Structure-based virtual screening and molecular dynamics elucidate novel inhibitors of HIV-1 TAR RNA from enamine library

Abstract

The HIV-1 trans-activation response (TAR) RNA is a key regulatory element necessary for viral replication, presenting a high-value target for anti-HIV therapeutics. Recent advances in RNA-targeted drug discovery have leveraged large, diverse compound libraries, such as the Enamine collection, using structure-based virtual screening to uncover novel modulators of RNA structure and function. In this study, we employed an integrative in silico approach combining high-throughput virtual screening of the Enamine library against the central bulge region of HIV-1 TAR RNA with molecular dynamics (MD) simulations to assess binding stability and conformational adaptability of lead molecules. Virtual screening pinpointed multiple candidate ligands predicted to disrupt the critical Tat-TAR interaction interface, with subsequent MD simulations over 500-nanosecond timescales revealing ligand-induced stabilization of druggable RNA conformations and dynamic remodelling of key binding pockets. Notably, some top-scoring compounds featured strong and persistent contacts within the U23-C25 bulge and the adjacent major groove, recapitulating pharmacologically relevant mechanistic disruption. These findings reinforce the synergy of structure-based screening and MD for identifying and validating RNA-targeted small molecules and provide promising new scaffolds for further experimental testing as selective HIV-1 replication inhibitors.

Introduction

The HIV-1 trans-activation response (TAR) RNA is a highly structured, critical, and dynamic regulatory element located at the 5′ end of all nascent viral transcripts^1,2. It serves as the primary binding site for the viral trans-activator protein Tat, which, in concert with numerous host cellular factors, is indispensable for transcriptional elongation and robust HIV-1 replication^2,3. Binding of Tat to TAR relieves transcriptional pausing, confers processivity to the viral RNA polymerase, and amplifies gene expression, altogether forming a pivotal regulatory axis essential to establishing and maintaining productive HIV-1 infection^2,4. Perturbation of the Tat-TAR axis not only suppresses viral replication but influences viral latency, making disruption of this interaction a long-standing and promising therapeutic strategy for tackling both active and latent HIV-1 reservoirs^1,2,5.

Although TAR RNA is a promising therapeutic target, its flexible structure and high conformational entropy pose significant challenges for small-molecule design. Unlike proteins with deep hydrophobic pockets, TAR features shallow solvent-exposed grooves that hinder conventional structure-based drug discovery^6,7. Yet, recent high-resolution crystallographic and NMR studies have revealed that HIV-1 TAR is far from static; its secondary and tertiary architectures comprise a complex ensemble of rapidly interconverting conformations, particularly within its hallmark UCU and three-nucleotide bulges as well as its lower and upper stem-loop regions. These conformationally plastic domains have emerged as “hot spots” for molecular recognition, allosteric regulation, and small-molecule binding, opening new avenues for drug development^5,8,9.

Importantly, studies on full-length TAR, as opposed to truncated mini-TAR constructs, demonstrate that the conformational landscape of the complete RNA is further modulated by structural coupling between the lower and upper stem regions, dynamic helical stacking, and interactions with neighbouring elements in the 5′ UTR. The flexible and unstable regions of TAR RNA are now known to be directly involved in Tat-driven transcription and in helping the virus evade innate immune defences like Protein Kinase R (PKR) activation. This highlights the importance of developing ligands that can specifically bind to and control these shape-changing regions¹⁰.

The ongoing evolution of RNA-targeted chemical libraries, highlighted by focused screening sets like the Enamine library, has ushered in an era where structure-enabled discovery of selective RNA binders is feasible. These specialized libraries, enriched for shape, charge, and hydrogen-bonding motifs compatible with RNA surfaces, have synergized with the latest high-throughput docking, virtual screening, and free energy prediction methodologies. Such platforms permit both broad and focused interrogation of potential ligand-RNA recognition principles under physiologically relevant conditions¹¹.

