In Silico Screening of BiTE Molecule Targeting CD22 and T-Cell Receptor Complex for Acute Lymphoblastic Leukemia

ABSTRACT

Acute Lymphoblastic Leukemia (ALL) is a rapidly progressing cancer that affects the blood and bone marrow, and it is caused by the uncontrolled division of white blood cells known as lymphoblasts. In this research, we used AlphaFold 3 to generate 3D molecular structures of the T-Cell Receptor (TCR) and the CD22 receptor, using their respective amino acid sequences as inputs. We identified 20 humanized single-chain variable fragment (scFv) antibody fragments found in the Structural Antibody Database (SAbDab) database. The humanized scFvs were docked to both the TCR and CD22 receptors to simulate the binding region and the binding energy between the two molecules. Then, we determined the most optimal binding region of the 10 models simulated by HDock for each of the docking simulations and compared both the binding energy and binding site to determine the most optimal scFv. Finally, Protein-Ligand Interaction Profiler (PLIP) and PRODIGY softwares were utilized to further analyze the binding affinity and interactions between the scFv and TCR or CD22 proteins, thereby gaining insights into specific amino acids directly involved in binding. The results identified scFv 9MFN as exhibiting strong binding to CD22, with a binding energy (ΔG) of −21.2 kcal/mol. Similarly, scFv 9B7R demonstrated favorable binding to the TCR receptor complex, with a ΔG of −13.3 kcal/mol, suggesting their potential utility in designing BiTE molecules for ALL therapy. The current approach highlights the application of in silico screening to accelerate therapeutic antibody development against ALL and help in reducing the global burden of the disease.

INTRODUCTION.

Acute Lymphoblastic Leukemia (ALL) is a type of cancer that can spread quickly, originating in the bone marrow and occurring in the blood.[1] It is caused by the uncontrolled overproduction of immature white blood cells called lymphoblasts. Some symptoms that ALL patients experience include fatigue, pale skin, joint pain, and easy bruising or bleeding. The main approach to treating the disease is chemotherapy in 3 phases: induction, consolidation, and maintenance. There are also more focused/targeted approaches, such as Chimeric Antigen Receptor (CAR) T-cell therapy, which involves modifying a patient’s T cells to identify better and target leukemia cells. However, there are issues and side effects with current treatment methods, such as chemotherapy, including hair loss and nausea, as well as resistance to standard treatments, making them ineffective. Thus, the limitations of current treatment approaches provide the primary motivation for this study to explore novel, targeted therapeutic strategies for ALL.

Because ALL often involves malignant B-cells that overexpress specific surface markers, understanding these targets—such as CD22 on B-cells and T-cell Receptor (TCR) on T-cells—is essential for developing more precise immunotherapies. CD22 is a cell surface receptor located on the surface of B-cells and is appropriately expressed in mature cells. CD22 works to inhibit the function of B-cell Receptors, which are key in activating B-Cells and initiating immune responses, aiding the immune system in stopping overactivation. These receptors are often overexpressed in ALL affected cells. On the other hand, TCR is located on the surface of T-cells. It is a complex of proteins that includes smaller alpha- and beta-subunits targeted to recognize specific antigens, initiate T Cell Activation, and kill cells. Antibodies are proteins produced by B cells that function as part of the immune system to identify foreign substances, or antigens. Once they identify antigens, they bind to them, acting as markers that signal other immune cells to destroy the antigens. A single-chain variable fragment (scFv) is an antibody fragment made by linking the variable regions of the heavy (VH) and light (VL) chains into one single protein that can still bind to a specific antigen. BiTE molecules are engineered proteins that bind T cells to cancer cells.Antibodies are proteins produced by B cells that function as part of the immune system to identify foreign substances, or antigens. Once they identify antigens, they bind to them, acting as markers that signal other immune cells to destroy the antigens. BiTE molecules (Bispecific T-cell Engagers) are engineered proteins that bind T cells to cancer cells, Figure 1.[2] BiTE technology is often used in cancer treatments to help T cells efficiently identify and kill cancer cells, thereby eliminating cancerous cells. BiTE molecules are now being applied through drugs given to patients to treat various cancers.[2]

