Resarch Field
Prof. Dr. Serdar DURDAĞI
I am a scientist dedicated to advancing our understanding of protein-protein and protein-ligand interactions, with a focus on designing computational frameworks to address complex biological challenges. My research bridges molecular biology, chemistry, and computational science, with applications in rational drug design, molecular modeling, protein engineering, and molecular simulations. My overarching goal is to develop effective therapeutic strategies by harnessing both theoretical and computational approaches to explore the behavior and interactions of biological molecules.
At the heart of my research is the development of novel computational methodologies, with an emphasis on machine learning-based models for drug discovery. My lab has pioneered innovative approaches for structure- and ligand-based virtual screening, enabling the efficient exploration of ultra-large molecular libraries to identify promising therapeutic candidates. In addition to our de novo drug design efforts, drug repurposing has become a key area of focus, as we explore existing drugs for new therapeutic indications, expediting the drug development pipeline.
My research team’s work extends beyond theoretical studies—collaboration with experimental research groups plays a crucial role in validating our findings. Through these partnerships, we integrate computational models with experimental data to address complex biological questions and develop reliable insights into disease mechanisms. This synergy allows us to contribute to the rational design of innovative therapeutics that target diseases with high unmet medical needs.
Over the years, my contributions to the scientific community have been recognized through multiple prestigious national and international awards, reflecting the impact of my work. These honors include:
• Turkish Academy of Sciences (TÜBA) Academy Award and Academy Medal (2023)
• Health Institutes of Türkiye (TÜSEB) Aziz Sancar Incentive Award (2017)
• The Scientific and Technological Research Institution of Türkiye (TÜBİTAK) Incentive Award (2016)
• Science Academy BAGEP Award (2014)
• Bahcesehir University (BAU) Respect for Science Awards:
◦ 25th Year Special Science Award (2023)
◦ Contribution to Science Award (2023)
◦ Contribution to Society Award (2021)
◦ Contribution to Science Award (2017)
In addition to my research achievements, I am honored to represent Türkiye on the international stage as the Ambassador of the Biophysical Society (BPS) for the 2024-2026 term. In this role, I aim to foster scientific exchange, interdisciplinary collaboration, and community engagement within the biophysics field, contributing to the global scientific dialogue.
Through my work, I remain committed to the development of computational solutions for biomedical challenges, providing a deeper understanding of biological systems and laying the foundation for the discovery of next-generation therapeutics.
Tool development: To translate these advances into impact, we have developed DRGSCROLL, a receptor-flexible docking engine that performs continuous χ-angle side-chain sampling and recurrent pose-receptor refinement to capture pocket plasticity and reduce false negatives in challenging targets; and mol2drug, a mobile AI application that converts molecular depictions (e.g., drawings or images of small molecules) into rapid predictions of activity/targets and drug-likeness. These platforms operationalize our methods for large-scale virtual screening and on-the-go decision support, and are actively used across our projects and collaborations.
To support reproducibility and community uptake, we maintain an actively updated GitHub organization—DURDAGILAB—hosting many useful open-access tools developed by our group. You can find example datasets, step-by-step tutorials, and ready-to-run notebooks on our GitHub page: https://github.com/DURDAGILAB. We welcome issues, forks, and contributions from collaborators and the wider community.
Here at DurdağıLab, we apply Molecular Mechanics (MM) and Molecular Dynamics (MD) simulations to reveal how biomolecules move, recognize ligands, and transmit signals. Using high-fidelity force fields, enhanced sampling, and free-energy methods (MM/PB(GB)SA, FEP, alchemical workflows), we map conformational landscapes, detect cryptic/allosteric pockets, and quantify binding affinities across kinases, proteases, scaffold proteins, ion channels, and membrane targets. These pipelines guide hit discovery and lead optimization by linking atomistic interactions to potency, selectivity, and resistance profiles.