Virtual screening now enables the rapid and efficient computational triage of thousands of candidate ligands, harnessing RNA-ligand docking to prioritize molecules for experimental validation. Subsequent molecular dynamics (MD) simulations offer atomistic insight into the real-time, nanosecond-to-microsecond scale motions of both TAR and its ligands, capturing effects such as induced fit, allosteric shifting, and dynamic contacts essential for high-affinity, selective binding. Integration of these approaches has proven powerful for illuminating druggable conformational states, mapping cryptic pockets transiently available in the RNA ensemble, and identifying the kinetic and thermodynamic factors that underlie effective Tat-TAR inhibition^5,8,9.

Thus, in the present study, we deploy a comprehensive computational pipeline leveraging structure-based virtual screening alongside extensive MD simulations to systematically interrogate interactions between Enamine library and the dynamic HIV-1 TAR RNA ensemble. Our overarching goal is the discovery and characterization of novel chemotypes that stabilize inhibitory TAR conformations, thereby paving the way for the rational design of new-generation, RNA-targeted antivirals against HIV-1​⁵.

Materials and methods

A structure-based virtual screening workflow was employed to identify novel small-molecule binders of the HIV-1 TAR RNA using the Enamine library, which was accessed on July 23, 2025. The ligands of the library were prepared for docking using LigPrep of the Schrödinger Suite. Molecular docking studies were conducted using the NMR structure of HIV-1 TAR RNA (PDB ID: 2L8H) as the receptor. The initial round of docking involved screening the entire Enamine library against TAR; Grid selection was done based on binding pose and was focused around the highly conserved three-nucleotide bulge (nt 23–25) and the adjacent upper stem/major groove region (Fig. 1A), which is critical for Tat protein interaction and recognized as major hotspots for RNA-targeted modulation^12–14. Compounds were scored according to predicted binding energy and interaction profiles.

Fig. 1.

(A) Secondary structure of TAR RNA, highlighting its key functional elements: the apical loop (magenta), upper stem (blue), three-nucleotide UCU bulge (red), and lower stem (green), with nucleotide numbering indicated. (B) Three-dimensional structure, coloured according to the same scheme, illustrating how the loop, bulge, and helical stems are spatially arranged. This integrated view emphasizes the conformational organization of TAR RNA regions relevant for ligand recognition and Tat-TAR interactions.

From the primary screening, the top 1,000 compounds showing the most favourable binding energies were subjected to a secondary round of docking, employing more stringent scoring and rescoring algorithms to account for both binding affinity and key molecular interactions with the bulge and major groove signatures. The top 100 hits from this refinement were further redocked to verify pose stability and interaction reproducibility. After comparative analysis of binding energies and binding-site engagement, the top 20 compounds were visually inspected for key contacts: optimal candidates formed hydrogen bonds and stacking interactions with bulged nucleotides (notably U23 and C24) (Fig. 1A, B) and neighbouring residues within the major groove, mimicking or disrupting native Tat-TAR recognition motifs¹⁵. Based on these criteria, four representative small molecules were selected for detailed downstream characterization, including those with the most promising occupancy of the bulge/major groove interface and distinct chemotypes.

Target preparation

The HIV-1 TAR RNA structure (PDB ID: 2L8H) was retrieved from the RCSB Protein Data Bank and processed using the Schrödinger suite to ensure a biologically relevant starting conformation for docking studies. The NMR structure comprises an ensemble of multiple conformations reflecting its intrinsic structural flexibility. For the purpose of molecular docking, a single representative model was selected. Specifically, the first model from the NMR ensemble was used, as it is lowest energy structural model satisfying experimental restraints and retains the key structural features of TAR RNA, including the three-nucleotide bulge (U23-C25) and the major groove architecture relevant for Tat binding. While this approach provides a consistent framework for docking, the dynamic nature of TAR RNA was further explored through subsequent molecular dynamics simulations. The preparation workflow included comprehensive optimization, which is crucial for correctly modelling hydrogen bonding patterns and electrostatic interactions within the RNA and between RNA and ligands.

