy

The mean MM-GBSA binding energy for the best compound in the training set (5g) and for the three best analogues AU_1-3-7-9, AU_1-7-6-11, and AU_1-6-2-10 was calculated over the last 50 nanoseconds of the trajectory (Table 10).

 

DISCUSSIONS

 

Validation of the Molecular Docking Protocol

The values confirm that the CDOCKER protocol effectively reproduced the native intermolecular interactions within the PptT binding pocket. Consequently, the validated protocol was deemed robust enough to explore the binding orientations and thermodynamic profiles of the 25 amidinourea derivatives.

Molecular docking, statistical quality and robustness of the QSAR model

We note that the model does not exhibit overfitting and possesses a high degree of internal consistency. Regarding external validation, the model’s ability to accurately predict the antitubercular activity of new amidinourea derivatives is demonstrated. The Y-randomization test confirms that the correlation between the PptT binding free energy and anti-tuberculosis activity is not due to chance. The structural applicability domain (AD) using the leverage approach results confirm that all compounds reside within the model's domain of validity, ensuring that the predicted anti-tuberculosis activities are structurally and statistically good.

Per-residue interaction energy analysis

The energy profiles, illustrated in Figure 4, reveal that the binding stabilization is primarily localized within a specific cluster of residues and the co-crystallized COA301 cofactor.

Structural basis of differential biological activities

The analysis highlights three representative types of interactions: hydrogen bonds, van der Waals interactions, and electrostatic interactions. The strong electrostatic interaction between Glu157 and 5g is a primary driver of potency. As seen in the 2D maps (Figure 5), 5g establishes a salt bridge and pi-cation interaction, whereas less active molecules like 4e only maintain a single, weaker hydrogen bond.  Tyr160 and Phe173 form the structural framework of the pocket, providing the necessary hydrophobic anchoring for the aromatic scaffolds. Cofactor COA301 interactions also matter. Consistent with the energy decomposition, the most active molecules maintain high-affinity contact with the Coenzyme A cofactor, a relationship that diminishes significantly in the least active compound, 4e. The design process for the new analogues prioritized preserving the critical electrostatic network established with the Glu157, while simultaneously maximizing the van der Waals fit within the characteristic hydrophobic canal. Furthermore, maintaining the stabilizing interactions with key residues Tyr160 and Phe173, as well as the COA301 cofactor, is essential for preserving structural alignment. Finally, investigating targeted molecular modifications at the para position of the phenyl ring presents a promising pathway to introduce or intensify stabilizing interactions, which could simultaneously improve both target affinity and metabolic stability. 

Docking and predicted activities of new analogues

The systematic substitutions at the R1 to R4 positions led to a distinct increase in target binding affinity compared to the parent training series. Most notably, analogue AU_1-3-7-9 (Figure 6) exhibited a remarkably favorable net free energy of binding (∆∆Gbind=-8.95 kcal.mol-1), translating to an outstanding predicted potency of image=20 nM. This marks a 10-fold increase in anti-tuberculosis activity over the best experimental compound 5g image=210 nM. 

Interaction Energies per Residue of New Amidinourea Analogues

The predicted biological activity of the best AU_1-3-7-9 analogue can be explained by a significant increase in hydrophobic interactions at positions R1 and R4 with the Lys78 residue. All of these interactions observed with the residues justify the activity of the new analogues. The new analogues interacts with several residues in the active site of PptT. These residues play an essential role in enhancing the inhibitory activity of the new amidinourea analogues. The diagram shows that the best analogue retains its affinity for Phe173 and has increased affinity for Glu157, Tyr160, Trp170, and COA301. Furthermore, this new analogue establishes a stabilizing interaction with Lys78. Consequently, interactions with the enzyme are more favorable for the best analog than for the 5g ligand.

Molecular dynamics

The temporal evolution of various properties of the bound inhibitors, such as root mean square deviation (RMSD) from the initial conformation, radius of gyration (rGyr), number of intramolecular hydrogen bonds (intraHB), molecular surface area (molSA), solvent-accessible surface area (SASA), and polar surface area (PSA), is satisfactory. Examination of the root-mean-square deviations (RMSD) highlights good structural convergence. The conformational equilibrium of the systems is not affected. The [system/complex] maintains the radius of gyration (rGyr) an invariant value at 5.1 Å, reflecting high rigidity and dynamic stability.  

The display rGyr values of analogues confirm the preservation of a folded and compact conformation. 

Pharmacokinetic profile evaluations

The values given by QikProp and pkCSM programs remain within the prescribed limits. For mycoCSM program, we have a high Maximum Recommended Tolerated Doses (MRTD) because it is greater than 0.477 log (mg/kg/day).