Where classical models reach their limits, we integrate Quantum Mechanics (QM) and hybrid QM/MM methods to capture electronic effects-protonation/tautomer states, polarization, metal coordination, covalent warheads, and reaction pathways in enzyme active sites. Density Functional Theory (DFT) calculations refine stereoelectronics, reactivity, and spectral signatures, while QM/MM free-energy surfaces explain catalysis and inhibitor mechanism of action. Together, our MM→QM/MM continuum delivers decision-grade insights that de-risk candidates, rationalize SAR, and accelerate translation from in silico hypotheses to experimentally validated therapeutics.
Selected Publications:
-
Durdağı, S., Dağ, Ç., Doğan, B., Yiğin, M., Avşar, T., Büyükdağ, C., … DeMirci, H. (2022). The neutralization effect of montelukast on SARS-CoV-2 is shown by multiscale in silico simulations and combined in vitro studies. Molecular Therapy, 30(2), 963–974. https://doi.org/10.1016/j.ymthe.2021.10.014. PubMed
-
Calis, S., Doğan, B., Durdagi, S., Yapicier, O., Kılıç, T., Turanlı, E. T., & Avşar, T. (2022). A novel BH3 mimetic Bcl-2 inhibitor promotes autophagic cell death and reduces in vivo glioblastoma tumor growth. Cell Death Discovery, 8(1), 433. https://doi.org/10.1038/s41420-022-01225-9. PubMedNature
-
Rosenhouse-Dantsker, A., Noskov, S., Durdagi, S., Logothetis, D. E., & Levitan, I. (2013). Identification of novel cholesterol-binding regions in Kir2 channels. Journal of Biological Chemistry, 288(43), 31154–31164. https://doi.org/10.1074/jbc.M113.496117. PubMed
Tool development: To translate these advances into impact, we have developed DRGSCROLL, a receptor-flexible docking engine that performs continuous χ-angle side-chain sampling and recurrent pose-receptor refinement to capture pocket plasticity and reduce false negatives in challenging targets; and mol2drug, a mobile AI application that converts molecular depictions (e.g., drawings or images of small molecules) into rapid predictions of activity/targets and drug-likeness. These platforms operationalize our methods for large-scale virtual screening and on-the-go decision support, and are actively used across our projects and collaborations.
To support reproducibility and community uptake, we maintain an actively updated GitHub organization—DURDAGILAB—hosting many useful open-access tools developed by our group. You can find example datasets, step-by-step tutorials, and ready-to-run notebooks on our GitHub page: https://github.com/DURDAGILAB. We welcome issues, forks, and contributions from collaborators and the wider community.
We create novel, patentable chemotypes from scratch by uniting AI-driven generative models with physics-based design. Starting from target structures, e-pharmacophore hypotheses, and pocket dynamics, we use 3D-aware graph/diffusion generators, fragment growing, and scaffold hopping to propose molecules that satisfy shape/electrostatic complementarity and key interaction motifs. Flexible docking (e.g., our developed DRGSCROLL platform), MD, and free-energy calculations rapidly triage ideas and prioritize chemotypes with favorable binding energetics and selectivity profiles.
Design is closed-loop and multi-objective: potency, ADMET/PK, solubility, permeability, hERG/CYP liability, and synthetic accessibility are optimized simultaneously with uncertainty-aware active learning. Retrosynthesis planning and synthesizability scoring ensure practical routes, while QM/QM-MM refinements de-risk covalent or metal-binding mechanisms. Partnering with Istanbul MedChem and experimental collaborators, we iterate in silico → in vitro to accelerate first-in-class leads across oncology and beyond.
Selected Publications:
Ikram, S., Sayyah, E., & Durdağı, S. (2024). Identifying potential SOS1 inhibitors via virtual screening of multiple small molecule libraries against KRAS–SOS1 interaction. ChemBioChem, 25(12), e202400008. https://doi.org/10.1002/cbic.202400008 PubMed
-
Ekhteiari Salmas, R., Seeman, P., Stein, M., & Durdagi, S. (2018). Structural investigation of the dopamine-2 receptor agonist bromocriptine binding to dimeric D2HighR and D2LowR states. Journal of Chemical Information and Modeling, 58(4), 826–836. https://doi.org/10.1021/acs.jcim.7b00722 PubMed
-
Durdagi, S., Dogan, B., Erol, I., Kayık, G., & Aksoydan, B. (2019). Current status of multiscale simulations on GPCRs. Current Opinion in Structural Biology, 55, 93–103. https://doi.org/10.1016/j.sbi.2019.02.013 PubMed
We build experiment-ready protein models that capture the right biology, from sequence to full, environment-aware structures. Starting with high-confidence templates and AI predictions (e.g., AlphaFold), we curate domains, loops, oligomeric states, cofactors/metal sites, protonation/tautomer states, PTMs, and, when relevant, membrane embedding and glycosylation. Each model is stress-tested via energy minimization, rotamer/loop refinement, and validation metrics, then cross-checked against available cryo-EM/X-ray/NMR data to ensure geometric and stereochemical quality.