Ligand library preparation

The Enamine library (28,000 molecules) was prepared for docking using Schrödinger’s LigPrep tool. LigPrep efficiently generates high-quality, three-dimensional ligand structures starting from 2D or 3D inputs. During preparation, protonation states of ligands were predicted and assigned corresponding to physiological pH (7.0 ± 1.0) using Epik, ensuring realistic ionization states for docking¹⁶. Possible tautomers, stereoisomers, and ring conformations were enumerated to fully capture relevant chemical diversity and conformational space for each compound. Ligands were then subjected to energy minimization using the OPLS4 force field to optimize geometry and avoid steric clashes¹⁷. This comprehensive preparation ensures the ligand structures are chemically realistic and conformationally relevant for downstream virtual screening.

Molecular Docking of HIV-1 TAR with Enamine library

The processed HIV-1 TAR RNA structure (PDB ID: 2L8H), prepared as described previously, was used as the receptor for molecular docking using Schrödinger’s Glide module^18,19. A receptor grid was defined centering on the highly conserved three-nucleotide bulge (nucleotides 23–25) and the adjacent upper stem region, representing the essential binding hotspot for Tat protein and known to be a critical ligandable pocket for small molecules^15,20. Initial docking was performed in High Throughput Virtual Screening (HTVS) mode to screen the entire Enamine library. Top-scoring 1,000 ligands were subjected to a secondary screening using Glide standard precision (SP) mode to refine pose predictions and binding affinity estimates. Visual inspection and interaction analysis focused on hydrogen bonding, π-stacking, and electrostatic contacts, particularly with key bulge residues and the major groove phosphates. Based on docking scores and binding pattern analysis, four top candidate compounds were selected from the top 20 hits for advanced computational characterization and potential experimental validation.

Binding free energy

To further refine the binding affinities predicted by docking, the top 20 hit compounds from the Glide SP docking stage were subjected to binding free energy analysis¹⁸. The total binding free energy was broken down into key energetic components, including van der Waals interactions, electrostatic contributions, and solvation effects, which collectively describe the physical forces stabilizing the RNA-ligand complex. This approach enables a more nuanced evaluation beyond the composite GlideScore, highlighting energetic hotspots and favourability of specific interactions. Compounds were re-ranked according to their refined binding energies, allowing prioritization of ligands exhibiting strong and balanced contributions from hydrophobic contacts, electrostatic complementarity, and solvation dynamics. This selective energy-based filtering enhanced the identification of molecules with genuine binding potential to the TAR RNA three-nucleotide bulge and upper stem region, supporting subsequent molecular dynamics simulation studies and rational drug design efforts. It should be noted that the binding free energy values reported in this study are derived from docking-based scoring functions and energy decomposition analyses, which provide approximate estimates of ligand binding affinity. These values are not based on ensemble averaging over molecular dynamics trajectories and therefore do not include associated statistical deviations. While this approach enables rapid comparative screening of ligands, more rigorous free energy calculations such as MM-GBSA would be required to obtain statistically averaged binding energies, which may be considered in future studies.

Visualization and analysis of complexes

Structural visualization and analysis of docking poses and molecular dynamics trajectories were performed using Maestro (Schrödinger), PyMOL, and Discovery Studio to comprehensively elucidate RNA-ligand interactions^21,22. Maestro was primarily used for interactive inspection of docking results, binding site residues, and the dynamic behaviour of complexes during MD simulations, leveraging its integrated tools for hydrogen bond, π-π stacking, and electrostatic interaction analysis²³. PyMOL was employed for high-quality three-dimensional rendering of RNA-ligand complexes, enabling detailed visualization of molecular conformations, residue-specific contacts, and dynamic trajectory frames with customizable representations such as cartoon, sticks, and surfaces²¹. Discovery Studio complemented these tools by facilitating two-dimensional ligand interaction diagrams highlighting ligand contacts with non-covalent bonds, hydrophobic regions, and polar residues, aiding in clear depiction of key binding site interactions²². Combined use of these platforms ensured robust structural interpretation, clear graphical representation, and comprehensive interaction mapping, critical to understanding ligand binding modes and the conformational dynamics underpinning inhibitory potential.