To design such therapeutic molecules effectively, computational tools play a crucial role in predicting and analyzing how these engineered proteins will interact with their biological targets. Molecular docking simulation is a computational technique that predicts how specific molecules bind to proteins.[3] These simulations are often used in drug discovery to aid in designing molecules by allowing scientists to predict how molecules interact with target proteins. Some docking software are designed to give very accurate results, some work better for certain molecules (like proteins, antibodies, or small drugs), and some allow molecules to move and bend (flexible docking) while others keep them fixed in place (rigid docking). AlphaFold is an artificial intelligence software capable of predicting the 3D structure of proteins from their amino acid sequences.[4]  It is beneficial because it has high accuracy in predicting the exact folds of proteins. Traditionally, determining protein folds requires time-consuming and expensive lab experiments like X-ray crystallography and nuclear magnetic resonance (NMR).  Alphafold is often used in drug discovery, disease research, and biotechnology because it can accelerate research in these fields by predicting over 200 million proteins, including almost all known proteins on the planet. In this research, we performed comprehensive computational simulations (using AlphaFold 3 and HDOCK software) and identified two scFvs that bind specifically to the TCR and the CD22 receptor, respectively.  This BiTE will simultaneously bind CD22 on leukemia cells and the TCR on T-cells, bringing them together to trigger targeted killing of Acute Lymphoblastic Leukemia (ALL) cells.

MATERIALS AND METHODS.

Sequence Retrieval and Protein Modeling.

Uniprot [5] is a database and was used it to identify protein sequences in structures such as the T-cell receptor and CD22, and to gain additional information about their composition. Alphafold is an Artificial Intelligence (AI) tool that predicts and displays 3D protein structures from amino acid sequences.[4] We use it to create accurate, applicable protein models for our simulations and analyses.

Antibody (scFv) Selection.

The Protein Data Bank is an extensive database of protein 3D structures.[6] We use it to view 3D structures and other data of key biological structures in our research, such as the T Cell Receptor complex. Protter is a web tool that provides visualizations of protein sequences, including information and structures like the cell membrane.[7] We use PROTTER to create clean diagrams that help us visualize the structures and features of the proteins we analyze. The SAbDab database, or Structural Antibody Database, is a resource that contains structural and sequence information on a large number of antibodies and their scFv sequences.[8]

Binding Site Prediction and Docking Validation.

To independently validate the docking sites, binding pocket predictions were generated using PrankWeb and ScanNet, which apply machine-learning and deep-learning algorithms to identify functionally essential residues on protein surfaces.

Molecular Docking Simulations and Analysis.

The molecular docking simulations using HDock were performed. Detailed interaction profiling—including hydrogen bonds, hydrophobic contacts, salt bridges, and π-stacking interactions—were performed using PLIP.[9] PLIP analyzes and identifies non-covalent interactions (like hydrogen bonds and hydrophobic contacts) between proteins and ligands. Binding affinity (ΔG) and interface stability were further evaluated using PRODIGY.[10] PRODIGY predicts the binding affinity (ΔG) and stability of protein–protein complexes.

Molecular Dynamics Simulation.

Finally, molecular dynamics (MD) simulations of the top scFv–receptor complexes (9mfn–CD22 and 9b7r–TCR) were conducted using GROMACS to assess structural stability over time and confirm the persistence of key interactions under physiologically relevant conditions.[11] GROMACS is a molecular dynamics simulation software used to study the motion and stability of biomolecules over time. We used it to predict the binding between either the TCR or CD22, and a particular scFv found in humans to study how they interact and bind.

RESULTS.

This paper uses computational tools to identify and evaluate antibody fragments that could simultaneously target CD22 on leukemia cells and the T-cell receptor on T-cells to create an effective BiTE therapy for ALL. The study integrates AI-based structure prediction, molecular docking, and molecular dynamics to rapidly screen and validate the strongest antibody candidates for future cancer treatment.

Molecular Docking Simulations and Analysis.

The protein CD22 and the T-cell receptor were used for molecular docking, and the 20 different scFvs were docked to each of the receptors individually, producing a 3D visualization of the docking as well as 10 possible docking combinations between each scFv and amino acid, with each combination producing numerical values for the docking score, confidence score, ligand rmsd (Å), and interface residues.

