Computational drug designTarget: CXCR4 · PDB ID: 3ODU

Designing molecules for CXCR4

Full thesis title

Structure-Based Design and Computational Evaluation of Amentoflavone Derivatives as Novel CXCR4 Antagonists

Degree and thesisBachelor's thesis

Institutionİnönü University, Department of Molecular Biology and Genetics

SupervisorDr. Seçil Demiral

AuthorUğur Cem Yıldız

I designed five amentoflavone derivatives, docked them against CXCR4 and compared their predicted ADME and toxicity profiles. I wanted to choose candidates for experimental testing.

5Designed derivatives

−11.9Docking score for two candidates · kcal/mol

3ODUCXCR4 crystal structure

Main finding

Deoxy and Fluoro-Deoxy tied for the most favorable docking score. I prioritized Fluoro-Deoxy after considering the docking and toxicity predictions together. Deoxy met all four Lipinski criteria.

Experimental testing is still needed to assess efficacy and safety.

Research questions and hypothesis

I examined how changes to amentoflavone affected docking scores, predicted receptor contacts and ADME/toxicity profiles.

Molecule design: Can selective deoxygenation and amination improve the predicted CXCR4 docking score relative to amentoflavone and Plerixafor?
Receptor contacts: Do the predicted poses place the derivatives in contact with binding-pocket residues through hydrogen bonds and aromatic interactions?
Drug-likeness and toxicity: Which derivatives have ADME and toxicity predictions that support further experimental evaluation?

Working hypothesis: I hypothesized that removing selected hydroxyl groups would reduce non-specific polar contacts and steric hindrance, lower polar surface area and improve the predicted fit around Asn43, Tyr55 and Trp104.

Computational method

I prepared the receptor and ligands, ran docking simulations, then compared predicted interactions and compound profiles.

Step 1

Receptor preparation

I used the human CXCR4 crystal structure 3ODU from RCSB PDB, resolved at 2.50 Å with the antagonist IT1t. I removed crystallographic waters, co-crystallized antagonists and non-protein heteroatoms, then added polar hydrogens and assigned Kollman united-atom partial charges in AutoDock Tools.

Step 2

Derivative design

I drew five amentoflavone derivatives in MolSoft ICM-Chemist. I generated 3D coordinates and minimized their energy with the NCI Online SMILES Translator, then assigned Gasteiger charges and rotatable bonds in AutoDock Tools.

Step 3

Molecular docking

I ran AutoDock Vina v1.1.2 with a grid centered on the orthosteric pocket: x=19.78, y=−6.85, z=68.11, with dimensions of 52 × 52 × 52 Å. Exhaustiveness was set to 32.

Step 4

Interaction and toxicity prediction

I inspected 2D and 3D ligand–receptor contacts in BIOVIA Discovery Studio Visualizer. SwissADME provided physicochemical, Lipinski and ADME predictions. ProTox-II provided predicted oral LD50, toxicity class, organ toxicity, mutagenicity and carcinogenicity.

Table 1. Amentoflavone derivatives and design rationale
Derivative	Structural change	Design rationale
Deoxy-Amentoflavone	Selective deoxygenation only; no added functional group	Tests deoxygenation alone and its effect on non-specific hydrogen bonding
Fluoro-Deoxy-Amentoflavone	Selective deoxygenation + F at para position of one terminal phenyl ring	Tests fluorine substitution in place of an amino group, comparing docking and predicted toxicity
Deoxy-Amino-Amentoflavone	Selective deoxygenation + NH₂ at para position of terminal phenyl ring	Tests predicted contacts with Asn43 and Tyr55
MethylAmino-Deoxy-Amentoflavone	Selective deoxygenation + NHCH₃ at para position of terminal phenyl ring	Evaluates effect of N-methylation on binding geometry and lipophilicity
Di-Amino-Amentoflavone	Selective deoxygenation + NH₂ at para positions of both terminal phenyl rings	Tests whether adding a second amino group improves the docking score

Tools and data

Receptor structureHuman CXCR4 (PDB ID: 3ODU ↗, 2.50 Å resolution, X-ray diffraction)

Docking softwareAutoDock Vina (v1.1.2) with exhaustiveness = 32

Structure preparationAutoDock Tools (v1.5.7), MolSoft ICM-Chemist, NCI SMILES Translator

Interaction analysisBIOVIA Discovery Studio Visualizer (Dassault Systèmes), OpenBabel (v3.1.1)

ADME predictionsSwissADME web server (Lipinski compliance, TPSA, Log P, CYP inhibition)

Toxicity predictionsProTox-II platform (acute oral toxicity, organ toxicity, mutagenicity, carcinogenicity)

Docking results and compound profiles

All five derivatives received more negative docking scores than amentoflavone (−11.1 kcal/mol) and Plerixafor (−9.4 kcal/mol) under the same protocol. Stronger binding and greater efficacy would need to be demonstrated experimentally.