Molecular dynamics simulation of free RNA and complexes

The top four ligands complexed with HIV-1 TAR RNA (PDB ID: 2L8H) were subjected to extensive molecular dynamics (MD) simulations to evaluate the stability and dynamic behaviour of the RNA-ligand complexes. MD simulations were performed using the OPLS5 force field, known for its accuracy in modelling nucleic acid systems²⁴. The complexes were solvated in a SPC explicit water model within an orthorhombic simulation box with a buffering distance of 10 Å, and neutralized with NaCl ions to mimic physiological ionic strength²⁵. Prior to production runs, energy minimization was executed to resolve steric clashes and relax the system using a steepest descent algorithm²⁶. Equilibration was conducted in two phases: initial NVT ensemble equilibration was done to stabilize temperature at 300 K, followed by NPT ensemble equilibration, which was done at 1 atm pressure to allow for density stabilization and box dimension adjustments under constant pressure and temperature conditions²⁷.

Subsequently, the production stage of MD simulation was carried out for 500 ns under canonical ensemble conditions, maintaining 300 K and 1 atm using the Nosé-Hoover thermostat and the Martyna-Tobias-Klein barostat, ensuring a realistic biological environment¹⁴. Trajectory data were recorded at appropriate intervals for subsequent detailed analysis. This rigorous simulation protocol was designed to capture the conformational flexibility, RNA-ligand interaction dynamics, solvent effects, and potential induced fit phenomena, providing atomistic insights into the mechanistic basis of ligand binding and stabilization of inhibitory RNA conformations. MD results were further analysed through root-mean-square deviations (RMSD), radius of gyration (Rg), root-mean-square fluctuations (RMSF) and solvent accessible surface area (SASA) to characterise the effect of ligand binding on TAR RNA structural dynamics²⁸.

Results and Discussion

Docking results

Comprehensive structure-based virtual screening of the Enamine library against the HIV-1 TAR RNA (PDB ID: 2L8H) led to the identification of four ligands- 2184, 14,204, 25,194, and 17,937 (Chemical structure shown in Fig. 2) selected based on their superior docking performance, physicochemical profiles, and detailed interaction analyses. For clarity, all ligands are consistently referred to by their catalogue identifiers (2184, 14204, 25194, and 17937) throughout the manuscript.

Fig. 2.

2D representation of top 4 selected hits A. 2184 (B) 14,204 (C) 25,194 (D) 17,937.

The docking scores for these compounds ranged from − 7.77 to −8.18 kcal/mol, reflecting high predicted binding affinity to TAR RNA’s conserved and functionally critical regions. Each ligand displayed a favourable molecular weight - between 281.3 and 394.4 Da, consistent with oral drug-likeness and facilitating optimal pharmacokinetic properties.

Further investigation into the binding modes revealed that the selected ligands consistently targeted the bulge and upper major groove of the TAR RNA (Fig. 1A), engaging a recurring panel of conserved nucleotides (notably U23, G36, C37, and A22) that are central to Tat recognition and viral transcriptional regulation. These interactions were mediated by robust hydrogen bond networks and extensive π-π stacking or π-anion interactions recognized as essential contributors to both molecular specificity and RNA recognition. For example, ligand 2184 primarily featured hydrogen bonding and T-shaped π-π contacts with U23, G36, and aromatic side chains of C37 and A22, stabilizing its binding orientation (Fig. 3A). By contrast, ligand 14,204 exhibited the lowest docking score of −8.08 kcal/mol and demonstrated an even broader interaction profile, forming additional π-anion contacts and bridging across C29, G28, A27, and the phosphate backbone, indicative of multipoint anchoring and affinity (Fig. 3B). Ligand 25,194 engaged in an intricate hydrogen bonding network with seven nucleotides and leveraged both π-anion and T-shaped π-π interactions (Fig. 4A). Meanwhile, compound 17,937 uniquely formed halogen (F) bonds and classical π-π stacking, enhancing its contact diversity (Fig. 4 B).