MM-GBSA Binding Free Energy

The negative values obtained during enzyme-ligand complex formation indicate that the interactions occur spontaneously. These values, which remain lower for the new analogues, compared to that of the 5g ligand, denote a higher affinity between the new inhibitors and the target protein. These values are also compared to the relative free energy of binding (PBSA). The biological activities fall within the same order of magnitude as each of these two thermodynamic quantities. This convergence confirms the reliability of the computational approaches and allows us to propose these designed analogues as potential inhibitors25.

Limitation of the study

We note here that intuitive substitution has limited the potential for generating a large number of analogues and, consequently, compounds with the prospect of improved activity. Furthermore, our computational resources do not allow for molecular dynamics simulations exceeding 200 ns. 

 

CONCLUSIONS 

 

This study utilized molecular docking, binding free energy decomposition, and Quantitative Structure-Activity Relationship (QSAR) workflows to characterize and strategically enhance a series of twenty-five (25) amidinourea derivatives targeting Mycobacterium tuberculosis Phosphopantetheinyl Transferase (PptT). The constructed univariate model demonstrates impressive predictive strength, with its physics-based relative free energy of complex formation ∆∆Gbind accounting for 87% of the total variation in experimental biological activity. Deep analysis of the receptor-ligand complex networks shows that structural stability is primarily governed by strong hydrogen bonding networks and key non-polar contacts within the enzyme's conserved hydrophobic canal. By applying these insights to a targeted fragment library, we successfully engineered novel sub-series options. The top-designed candidate, analogue AU_1-3-7-9, displayed exceptional active-site complementarity, highlighted by an advanced electrostatic salt-bridge network with Glu157, hydrophobic interactions with Lys75 and Phe164, localized interactions with Lys78, Lys168, and a powerful binding synergy with the embedded COA301 cofactor. These optimizations yielded a highly favorable thermodynamic state (∆∆Gbind=-8.95 kcal.mol-1) and a projected biological potency of 20 nM. This reflects a significant 10-fold improvement over the best experimental hit in the dataset. Other more robust methods involving, for example, Gibbs free enthalpy as a descriptor, combining a virtual chemical library screened by a pharmacophore model could make it possible to obtain PptT inhibitors with a higher power factor.

 

ACKNOWLEDGEMENTS

 

The authors gratefully acknowledge Prof. Paola Gramatica for granting a free license for the QSARINS software.
 

AUTHOR’S CONTRIBUTION

 

Kouakou KJL: conceptualization, writing original draft. Kra KA:  conceptualization, investigations, computing, and editing. Esmel EA: conceptualization, investigations, technical, computational coaching, software, and editing. Kéita M:conceptualization, Intellectual input and final manuscript editing. Megnassan E: Intellectual input and final manuscript editing. Final manuscript was checked and approved by all authors.    

 

DATA AVAILABILITY   

 

Data will be available on request to anyone from the correspondence author.

 

CONFLICT OF INTERESTS         

 

The authors declared no conflict of interests

 

REFERENCES

 