To connect structure to function, we simulate proteins across their conformational landscape with MD and enhanced sampling, map allosteric networks, and expose cryptic/induced pockets that classical static models miss. Flexible/ensemble docking and free-energy calculations quantify ligand recognition and selectivity, while in silico mutagenesis highlights resistance liabilities and mechanistic hotspots. These protein-centric insights anchor our de novo design, repurposing, and lead-optimization pipelines.
Selected publications (APA):
Kahveci, K., Düzgün, M. B., Atis, A. E., Yılmaz, A., Shahraki, A., Coskun, B., Durdagi, S., & Birgul Iyison, N. (2024). Discovering allatostatin type-C receptor specific agonists. Nature Communications, 15, 3965. https://doi.org/10.1038/s41467-024-48156-w. Nature
Durdagi, S., Dağ, Ç., Dogan, B., Yigin, M., Avsar, T., Buyukdag, C., … DeMirci, H. (2021). Near-physiological-temperature serial crystallography reveals conformations of SARS-CoV-2 main protease active site for improved drug repurposing. Structure, 29(12), 1382–1396.e6. https://doi.org/10.1016/j.str.2021.07.007. PubMed
Sahaboglu, A., Miranda, M., Canjuga, D., Avci-Adali, M., Savytska, N., Secer, E., … Durdagi, S. (2020). Drug repurposing studies of PARP inhibitors as a new therapy for inherited retinal degeneration. Cellular and Molecular Life Sciences, 77(11), 2199–2216. https://doi.org/10.1007/s00018-019-03283-2. Bahçeşehir University
We dissect how proteins work at the atomic and electronic levels by combining Molecular Mechanics/Molecular Dynamics (MM/MD) with Quantum Mechanics (QM) and hybrid QM/MM. Using enhanced sampling (e.g., metadynamics, umbrella sampling) and free-energy protocols, we map conformational pathways, allosteric networks, and protonation/tautomer shifts that govern recognition, catalysis, and gating. This MM layer reveals the relevant states and transition routes; the QM/QM-MM layer then resolves bond-making/breaking, metal coordination, polarization, and covalent warhead chemistry with DFT-quality accuracy.
In practice, flexible docking (i.e., DRGSCROLL) seeds catalytically competent or allosterically engaged poses, MD explores their stability, and QM/MM calculates reaction coordinates and barriers to pinpoint mechanism and resistance hotspots. We apply this workflow to kinases, proteases, ion channels, and nucleic-acid–associated proteins to explain SAR, anticipate mutation effects, and rationalize inhibitor MoA—delivering decision-grade mechanistic insights that directly guide design and experimental validation
Selected Publications:
-
Calis, S., Dogan, B., Durdagi, S., Celebi, A., Yapicier, O., Kilic, T., Tahir Turanli, E., & Avsar, T. (2022). A novel BH3 mimetic Bcl-2 inhibitor promotes autophagic cell death and reduces in vivo glioblastoma tumor growth. Cell Death Discovery, 8(1), 433. https://doi.org/10.1038/s41420-022-01225-9 PubMed
-
Durdagi, S., Deshpande, S., Duff, H. J., & Noskov, S. Y. (2012). Modeling of open, closed, and open-inactivated states of the hERG1 channel: Structural mechanisms of the state-dependent drug binding. Journal of Chemical Information and Modeling, 52(10), 2760–2774. https://doi.org/10.1021/ci300353u Bahçeşehir University
-
Wang, Y., Guo, J., Perissinotti, L. L., Lees-Miller, J., Teng, G., Durdagi, S., Duff, H. J., & Noskov, S. Y. (2016). Role of the pH in state-dependent blockade of hERG currents. Scientific Reports, 6, 32536. https://doi.org/10.1038/srep32536
We build a full-stack ML pipeline that converts raw oncology data into design-ready insights. Curated bioactivity matrices (single-point and dose–response), mutation annotations, and screening metadata are harmonized and de-duplicated, then fused with protein-aware features—pocket geometry, residue physicochemistry, dynamics-derived descriptors—and ligand representations that include classical cheminformatics as well as mathematically inspired features. Docking/MD outputs from our flexible engine (DRGSCROLL) provide physics-grounded pose and interaction fingerprints that enrich the feature space for structure-based learning.