Fig. 3.

Integrated visualisation of I, II. 3D binding poses and III, IV. 2D interaction maps depicting the molecular recognition of RNA by small molecules (A) 2184 (B) 14,204.

Fig. 4.

Integrated visualisation of I, II. 3D binding poses and III, IV. 2D interaction maps depicting the molecular recognition of RNA by small molecules (A) 25,194 (B) 17,937.

Quantitative analysis of topological polar surface area (TPSA), a key determinant of molecular transport and bioavailability, showed values ranging from 43.3 to 152.2 Ų, in line with accepted norms for nucleic acid-targeted small molecules²⁹. Calculated aqueous solubility scores (logS from − 6.7 to −1.7) reinforced the drug-likeness and development potential. Glide g-scores and model energies corroborated the high binding strengths and were consistent across both empirical scoring and the deeper energetic decomposition Tables 1 and 2.

Table 1.

Binding energy of Top 4 selected hits with 2L8H.

Ligand name	Catalogue ID	Molecular weight	Docking score (kcal/mol)	glide g score	glide energy	glide model	TPSA	log S
2184	Z7606412589	321.808	−7.768	−7.768	−71.763	−101.696	43.32	−6.701
14,204	Z8102060370	394.389	−8.085	−8.085	−87.678	−123.902	152.23	−3.793
25,194	Z1281737251	281.316	−7.864	−7.864	−81.016	−117.299	82.7	−3.91
17,937	Z1162526590	317.322	−8.181	−8.181	−83.646	−126.462	88.05	−1.69

Table 2.

Summary of principal nucleotide interactions and ligand features in RNA complexes.

Complex	Key Interacting Nucleotides	Main Interaction Types
2184	U23, G36, A27, C37, A22, C27	H-bond, π- π T-shaped, van der Waals
14,204	U23, G36, C37, C29, A22, A27, G28	H-bond, π -Anion, Pi-Pi T-shaped, van der Waals
25,194	U23, G21, C37, A20, A22, A38, G36	H-bond, π- π T-shaped, Pi-Anion, van der Waals
17,937	A22, G36, C29, U23, C37	H-bond, Halogen (F), van der Waals, π - π stacked

Absorption, distribution, metabolism and excretion prediction of the ligands

The predicted ADME (Absorption, Distribution, Metabolism, and Excretion) properties of the selected ligands were assessed using QikProp (Schrödinger). Each molecule’s molecular weight ranged from 281.32 to 394.39 Da, well within the recommended ≤ 500 Da typically observed for orally available drugs, supporting favourable drug-likeness. Solvent accessible surface area (SASA), a measure of a molecule’s polar and nonpolar surface exposure fell between 570 and 662 Ų; the recommended range is 300–1000 Ų, indicative of appropriate molecular polarity and solubility. Hydrophobic (FOSA), hydrophilic (FISA), and π-character (PISA) components of SASA also conformed to their optimal ranges, which is important for balancing solubility and permeability^30,31. Hydrogen bond donors (donorHB, range: 1.25–4.25) and acceptors (accptHB, range: 2.5–9.75) were within the accepted cutoffs for oral drugs (donorHB ≤ 6; accptHB 2–20), supporting satisfactory permeability and bioavailability. All compounds were free of reactive functional groups (#rtvFG = 0), minimizing risks for instability or rapid biotransformation. CNS activity, predicted as a value from − 2 (inactive) to + 2 (active), ranged from − 2 to 1, and is considered well tolerated within the full range³¹.