Table 2. Docking scores and predicted compound profiles
Compound	Docking score (kcal/mol)	MW (g/mol)	TPSA (Å²)	Lipinski violations	Bioavailability score	Predicted toxicity class	Predicted Mutagenicity
Fluoro-Deoxy-Amentoflavone	−11.9	524.45	141.34	1	0.55	Class 5 (2500 mg/kg)	Predicted inactive
Deoxy-Amentoflavone	−11.9	490.46	121.11	0 (Fully compliant)	0.55	Class 5 (2430 mg/kg)	Predicted inactive
Deoxy-Amino-Amentoflavone	−11.8	521.47	167.36	1	0.55	Class 3 (159 mg/kg)	Predicted active
MethylAmino-Deoxy-Amentoflavone	−11.8	535.50	153.37	1	0.55	Class 3 (159 mg/kg)	Predicted active
Di-Amino-Amentoflavone	−11.5	520.49	173.15	1	0.55	Class 3 (159 mg/kg)	Predicted active; carcinogenicity also predicted
Natural Amentoflavone (Parent)	−11.1	538.46	181.80	2	0.17	Class 5 (3919 mg/kg)	Predicted inactive
Plerixafor (Clinical Reference)	−9.4	502.78	78.66	2	0.17	Class 4 (550 mg/kg)	Predicted inactive

Fluoro-Deoxy-Amentoflavone−11.9 kcal/mol
Deoxy-Amentoflavone−11.9 kcal/mol
Deoxy-Amino-Amentoflavone−11.8 kcal/mol
MethylAmino-Deoxy-Amentoflavone−11.8 kcal/mol
Di-Amino-Amentoflavone−11.5 kcal/mol
Natural Amentoflavone (parent)−11.1 kcal/mol
Plerixafor (clinical reference)−9.4 kcal/mol

All five derivatives received more negative docking scores than amentoflavone and Plerixafor under the same protocol. Experimental testing is needed to assess binding, efficacy and safety.

1. Deoxygenation and predicted Tyr55 contacts

Deoxygenation produced the largest change in docking score across the series. In the interaction maps, amentoflavone contacted Cys196, Asn43 and Ser459 but not Tyr55. The Deoxy pose included predicted hydrogen bonds with Tyr55 and Asp107 and aromatic stacking with Trp104.

2. Why I prioritized Fluoro-Deoxy

Fluoro-Deoxy and Deoxy both scored −11.9 kcal/mol. Fluoro-Deoxy had a predicted toxicity class of 5 (LD50: 2500 mg/kg), inactive mutagenicity and carcinogenicity predictions, and no predicted CYP2C19 inhibition. Its interaction map showed contacts with Tyr55, Asn43, Cys196, Tyr429 and Trp104. These combined predictions supported further testing; they do not establish safety.

3. The second amino group in Di-Amino

Di-Amino received the least favorable score among the five derivatives (−11.5 kcal/mol). Its interaction map showed an unfavorable donor–donor contact between the second amino group and Asn43. This contact may help explain the score in the static docking model.

Structures and interaction maps

Three-dimensional ribbon structure of human CXCR4 receptor (PDB ID: 3ODU) — **Figure 1:** Human CXCR4 crystal structure 3ODU (2.50 Å resolution), shown as a ribbon model with seven transmembrane helices and the orthosteric pocket.

Two-dimensional chemical structure of natural Amentoflavone — **Figure 2:** Amentoflavone chemical structure: two apigenin units linked through a C-3'–C-8'' bond.

2D interaction map of natural Amentoflavone bound to CXCR4 — **Figure 3:** Predicted amentoflavone–CXCR4 contacts from BIOVIA Discovery Studio: hydrogen bonds with CYS A:196, ASN A:43 and SER A:459, and aromatic stacking with TRP A:104 and HIS A:123. No TYR A:55 contact was identified.

2D interaction map of Deoxy-Amentoflavone bound to CXCR4 — **Figure 4:** Predicted Deoxy–CXCR4 contacts: hydrogen bonds with TYR A:55 and ASP A:107, and aromatic stacking with TRP A:104.