  1. Rastogi N, Legrand E, Sola C. The mycobacteria: An introduction to nomenclature and pathogenesis. Rev Sci Tech 2001; 20: 21-54. https://doi.org/10.20506/rst.20.1.1265
  2. Koloma I. Role of biological sex in the immune response against the Mycobacterium tuberculosis complex (Doctoral dissertation, USTTB). 2021
  3. Tuberculosis Global Report 2025.
  4. Germain MA, Latrille CFR, Philippon A, et al. Lymph: From Antiquity to the 21st L'Harmattan Publishers; 2023.
  5. Ottavi S, Scarry SM, Mosior J, et al. In vitro and in vivo inhibition of the mycobacterium tuberculosis phosphor pante theinyl transferase by amidinoureas. J Med Chem 2022; 65: 1996-2022. https://doi.org/10.1021/acs.jmedchem.1c01565
  6. Discovery Studio molecular modeling and simulation program n.d.
  7. Ballinger E, Mosior J, Hartman T, et al. Opposing reactions in coenzyme A metabolism sensitize Mycobacterium tuberculosis to enzyme inhibition. Science (1979) 2019; 363: eaau8959. https://doi.org/10.1126/science.aau8959
  8. Gramatica P, Cassani S, Roy PP, et al. QSAR modeling is not push a button and find a correlation: a case study of toxicity of (benzo-)triazoles on algae. Mol Inform 2012; 31: 817-35.https://doi.org/https://doi.org/10.1002/minf.201200075
  1. Gramatica P, Chirico N, Papa E, et al. QSARINS: A new software for the development, analysis, and validation of QSAR MLR models. J Comput Chem 2013; 34: 2121-32. https://doi.org/https://doi.org/10.1002/jcc.23361
  2. Gramatica P, Cassani S, Chirico N. QSARINS-chem: Insubria datasets and new QSAR/QSPR models for environmental pollutants in QSARINS. J Comput Chem 2014; 35: 1036-44. https://doi.org/https://doi.org/10.1002/jcc.23576
  3. Gramatica P, Sangion A. A historical excursus on the statistical validation parameters for QSAR models: A clarification concerning metrics and terminology. J Chem Inf Model  2016; 56: 1127-31.https://doi.org/10.1021/acs.jcim.6b00088
  1. Gramatica P. Principles of QSAR Modeling: Comments and suggestions from personal experience. Int J Quantitative Structure-Property Relationships 2020; 5: 61-97.https://doi.org/10.4018/IJQSPR.20200701.oa1
  1. Gramatica P. Origin of the OECD Principles for QSAR validation and their role in changing the QSAR paradigm worldwide: An Historical Overview. J Chemom 2025; 39: e70014. https://doi.org/https://doi.org/10.1002/cem.70014.
  2. Chirico N, Sangion A, Gramatica P, et al. QSARINS-Chem standalone version: A new platform-independent software to profile chemicals for physicochemical properties, fate, and toxicity 2021; 42. John Wiley & Sons, Ltd.https://doi.org/https://doi.org/10.1002/jcc.26551
  1. Schrödinger LLC, D. E. Shaw Research, Desmond Molecular Dynamics System 2018.
  2. De Vivo M, Masetti M, Bottegoni G. Role of molecular dynamics and related methods in drug discovery. J Med Chem 2021; 59(9): 4035-4061.https://doi.org/10.1021/acs.jmedchem.5b01684
  1. Frecer V, Miertus S. Antiviral agents against COVID-19: Structure-based design of specific peptidomimetic inhibitors of SARS-CoV-2 main protease. RSC Adv 10 (2020) 40244-40263. https://doi.org/10.1039/D0RA08304F
  1. T Hou, J Wang, Y Li, et al. Assessing the performance of the MM/PBSA and MM/GBSA Methods. The accuracy of binding free energy calculations based on molecular dynamics simulations. J Chem Inf Model 2011; 51: 69-82.https://doi.org/10.1021/ci100275a
  1. Ntie-Kang F, Kannan S, Wichapong K, et al, Binding of pyrazole-based inhibitors to Mycobacterium tuberculosis pantothenatesynthetase: Docking and MM-GB(PB)SA analysis. Mol Biosyst 2014; 10: 223-239.https://doi.org/10.1039/C3MB70449A
  1. Pires DE, Blundell TL, Ascher DB. PkCSM: Predicting small-molecule pharmacokinetic properties using graph-based signatures (Theory-How to EnterpretPkCSM Result) 2015.https://doi.org/10.1021/acs.jmedchem.5b00104
  1. Pires DE, Ascher DB. MycoCSM: using graph-based signatures to identify safe potent hits against mycobacteria. J Chem Inform Model 2020; 60(7): 3450-3456.https://doi.org/10.1021/acs.jcim.0c00362
  1. Leblanc C, Prudhomme T, Tabouret G, et al. 4′-Phosphopantetheinyl Transferase PptT, a new drug target required for Mycobacterium tuberculosis growth and persistence in vivo. PLoS Pathog 2012, 8, No. e1003097. https://doi.org/10.1371/journal.ppat.1003097
  1. Vickery CR, Kosa NM, Casavant EP, et al. Structure, biochemistry, and inhibition of essential 4′-phosphopantetheinyl transferases from two species of mycobacteria. ACS Chem Biol 2014; 9: 1939−1944.https://doi.org/10.1021/cb500263p
  1. Jung J, Bashiri G, Johnston JM, et al. Crystal structure of the essential Mycobacterium tuberculosis Phosphopantetheinyl Transferase PptT, solved as a fusion protein with maltose binding protein. J Struct Biol 2014; 188: 274−278.https://doi.org/10.1016/j.jsb.2014.10.004
  1. Yavo BU,  Esmel  AE,  Kouakou  KJL,  Kéita  M, Eugène      Molecular    modeling    and virtual screening   for computer-aided   design of HCV NS5B   polymerase   inhibitors.  Universal   J Pharm Res 2026; 11(3): 43-57.  http://doi.org/10.22270/ujpr.v11i3.1586