Crucially, the predicted brain/blood partition coefficient (QPlogBB), which estimates a compound’s capacity to penetrate the blood-brain barrier, ranged from − 2.93 to 0.19 for these ligands. The optimum accepted range for QPlogBB is approximately − 3.0 to + 1.2. Values below zero suggest poor BBB penetration, potentially desirable for non-CNS targeted antivirals³². Higher QPlogBB values would indicate greater CNS permeability, relevant for drugs aimed at neurological targets or indications³⁰. Qualitative human oral absorption (categories: 1, 2, or 3 for low, medium, or high absorption) was either Category 2 or 3 for all compounds, and predicted percent absorption values were well above 25% (the poor absorption threshold), generally considered desirable for oral drugs (> 80% considered excellent)³³. All candidates met Lipinski’s “rule of five” (MW < 500; logP < 5; donorHB ≤ 5; accptHB ≤ 10), further supporting their status as drug-like molecules likely to be orally bioavailable. These indicate that the selected ligands possess favourable ADME characteristics for small-molecule drug development against RNA targets, showing appropriate molecular size, balanced polarity, permeability, oral absorption, and limited CNS penetration³⁰ Table 3.

Table 3.

Predicted ADME properties of selected ligands calculated using QikProp (Schrödinger).

Ligand name	M.W.	CNS	SASA	FOSA	FISA	PISA	Volume	donor HB	acceptor HB	QP log BB	Human Oral Absorption	Percent Human Oral Absorption
2184	321.808	1	598.93	145.894	76.373	308.433	1022.992	2	2.5	0.191	3	100
14,204	394.389	−2	661.989	139.668	311.876	210.444	1161.842	4	9.75	−2.928	2	47.414
25,194	281.316	−1	570.429	79.532	125.955	364.942	944.257	3	5.5	−0.999	3	89.814
17,937	317.322	−2	585.213	196.86	176.228	178.726	993.917	1.25	4.25	−1.27	3	78.515

Molecular dynamics simulation of the best docked ligands with HIV-1 TAR

Root mean square deviation (RMSD)

The simulation results show that certain ligands (2184 and 14204, Fig. 5A) stabilize the HIV TAR RNA structure, as evidenced by their lower RMSD fluctuations and minimal deviation compared with other complexes. These data suggest favourable binding conformations and support their potential for functional inhibition. Conversely, higher RMSD in others (25194 and 17937, Fig. 6A) may reflect induced flexibility, which could be essential for antagonistic activity, or conversely, suboptimal binding that destabilizes the RNA fold. Such nuanced analysis of RNA-ligand dynamics, contextualized by RMSD evolution and comparative trajectory stability, forms an integral part of evaluating lead compounds for further in silico and experimental validation in RNA-targeted drug discovery.

Fig. 5.

Comparative molecular dynamics analysis of HIV-1 TAR RNA in the apo form and in complex with ligands 2184 and 14,204. (A) Time evolution of structural stability represented as RMSD of the RNA backbone, (B) compactness measured by radius of gyration (Rg) of the RNA, (C) residue-level flexibility represented as RMSF of RNA nucleotides, and (D) solvent accessible surface area (SASA) of the RNA across 500 ns simulations. The black curve represents RNA alone, while colored curves correspond to RNA in complex with ligands, enabling comparison of ligand-induced structural effects.

Fig. 6.

Comparative molecular dynamics analysis of HIV-1 TAR RNA in the apo form and in complex with ligands 25,194 and 17,937. (A) Time evolution of structural stability represented as RMSD of the RNA backbone, (B) compactness measured by radius of gyration (Rg) of the RNA, (C) residue-level flexibility represented as RMSF of RNA nucleotides, and (D) solvent accessible surface area (SASA) of the RNA across 500 ns simulations. The black curve represents RNA alone, while colored curves correspond to RNA-ligand complexes, highlighting ligand-induced conformational changes.

Radius of gyration (Rg)

The comparative Rg analysis demonstrates that binding of small molecules to the major groove of HIV-1 TAR RNA can substantially influence the RNA’s global fold. Complexes 25,194 and 17,937.

(Fig. 6B) show the largest deviations in Rg, consistent with RMSD trends that predicted increased structural flexibility or local unwinding, possibly reflecting induced fit mechanisms or partial destabilization following ligand accommodation. In contrast, complexes 2184 and 14,204 (Fig. 5B) consistently maintain tight Rg distributions, reinforcing their candidate status for stabilizing effects and supporting a mechanism involving preservation of RNA compactness and secondary structure.