On top of this foundation, we train graph neural networks (e.g., D-MPNN/Chemprop) and joint protein–ligand encoders to predict potency, selectivity, and resistance resilience across targets central to our Lab. Model development emphasizes rigor: stratified cross-validation, scaffold splits, calibration, and conformal prediction for well-quantified uncertainty. Attribution tools (feature/atom-level) and counterfactual analyses help us explain SAR drivers, highlight mutation-sensitive interactions, and generate testable hypotheses.
Design is closed-loop and multi-objective. Diffusion/transformer generators and fragment-growing agents propose novel, patentable chemotypes conditioned on pocket and pharmacophore constraints. We co-optimize efficacy with ADMET/PK (permeability, solubility, hERG/CYP, DDI risk) and synthetic tractability using retrosynthesis scores and route planners. Bayesian/active learning cycles prioritize compounds that are both informative and makeable; DRGSCROLL + MD + FEP/QM/MM triage candidates to ensure that high ML scores correspond to physically credible binding modes and energetics.
Safety-by-design is built in from the start. Multi-task toxicity classifiers (cardiotoxicity, hepatotoxicity), off-target GPCR/ion-channel panels, and metabolism/reactivity predictors run in parallel with efficacy models. Out-of-distribution and domain-applicability checks guard against spurious extrapolation when exploring novel chemical space, while uncertainty thresholds gate advancement to physics and synthesis stages.
Finally, we translate predictions into experiments with Istanbul MedChem and clinical collaborators. Prospective validation cycles (small, hypothesis-driven batches) feed back outcomes to recalibrate models, refine priors for resistant mutants, and lock in robust SAR. This iterative ML→physics→chemistry→biology loop accelerates lead identification, reduces false positives, and focuses resources on the highest-likelihood anticancer candidates.
Selected publications
-
Sayyah E., Oktay L., Tunç H., Durdağı S. Developing Dynamic Structure-Based Pharmacophore and ML-Trained QSAR Models for Novel Resistance-Free RET Tyrosine Kinase Inhibitors. ChemMedChem (2024). PubMed
-
Şahin K., Orhan M. D., Avşar T., Durdağı S. Hybrid In Silico and TR-FRET-Guided Discovery of Novel BCL-2 Inhibitors. ACS Pharmacology & Translational Science (2021). PMC
-
Çalış S., Doğan B., Durdağı S., et al. A novel BH3-mimetic BCL-2 inhibitor promotes autophagic cell death and reduces in vivo glioblastoma tumor growth. Cell Death Discovery (Nature Portfolio) (2022).
At DurdağıLab, we use pharmacophore modeling to distill the essential 3D interaction features that drive binding—hydrogen-bond donors/acceptors, hydrophobic/aromatic hotspots, and charged centers—into precise hypotheses for ligand recognition. We build both ligand-based and structure-based pharmacophores and apply e-pharmacophore workflows to encode energetic contributions of key contacts. These models guide de novo design, scaffold hopping, and rapid triage of ultra-large libraries for oncology and other therapeutic areas.
To capture receptor plasticity, we generate dynamic/ensemble pharmacophores from molecular dynamics trajectories, enabling recognition of induced-fit and cryptic pockets that static models miss. Hits prioritized by pharmacophore fit are iteratively refined with our flexible docking, ML-assisted ranking (e.g., our developed platform mol2drug), and MD stability checks before experimental follow-up. This integrated pipeline has helped us accelerate hit identification and focus synthesis on chemotypes with the highest likelihood of success.