Predicted Deoxy-Amino-Amentoflavone contacts with CXCR4 — **Figure 5:** Deoxy-Amino interaction map, with predicted hydrogen-bond contacts involving TYR55, ASN43, CYS196 and ASP107.

Predicted Di-Amino-Amentoflavone contacts with CXCR4 — **Figure 6:** Di-Amino interaction map, including an unfavorable donor–donor contact with ASN43.

2D interaction map of Fluoro-Deoxy-Amentoflavone bound to CXCR4 — **Figure 7:** Fluoro-Deoxy interaction map, with predicted hydrogen-bond contacts involving TYR A:55, ASN A:43, CYS A:196 and TYR A:429, and aromatic contacts with TRP A:104.

Interactive 3D binding pose

Open the 3D view to load the viewer and structure (about 0.84 MB before compression).

Fluoro-Deoxy-Amentoflavone docked in the prepared CXCR4 structure (3ODU). This is the top-ranked Vina pose, with a score of −11.9 kcal/mol. ASN43, TYR55, CYS196, TYR429 and TRP104 are labeled to help locate the predicted contacts. Hydrogens are omitted from the display.

Predicted MethylAmino-Deoxy-Amentoflavone contacts with CXCR4 — **Figure 8:** MethylAmino-Deoxy interaction map, with predicted hydrogen-bond contacts involving TYR429, TYR55, ASN43 and ASP107.

Method record and study files

The JSON file contains the method details. The links below lead to the study files I can share.

My contribution

I designed five derivatives, prepared the receptor and ligands, ran docking analyses and compared compound profiles.

Receptor: CXCR4 · 3ODU
AutoDock Vina: v1.1.2 · exhaustiveness 32
Search box: 52 × 52 × 52 Å
Centre: x 19.78 · y −6.85 · z 68.11

Method record (JSON)Docking results (CSV)Displayed binding pose (PDB)

Limits of this record

The shared PDB is the prepared receptor and one ligand pose shown in the 3D viewer. The full ligand inputs, original Vina configuration file and run logs are not part of this publication package. I have not independently rerun the analysis for this publication.

Limitations and interpretation

How to read these results

Docking, ADME and toxicity results are computational predictions. They identify candidates to test experimentally; they do not demonstrate biological efficacy or clinical safety.

Static docking: Docking scores estimate the favorability of a pose within a simplified model. This workflow does not capture full receptor flexibility, solvent effects or membrane dynamics.
Prediction limits: SwissADME and ProTox-II use models based on existing chemical data. Predictions for new derivatives depend on how well those compounds fit each model's applicability domain.
Small compound set: I evaluated five designed derivatives. A larger compound library would be needed to explore other substitutions.
One receptor structure: 3ODU represents one antagonist-bound conformation. Molecular dynamics or docking against several conformations would help assess how receptor movement affects the results.

References

Banerjee, P. et al. (2018). ProTox-II: A webserver for the prediction of toxicity of chemicals. Nucleic Acids Research, 46(W1), W257–W263. doi:10.1093/nar/gky318 ↗
Burger, J. A., & Kipps, T. J. (2006). CXCR4: A key receptor in the crosstalk between tumor cells and their microenvironment. Blood, 107(5), 1761–1767. doi:10.1182/blood-2005-08-3182 ↗
Daina, A., Michielin, O., & Zoete, V. (2017). SwissADME: A free web tool to evaluate pharmacokinetics and drug-likeness. Scientific Reports, 7, 42717. doi:10.1038/srep42717 ↗
Lipinski, C. A. et al. (2001). Experimental and computational approaches to estimate solubility and permeability. Advanced Drug Delivery Reviews, 46(1), 3–26. doi:10.1016/s0169-409x(00)00129-0 ↗
Trott, O., & Olson, A. J. (2010). AutoDock Vina: Improving the speed and accuracy of docking. Journal of Computational Chemistry, 31(2), 455–461. doi:10.1002/jcc.21334 ↗
Wu, B. et al. (2010). Structures of the CXCR4 chemokine receptor in complex with small molecule and cyclic peptide antagonists. Science, 330(6007), 1066–1071. doi:10.1126/science.1194396 ↗
Xiong, X. et al. (2021). Insights Into Amentoflavone: A Natural Multifunctional Biflavonoid. Frontiers in Pharmacology, 12, 768708. doi:10.3389/fphar.2021.768708 ↗