These insights suggest that not all groove-binding ligands equally stabilize TAR RNA; the best candidates for functional inhibition appear to be those that induce minimal expansion while supporting sustained atomic compactness. Such data provide a rational basis for prioritizing ligands 2184 and 14,204 for further in silico and biochemical assays, as their effect on Rg supports favourable steric constraints on the RNA fold.

The observed differences in ligand-induced structural responses suggest distinct potential mechanisms of Tat inhibition. Ligands such as 2184 and 14,204, which maintain compact RNA conformations with stable Rg profiles, are likely to inhibit Tat binding by directly occupying and stabilizing the bulge major groove interface, thereby restricting access to the native recognition site. In contrast, ligands 25,194 and 17,937 induce increased structural expansion and flexibility, which may disrupt the spatial organization required for Tat binding. However, such expansion may also reflect reduced binding specificity or partial destabilization of the RNA fold, potentially leading to less consistent inhibitory effects. These findings indicate that stabilization of functionally relevant TAR conformations may represent a more reliable strategy for effective inhibition compared to global structural perturbation.

Root mean structure fluctuation (RMSF)

The RMSF results highlight localized and global increases in residue flexibility upon ligand binding, with the most pronounced effects seen in complexes 25,194 and 17,937 (Fig. 6C). These ligands, previously shown to expand RNA structure in Rg and RMSD analyses, further destabilize the RNA backbone and promote increased atomic fluctuation, especially at chain ends and internal loops. Conversely, ligands 2184 and 14,204 (Fig. 5C) generally maintain flexibility profiles closer to the control, suggesting a stabilizing interaction that may preserve native TAR RNA architecture and minimize unwanted conformational drift.

Residue-level RMSF data thus confirms that ligand selection can dramatically alter RNA atomic motion. Minimizing excessive flexibility, as achieved by 2184 and 14,204, should be prioritized in inhibitor optimization.

Solvent accessible surface area (SASA)

SASA results reinforce the ligand-dependent effects on RNA folding: complexes 25,194 and 17,937 (Fig. 6D) substantially increase solvent exposure, consistent with previously observed Rg and RMSD expansion and flexibility. This behaviour signals a destabilizing influence, possibly causing local unfolding or groove expansion. In contrast, ligands 2184 and 14,204 (Fig. 5D) mirror the reference RNA in SASA distribution, supporting their stabilizing effect on TAR RNA architecture.

These findings visualize the ability of certain small molecules to either promote compactness or induce unwinding in the RNA target. Enhanced SASA values generally correlate with greater conformational adaptation and solvent exposure, which may impact ligand efficacy, stability, and the ability to block protein-RNA recognition surfaces. Ligands yielding lower SASA values (2184,14204) remain ideal candidates for maintaining RNA functional integrity and inhibiting pathological interactions Table 4.

Table 4.

Comparative Analysis of Key Structural and Flexibility Parameters in Simulated RNA-Ligand Systems.

2L8H	Atom Count	RMSD Mean (nm)	Rg Mean (nm)	RMSF Mean (nm)	SASA Mean (nm²)	Groove Binding	Stability Trend
	930	2.0–2.5.0.5	15.5	2.4	5600	Major	Stable, compact
2184	969	2.3–2.8	15.0	3.2	5700	Major	Stable, compact
14,204	977	2.2–2.7	14.5	3.1	5750	Major	Most stable, compact
25,194	966	3.0–3.5.0.5	17.0	3.6	6100	Major	Expanded, dynamic
17,937	969	3.2–4.0.2.0	17.0	4.0	6300	Major	Expanded, flexible

Conclusion

This integrative computational study highlights the feasibility and promise of targeting the HIV-1 TAR RNA bulge groove interface with small molecules drawn from an RNA-focused library. By combining rigorous structure-based virtual screening of the Enamine library against the NMR structure of full-length TAR (2L8H) with extensive 500 ns molecular dynamics simulations, the work identifies four ligands, 2184, 14,204, 25,194, and 17,937 that bind key nucleotides in the U23- C25 bulge and adjacent major groove with favourable docking energies and well-balanced ADME profiles. The convergence of docking scores in the range of approximately − 7.8 to −8.2 kcal/mol, drug-like physicochemical properties, and stable bulge-major-groove engagement underscores the tractability of this RNA element as a druggable target when guided by a carefully curated RNA-compatible chemical space.