Selected publications:
-
Sahin, K., Orhan, M. D., Avsar, T., & Durdagi, S. (2021). Hybrid in silico and TR-FRET-guided discovery of novel BCL-2 inhibitors. ACS Pharmacology & Translational Science, 4(3), 1111–1123. https://doi.org/10.1021/acsptsci.0c00210
-
Kanan, D., Kanan, T., Dogan, B., Orhan, M. D., Avsar, T., & Durdagi, S. (2021). An integrated in silico approach and in vitro study for the discovery of small-molecule USP7 inhibitors as potential cancer therapies. ChemMedChem, 16(3), 555–567. https://doi.org/10.1002/cmdc.202000675
-
Sayyah, E., Oktay, L., Tunc, H., & Durdagi, S. (2024). Developing dynamic structure-based pharmacophore and ML-trained QSAR models for the discovery of novel resistance-free RET tyrosine kinase inhibitors through extensive MD trajectories and NRI analysis. ChemMedChem, 19(12), e202300644. https://doi.org/10.1002/cmdc.202300644
Our Lab engineers proteins by coupling atomistic modeling with data-driven mutation design to rewire function, stability, and recognition. We build multi-state structural models (active/inactive/intermediate), map allosteric networks, and simulate dynamics under realistic environments to predict how specific substitutions reshape conformational equilibria and ligand or partner specificity. Target classes include GPCRs and ion channels, where state-dependent interactions and gating are central to function (Durdagi, Deshpande, Duff, & Noskov, 2012). PubMed
We translate these insights into practical designs: for GPCRs, we modeled activation-state determinants and engineered residues to modulate receptor–ligand coupling and signaling (Salmas, Yurtsever, Stein, & Durdagi, 2015); for CCR2, we performed focused protein-engineering analyses to probe selectivity and binding determinants (Salmas, Yurtsever, & Durdagi, 2016). Together, these studies illustrate our end-to-end workflow—from multiscale modeling to mutation prioritization—that informs experimental validation and guides rational redesign. ResearchGateKing s College London
-
Durdagi, S., Deshpande, S., Duff, H. J., & Noskov, S. Y. (2012). Modeling of open, closed, and open-inactivated states of the hERG1 channel: Structural mechanisms of state-dependent drug binding. Journal of Chemical Information and Modeling, 52(10), 2760–2774. https://doi.org/10.1021/ci300353u PubMed
-
Salmas, R. E., Yurtsever, M., Stein, M., & Durdagi, S. (2015). Modeling and protein engineering studies of active and inactive states of human dopamine D2 receptor (D2R) and investigation of drug/receptor interactions. Molecular Diversity, 19(2), 321–332. https://doi.org/10.1007/s11030-015-9598-4 ResearchGate
-
Salmas, R. E., Yurtsever, M., & Durdagi, S. (2016). Protein engineering studies for C-C chemokine receptor type 2 (CCR2). Current Enzyme Inhibition, 12(2), 110–114. https://doi.org/10.2174/1573408011666150807190410 King s College London
At DurdağıLab, we engineer hybrid algorithms that fuse ligand-based learning (pharmacophore/QSAR, graph ML, similarity metrics) with target-driven physics (receptor-flexible docking, MD/ensemble modeling, QM/MM refinements). Starting from e-pharmacophore hypotheses and pocket dynamics, our pipelines use uncertainty-aware active learning to iteratively propose, score, and filter candidates—balancing potency, selectivity, ADMET, and synthesizability. Custom tools that we developed (e.g., DRGSCROLL for flexible docking; Mol2Drug for ML-ranking) allow us to search ultra-large libraries while accounting for induced fit, allostery, and resistance-linked mutations.
These algorithms are validated end-to-end: hits prioritized in silico are stress-tested via MD/free-energy methods and then advanced to experimental assays with our partners. By unifying statistical patterns in known ligands with physics-grounded target models, we cut false positives, surface cryptic pockets, and accelerate de novo design and repurposing across oncology and beyond.