The interaction between the CD22 protein and the 9mfn scFv (Figure 2D) as well as the interaction between TCR and 9b7r (Figure 2A). In this figure, the antibody is shown in pink and green colors while the amino acids of the protein interacting with the antibody are shown in blue. Once the molecular docking was complete, the binding energy was calculated for each of the 40 possible interactions (20 for CD22 and 20 for TCR) between the receptors and scFvs as shown in Figure 3. Based on the binding energies and visual analysis of the docking interfaces, the scFvs that demonstrated strong binding and optimal interaction sites with their respective receptors were CD22–scFv (PDB ID: 7O52, ΔG = −21.2 kcal/mol) and TCR–scFv (PDB ID: 9B7R, ΔG = −13.3 kcal/mol; no experimentally reported TCR complex structure available).

Once the antibodies were selected 9mfn and 9b7r for CD22 and TCR, respectively, the different properties of the antibodies were also computed as shown in Figure 4. It shows the variable region of the scFv, which is the binding (sticky) region that interacts with the receptor. The physicochemical properties of this region were analyzed to determine whether it is hydrophobic or hydrophilic and whether it carries a positive or negative charge. It was observed that the region contains hydrophilic residues and is predominantly neutral in charge.

Molecular Dynamics Simulation Analysis.

Molecular dynamics simulations of the selected receptor-scFv complex were performed using GROMACS, as shown in Figure 5. The different molecular dynamics analyses were computed to validate that the antibody remains bound to the receptor during the MD simulation. During the entire MD simulations, the scFv remains bound to the receptor. The Root Mean Square Deviation (RMSD) values shown for both the docked 9mfn scFv and CD22 Receptor as well as 9b7r and TCR shown in Figure 5 are mostly stable with the values on the graph not changing drastically. This is indicative of structural stability and overall structural change, so little differences in the RMSD value throughout the simulation time with both docked structures show that the structures do not change substantially over the course of time when in cell-like conditions simulated through MD simulation, contributing to the effectiveness of the BiTE. Further the Root Mean Square Fluctuation of both structures shown in Figure 5 shows specific spikes correlated with certain regions on the receptor which represents points of flexibility within the structures.

BiTE 3D Modeling.

Finally, we also designed the 3D structure of the BiTE molecule using AlphaFold. To the best of our knowledge, no experimentally resolved 3D structure of a BiTE molecule is currently available, so we generated a predicted structure using AlphaFold. The 3D structure shown in Figure 6 was generated of a complete BiTE molecule using the amino acid sequences of the selected scFvs (PDB ID = 9MFN and 9B7R) as well as a flexible linker in between with a specific linker length of 10 amino acids composed of glycine which is small and stable contributing to the BiTE molecule effectiveness because flexibility increases the likelihood of binding with both receptors at various angles. After analysis of various linker lengths ranging from 10 amino acids to 35 amino acids, a 10 amino acid length was visually determined to be the most suitable based on the effects that linker length had on the different generated 3D structures.

Toxicity and Thermal Stability.

After obtaining the 3D model of the complete BiTE molecule with a linker, its amyloidogenicity and thermal stability were also computed. The amyloidogenicity probability values range from 0 to 1, indicating which amino acid regions within the BiTE molecule are more likely to aggregate. Higher likelihood of aggregation indicates toxicity, so the graph shows the specific regions that are more likely to contribute to the toxicity of the BiTE. In addition, its melting point was computed to be 47.2 °C, with the individual scFvs 9b7r and 9mfn having melting points of 54.27 °C and 58.30 °C, respectively. All three of these melting point values indicate temperatures far exceeding the environmental conditions within and between cells in the human body. This suggests that both the complete BiTE molecule as well as the individual scFvs can keep their structure intact even at high temperatures, indicating that the BiTE is suitable for use in the cell environment.

DISCUSSION.

The BiTE molecule design has many helpful applications. It can be used in targeted ALL therapy; the identified scFv candidates (9mfn for CD22 and 9b7r for TCR) can be engineered into BiTE molecules to selectively redirect T cells toward ALL cells, reducing off-target toxicity compared to chemotherapy. In addition, the findings can accelerate drug discovery, enabling the in silico workflow (AlphaFold, docking, ML validation) to be applied broadly to design therapeutic antibodies against other cancers and immune disorders. Serving as an advancement in personalized medicine, this computational approach can be extended to patient specific ALL mutations or receptor variants, enabling customized BiTE design. The BiTE design strategy can potentially be used as a platform for combination therapies, and can be integrated with CAR-T, checkpoint inhibitors, or hypoxia-targeting drugs to enhance therapeutic outcomes. In addition, BiTE molecules are much smaller (~55 kDa) than full monoclonal antibodies (~150 kDa) because they lack the Fc region and consist only of two linked scFvs. Their smaller size allows better tumor penetration and closer T-cell–tumor cell interaction, improving immune synapse formation and cytotoxic activity.

However, this study’s limitations include discrepancies between computational predictions and reality, as docking and ML methods may not fully capture the dynamic complexity of in vivo interactions. There is also a lack of experimental validation, as binding affinities and interactions have not yet been confirmed by laboratory assays such as SPR or ITC.[12] Finally, the limited scope of the database may overlook other potentially stronger candidates because the antibody selection was restricted to scFvs available in SAbDab. The next step in the study is experimental validation, testing top BiTE candidates in vitro using ALL cell lines to confirm computational findings.

CONCLUSION

This study successfully identified two strong scFv candidates—9mfn for CD22 and 9b7r for the T-cell receptor—using an integrated in silico pipeline combining AlphaFold structure prediction, molecular docking, machine-learning validation, and molecular dynamics simulations. The stable and favorable interactions observed suggest that these scFvs can be engineered into an effective BiTE for targeted ALL therapy. Overall, this computational approach demonstrates a fast, cost-effective strategy for accelerating antibody discovery and provides a strong foundation for future experimental validation and development of next-generation cancer immunotherapies.

REFERENCES

1. C.-H. Pui, W. E. Evans, Treatment of Acute Lymphoblastic Leukemia. New England Journal of Medicine 354, 166-178 (2006).
2. H. Einsele et al., The BiTE (bispecific T-cell engager) platform: Development and future potential of a targeted immuno-oncology therapy across tumor types. Cancer 126, 3192-3201 (2020).
3. J. Fan, A. Fu, L. Zhang, Progress in molecular docking. Quantitative Biology 7, 83-89 (2019).
4. J. Abramson et al., Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493-500 (2024).
5. T. U. Consortium, UniProt: a hub for protein information. Nucleic Acids Research 43, D204-D212 (2014).
6. S. K. Burley et al., in Protein Crystallography: Methods and Protocols, A. Wlodawer, Z. Dauter, M. Jaskolski, Eds. (Springer New York, New York, NY, 2017), pp. 627-641.
7. U. Omasits, C. H. Ahrens, S. Müller, B. Wollscheid, Protter: interactive protein feature visualization and integration with experimental proteomic data. Bioinformatics 30, 884-886 (2013).
8. J. Dunbar et al., SAbDab: the structural antibody database. Nucleic Acids Research 42, D1140-D1146 (2013).
9. S. Salentin, S. Schreiber, V. J. Haupt, M. F. Adasme, M. Schroeder, PLIP: fully automated protein–ligand interaction profiler. Nucleic Acids Research 43, W443-W447 (2015).
10. A. Vangone, A. Bonvin, PRODIGY: A Contact-based Predictor of Binding Affinity in Protein-protein Complexes. Bio Protoc 7, e2124 (2017).
11. D. Van Der Spoel et al., GROMACS: Fast, flexible, and free. Journal Of Computational Chemistry 26, 1701-1718 (2005).
12. W. Hou, S. B. Cronin, A Review of Surface Plasmon Resonance-Enhanced Photocatalysis. Advanced Functional Materials 23, 1612-1619 (2013).

---

---

Posted by buchanle on Wednesday, June 3, 2026 in May 2026.

Tags: Acute Lymphoblastic Leukemia, Bispecific T-Cell Engager, Cancer Cell Hypoxia, CD22 Receptor, T-Cell Receptor