Beyond identifying specific hits, the study underscores the broader potential of structure-enabled RNA drug discovery. Careful target preparation, focused grid definition at a functionally critical bulge, and multi-step docking and rescoring helped to enrich ligands that mimic or disrupt native recognition motifs. Subsequent ADME predictions show that these scaffolds combine appropriate molecular size, balanced polarity, acceptable predicted absorption, and limited CNS penetration, aligning with desirable properties for non-CNS antivirals. While the predicted low BBB permeability of the identified ligands may limit their direct action against HIV reservoirs within the CNS, this property may still be advantageous in reducing potential CNS-related toxicity for peripherally acting antivirals. Given that the brain represents a well-established viral reservoir, these compounds are more suitably envisioned as components of combination antiretroviral therapy, where they may act synergistically with agents capable of effective CNS penetration. Furthermore, the identified scaffolds provide a valuable foundation for future medicinal chemistry optimization aimed at modulating pharmacokinetic properties, including BBB permeability, to expand their therapeutic scope toward CNS-targeted HIV intervention. The overall workflow thus provides a practical template for discovering chemotypes that engage complex RNA architectures while retaining developability.

Downstream, experimental validation will be essential to translate these computational insights into tangible therapeutic opportunities. Biophysical assays such as fluorescence anisotropy, NMR titration, and surface plasmon resonance can confirm direct TAR binding, map contact surfaces, and quantify affinities for the top hits and their analogues. Cellular reporter systems that read out Tat-dependent transcriptional activation will enable assessment of functional inhibition and preliminary off-target profiles. Parallel medicinal chemistry, guided by the interaction patterns around U23, A22, G36, C37, and adjacent residues, can be used to tune potency, selectivity, and pharmacokinetics. Expanding this strategy to diverse TAR sequences and related viral or host regulatory RNAs may ultimately contribute to a broader platform for RNA-targeted antivirals and help establish small-molecule modulation of structured RNA as a viable route for managing HIV-1 infection and latency. A limitation of the present study is the absence of trajectory-based free energy calculations, which would allow statistical estimation of binding energy fluctuations. Incorporation of such methods in future work will further enhance the quantitative assessment of ligand binding and stability.

Abbreviations

HIV

Human immunodeficiency virus

RNA

Ribonucleic acid

TAR

Trans-activation response

TAT

Trans-activator protein

UTR

Untranslated regions

RMSD

Root mean square deviation

RMSF

Root-mean-square fluctuations

Rg

Radius of gyration

SASA

Solvent accessible surface area

MD

Molecular dynamics

ADME

Absorption, Distribution, Metabolism, and Excretion

CNS

Central nervous system

PKR

Protein Kinase R

TPSA

Topological polar surface area

ns

nanosecond

Author contributions

PR, GS and RPB conceived the idea; PM and SPS executed the experiments in coordination with JG; PM made the figures and wrote the first draft. PR, GS and RPB edited and finalized the draft in coordination with other authors. All authors have approved it for publication.

Funding

DHR-start up (R.12020/02/2024-HR/E-Office:8292859), ICMR (17 × (3)/Ad-hoc/69/2022-ITR), DST-SERB (CRG/2022/000628), and CSIR (37/1743/23/EMR-II), Government of India. Schrodinger package is purchased through DST-CSRI (DST/CSRI/2021/7) grant.

Data availability

All data generated or analysed during this study are included in this article.

Declarations

Competing interests

The authors declare no competing interests.