Selected publications:
-
Tutumlu, G., Doğan, B., Avşar, T., Orhan, M. D., Çalış, Ş., & Durdağı, S. (2020). Integrating ligand- and target-driven virtual screening with TR-FRET and cell-based assays to identify novel BCL-2 inhibitors. Frontiers in Chemistry, 8, 167. https://doi.org/10.3389/fchem.2020.00167. Frontiers
-
Çömert Önder, F., Durdağı, S., Şahin, K., Özpolat, B., & Ay, M. (2020). Design, synthesis, and molecular modeling studies of novel coumarin carboxamide derivatives as eEF-2K inhibitors. Journal of Chemical Information and Modeling, 60(3), 1766–1778. https://doi.org/10.1021/acs.jcim.9b01083. Bahçeşehir University
-
Kanan, T., Kanan, D., Erol, İ., Yazdi, S., Stein, M., & Durdağı, S. (2019). Targeting the NF-κB/IκBα complex via fragment-based e-pharmacophore screening and binary QSAR models. Journal of Molecular Graphics and Modelling, 86, 264–277. https://doi.org/10.1016/j.jmgm.2018.09.014. Bahçeşehir University
Our Lab predicts therapeutic activity by uniting ligand-based machine learning (binary/disease-focused QSAR) with cheminformatics and pathway-aware filters to prioritize compounds with disease-relevant efficacy. We have applied binary-QSAR models to prospect for antivirals and neuroactive scaffolds, demonstrating how model-guided screening rapidly narrows vast libraries to testable shortlists (Is, Durdagi, Aksoydan, & Yurtsever, 2018; Oktay et al., 2021). PubMedPMC
These predictions are integrated with structure-based docking, MD/MM-GBSA, and mechanism checks to ensure physicochemical plausibility and on-target action, feeding directly into hit triage and repurposing campaigns. The pipeline has progressed to experimental validation—for example, the BH3-mimetic BCL-2 inhibitor BAU-243, discovered through multistep virtual screening and confirmed to reduce glioblastoma growth in vivo, exemplifies how our in silico calls translate into therapeutic impact (Çalış et al., 2022). Nature
Selected Publications:
-
Çalış, Ş., Doğan, B., Durdağı, S., Çelebi, A., Yapıcıer, O., Kılıç, T., Tahir Turanlı, E., & Avşar, T. (2022). A novel BH3 mimetic Bcl-2 inhibitor promotes autophagic cell death and reduces in vivo glioblastoma tumor growth. Cell Death Discovery, 8, 433. https://doi.org/10.1038/s41420-022-01225-9 Nature
-
Is, Y. S., Durdagi, S., Aksoydan, B., & Yurtsever, M. (2018). Proposing novel MAO-B hit inhibitors using multidimensional molecular modeling and binary QSAR models for prediction of therapeutic, ADME and toxicity properties. ACS Chemical Neuroscience, 9(7), 1768–1782. https://doi.org/10.1021/acschemneuro.8b00095 PubMed
Our Lab builds fast, decision-grade ADME/Tox filters that pair multi-task machine learning with physics-based modeling. Early in the pipeline, we screen chemotypes for oral drug-likeness (solubility, permeability, pKa/logP, TPSA), metabolism and clearance proxies (CYP liability, P-gp efflux), and safety risks (Ames, hepatotoxicity, PAINS/reactivity). For cardiotoxicity, we integrate ligand- and structure-based assessment of hERG block—an area where our group helped define pharmacophore, docking, and MD strategies that anticipate off-target channel binding before synthesis. PubMed
Beyond binary “safe/unsafe” flags, we analyze state-dependent and pH-dependent channel blockade to rationalize false positives and guide scaffold rehabilitation. This enables us to redesign problematic leads (e.g., long-QT promoters) into hERG-neutral analogs while retaining on-target potency, and to prioritize series least likely to encounter late-stage attrition in vivo. PubMed+1
-
Durdagi, S., Duff, H. J., & Noskov, S. Y. (2011). Combined receptor and ligand-based approach to the universal pharmacophore model development for studies of drug blockade to the hERG1 pore domain. Journal of Chemical Information and Modeling, 51(2), 463–474. https://doi.org/10.1021/ci100409y PubMed
-
Durdagi, S., Randall, T., Duff, H. J., Chamberlin, A., & Noskov, S. Y. (2014). Rehabilitating drug-induced long-QT promoters: In-silico design of hERG-neutral cisapride analogues with retained pharmacological activity. BMC Pharmacology and Toxicology, 15, 14. https://doi.org/10.1186/2050-6511-15-14 PubMed
-
Wang, Y., Guo, J., Perissinotti, L. L., Lees-Miller, J., Teng, G., Durdagi, S., Duff, H. J., & Noskov, S. Y. (2016). Role of the pH in state-dependent blockade of hERG currents. Scientific Reports, 6, 32536. https://doi.org/10.1038/srep32536 PubMed
At DurdağıLab, we systematically reposition approved and clinical-stage molecules by combining large-scale virtual screening with physics-based evaluation and rapid experimental follow-up. Using e-pharmacophore filters, flexible/ensemble docking, MD, and free-energy methods (MM/PB(GB)SA, FEP), we triage compound libraries against validated and emerging targets, including viral proteases and neuro-/oncology proteins. We integrate network-level insights and ADMET/PK constraints early, then iterate in silico → in vitro to confirm mechanism, potency, and selectivity—shortening timelines and de-risking translation.
Our repurposing pipeline has been applied to urgent public-health targets and rare diseases alike: serial crystallography and multiscale simulations informed selection/optimization of clinically available agents; and repositioned PARP inhibitors were evaluated for inherited retinal degeneration. These efforts illustrate how structure-guided, simulation-driven repurposing can rapidly surface testable therapies with known safety profiles.
Selected publications:
-
Durdağı, S., Dağ, Ç., Doğan, B., Yiğin, M., Avşar, T., Büyükdağ, C., … DeMirci, H. (2021). Near-physiological-temperature serial crystallography reveals conformations of SARS-CoV-2 main protease active site for improved drug repurposing. Structure, 29(12), 1382–1396.e6. https://doi.org/10.1016/j.str.2021.07.007
-
Durdağı, S., Dağ, Ç., Doğan, B., Yiğin, M., Avşar, T., Büyükdağ, C., … Yılmaz, S. (2022). The neutralization effect of montelukast on SARS-CoV-2 is shown by multiscale in silico simulations and combined in vitro studies. Molecular Therapy, 30(2), 963–974. https://doi.org/10.1016/j.ymthe.2021.10.014
-
Sahaboglu, A., Miranda, M., Canjuga, D., Avci-Adali, M., Savytska, N., Secer, E., … Durdagi, S. (2020). Drug repurposing studies of PARP inhibitors as a new therapy for inherited retinal degeneration. Cellular and Molecular Life Sciences, 77(11), 2199–2216. https://doi.org/10.1007/s00018-019-03283-2
We design cascade workflows that make billion-scale chemical space tractable. Fast AI/chemoinformatics filters and e-pharmacophore hypotheses narrow ultra-large vendor/ZINC libraries, then our receptor-flexible docking platform (e.g., DRGSCROLL) explores pocket plasticity with continuous side-chain sampling. Short/long MD/MM-GBSA and consensus rescoring promote true binders while retiring frequent false-positives. The result is a high-precision funnel that preserves diversity yet surfaces synthesis-ready hits.
Our group has applied these strategies across diverse targets—from enzyme active sites to challenging PPIs—by combining HTVS of multi-million libraries with physics-based refinement and targeted experimental follow-up. Representative studies include a ∼7-million-compound screen against CA IX, a hybrid algorithm for extensive-library screening of the ERCC1/XPF PPI, and a COVID-19 drug-repurposing pipeline that docked and dynamically rescored thousands of approved/clinical compounds.
Selected Publications:
-
Salmas, R. E., Senturk, M., Yurtsever, M., & Durdagi, S. (2016). Discovering novel carbonic anhydrase type IX (CA IX) inhibitors from seven million compounds using virtual screening and in vitro analysis. Journal of Enzyme Inhibition and Medicinal Chemistry, 31(3), 425–433. PubMed
-
Ghazy, S., Oktay, L., & Durdaği, S. (2024). A novel algorithm for the virtual screening of extensive small-molecule libraries against ERCC1/XPF protein–protein interaction for the identification of resistance-bypassing potential anticancer molecules. Turkish Journal of Biology, 48(2), 91–111. PubMed
-
Durdağı, S., Dağ, Ç., Doğan, B., Yiğin, M., Avşar, T., Büyükdağ, C., … Yılmaz, S. (2022). The neutralization effect of montelukast on SARS-CoV-2 is shown by multiscale in silico simulations and combined in vitro studies. Molecular Therapy, 30(2), 963–974. https://doi.org/10.1016/j.ymthe.2021.10.014
Our Lab develops predictive 3D/4D QSAR models that connect molecular structure to biological activity with high interpretability. In 3D QSAR, we employ ligand-based and receptor-aware alignments to extract steric, electrostatic, H-bonding, and hydrophobic fields that explain potency trends and guide chemical optimization. We pair these models with docking/MD to ensure bioactive conformations and binding modes are consistent with the target’s physics, enabling reliable SAR rationalization and prospective design (Durdagi et al., 2008; Durdagi et al., 2007). PubMedAmerican Chemical Society Publications
Extending to 4D QSAR, we encode conformational ensembles as the “fourth dimension,” capturing alignment and flexibility effects that classical single-conformer models miss. This ensemble-based strategy improves generalization across chemotypes and supports pharmacophore extraction for scaffold hopping and lead optimization, as demonstrated in our combined 4D-QSAR/target-based campaigns on bioactive isatin derivatives (Şahin, Sarıpınar, & Durdağı, 2021). PubMed
-
Durdagi, S., Mavromoustakos, T., Chronakis, N., & Papadopoulos, M. G. (2008). Computational design of novel fullerene analogues as potential HIV-1 PR inhibitors: Analysis of the binding interactions between fullerene inhibitors and HIV-1 PR residues using 3D QSAR, molecular docking and molecular dynamics simulations. Bioorganic & Medicinal Chemistry, 16(23), 9957–9974. https://doi.org/10.1016/j.bmc.2008.10.039 PubMed
-
Durdagi, S., Kapou, A., Kourouli, T., Andreou, T., Nikas, S. P., Nahmias, V. R., Papahatjis, D. P., Papadopoulos, M. G., & Mavromoustakos, T. (2007). The application of 3D-QSAR studies for novel cannabinoid ligands substituted at the C1′ position of the alkyl side chain on the structural requirements for binding to cannabinoid receptors CB1 and CB2. Journal of Medicinal Chemistry, 50(12), 2875–2885. https://doi.org/10.1021/jm0610705 American Chemical Society Publications
-
Şahin, K., Sarıpınar, E., & Durdağı, S. (2021). Combined 4D-QSAR and target-based approaches for the determination of bioactive isatin derivatives. SAR and QSAR in Environmental Research, 32(10), 769–792. https://doi.org/10.1080/1062936X.2021.1971760 PubMed
Our Lab treats docking as the decision engine of structure-based design. We combine fast, target-aware prescreening with receptor-flexible docking to capture pocket plasticity and side-chain rearrangements that simple rigid protocols miss. Multi-precision funnels (HTVS → focused docking → physics-based rescoring/short MD) reduce false positives while preserving chemotype diversity—an approach we’ve validated on challenging systems including protein–protein interfaces. PMC
We apply this pipeline end-to-end: from ultra-large screening against enzymatic targets (e.g., CA IX) to safety-critical ion channels (hERG) where docking guides liability mitigation and scaffold redesign. These studies demonstrate how carefully parameterized docking—paired with pharmacophore/ML filters and MD refinement—prioritizes synthesis-ready hits and explains SAR/mechanism at atomic resolution. PubMedAmerican Chemical Society Publications
Selected Publications:
-
Oktay, L., Erdemoğlu, E., Tolu, İ., Yumak, Y., Özcan, A., Acar, E., Büyükkılıç, Ş., Olkan, A., & