Phyloinformatics Lab

Anastasiia Duchenko passed her Qualifying Exams

Denis Jacob Machado February 04, 2026
qualifyingphdhonormillestonecellebrationdockingdrugai

Congratulations, Anastasiia!

Today, Feb. 4, 2026, our brilliant Ph.D. student, Anastasiia Duchenko, passed her qualifying exam, successfully completing a major milestone in her Ph.D. studies. We are all very proud of Anastasiia for her hard work, dedication, and commitment to science.

Below, we share a bit of what Anastasiia has been working on and her progress so far.

Developing cutting-edge AI-based methodologies to analyze pathogen evolution and improve public health approaches

Anastasiia Duchenko’s research focuses on developing cutting-edge AI-based methodologies to analyze pathogen evolution and improve public health approaches. The initial phase of her research addresses three critical challenges. First, viruses and other pathogens evolve rapidly, creating unexpected threats to public health; as clearly demonstrated during the COVID-19 pandemic, global preparedness for rapid viral evolution was limited. Second, there is a need for accessible, automated computational pipelines that provide reliable insights into pathogen evolution to guide vaccine and treatment development. Third, addressing health threats often requires extensive time and financial investments. While empirical evidence is invaluable and irreplaceable, automated computational pipelines can generate accurate predictions, reducing reliance on trial-and-error in vitro experimentation.

To address these challenges, she developed the automated AVOIDRUNE pipeline: Automated Viral Observation for Immune evasion and Docking - Resistance Unraveling in Neutralization Escape. AVOIDRUNE integrates bioinformatics tools to analyze molecular protein–protein interactions. The pipeline was developed and validated using SARS-CoV-2 as a model system due to the extensive availability of public structural and genomic data. AVOIDRUNE analyzes how single-nucleotide polymorphisms (SNPs) in viral proteins influence pathogen evolution and immune adaptation. The pipeline was demonstrated using SARS-CoV-2 Spike protein interactions with antibodies and with the host cell receptor. AVOIDRUNE is designed to be broadly applicable to other pathogens and to diverse protein–protein interaction studies.

The pipeline requires minimal input, relying primarily on empirical three-dimensional structures from the Protein Data Bank. For novel pathogens lacking experimental structures, protein sequences from the NCBI and GISAID databases were used to predict structures with AlphaFold. The first operational mode of the pipeline (antigen–antibody mode) analyzes interactions between the viral receptor-binding domain (RBD)—the Spike protein region responsible for host cell recognition and entry—and SARS-CoV-2-specific antibodies and controls. In the second mode (antigen–ligand mode), the viral RBD is analyzed against human angiotensin-converting enzyme 2 (hACE2), the primary receptor for viral entry.

AVOIDRUNE consists of three main modules: pre-processing, docking, and post-processing. During pre-processing, proteins of interest are prepared according to HADDOCK3 requirements. In the docking stage, the processed structures and configuration files are used to generate predicted binding poses for antibody–RBD and ACE2–RBD complexes across 20 viral variants, 50 antibodies, and 10 ACE2 structures. Docking experiments are executed automatically in multi-threaded mode on the UNC Charlotte High Performance Computing cluster. In the post-processing stage, predicted protein–protein complex structures generated by HADDOCK are analyzed with PRODIGY to calculate binding affinities, after which final scores and statistical summaries are recorded.

Using this approach, over 1,200 docking models were generated, including approximately 1,000 antibody–antigen and 200 ACE2–antigen complexes. Nonparametric statistical analyses (Kruskal–Wallis and pairwise Mann–Whitney U tests) were then applied to identify distinct immune escape strategies. The recently circulating JN.1 variant exhibited significantly lower antibody-binding affinity than the Delta variant (p < 0.05), suggesting immune escape. In contrast, the currently circulating LP.1.8 and B.A.3.2 variants showed increased antibody binding affinity, suggesting less efficient escape strategies. For ACE2 binding, only the Omicron and LP.1.8 variants showed statistically significant differences, with Omicron displaying reduced ACE2 binding efficiency and LP.1.8 showing increased binding. These computational predictions align with previously reported in vitro experimental results and epidemiological patterns, supporting the predictive accuracy of the pipeline. As a final validation step, molecular dynamics simulations (GROMACS 2024.2 MPI) were performed to evaluate the stability of predicted complexes over time.

The initial phase of this research was successfully completed, resulting in a containerized, multiplatform computational AVOIDRUNE pipeline (docker: phyloinformatics2022/avoidrune:latest). This pipeline addresses critical challenges in pathogen research by providing an automated, reliable, and reproducible computational framework that accelerates insights into rapid pathogen evolution while reducing time and resource demands in therapeutic development. This work forms the foundation of the first dissertation chapter, and a manuscript is currently in preparation for preprint submission. Future research will focus on advancing computational methodologies to understand pathogen adaptation and to improve therapeutic strategies, including expanding analytical frameworks, incorporating additional biological model systems, and establishing experimental collaborations for prospective validation.

References

  1. Alshahrani, M., Parikh, V., Foley, B., Raisinghani, N., & Verkhivker, G. (2025). Mutational Scanning and Binding Free Energy Computations of the SARS-CoV-2 Spike Complexes with Distinct Groups of Neutralizing Antibodies: Energetic Drivers of Convergent Evolution of Binding Affinity and Immune Escape Hotspots. International Journal of Molecular Sciences, 26(4), 1507.
  2. Chen, L., Kaku, Y., Okumura, K., Uriu, K., Zhu, Y., Ito, J., & Sato, K. (2025). Virological characteristics of the SARS-CoV-2 LP. 8.1 variant. The Lancet Infectious Diseases, 25(4), e193. 3. Giulini, M., Schneider, C., Cutting, D., Desai, N., Deane, C. M., & Bonvin, A. M. (2024). Towards the accurate modelling of antibody-antigen complexes from sequence using machine learning and information-driven docking. Bioinformatics, 40(10), btae583.
  3. Guo, C., Yu, Y., Liu, J., Jian, F., Yang, S., Song, W., … & Cao, Y. (2025). Antigenic and virological characteristics of SARS-CoV-2 variants BA. 3.2, XFG, and NB. 1.8.1. The Lancet Infectious Diseases.
  4. De Cae, S., Van Molle, I., van Schie, L., Shoemaker, S. R., Deckers, J., Debeuf, N., … & Schepens, B. (2025). Ultrapotent SARS coronavirus-neutralizing single-domain antibodies that clamp the spike at its base. Nature Communications, 16(1), 5040.
  5. Ma, Y., Liu, Q., Rehati, P., Zhang, Z., Wang, Y., Liu, Z., … & Huang, J. (2025). Structure-guided bispecific antibody engineering confers broad protection against KP. 3.1. 1 and sarbecoviruses. iScience, 28(8).
  6. Tong, Z., Tong, J., Lei, W., Xie, Y., Cui, Y., Jia, G., Li, S., Zhang, Z., Cheng, Z., Xing, X., et al. (2024). Deciphering a reliable synergistic bispecific strategy of rescuing antibodies for SARS-CoV-2 escape variants, including BA.2.86, EG.5.1, and JN.1. Cell Reports 43.

Know more about Anastasiia

Previous post
Denis was elected a Council Member of the Willi Hennig Society Next post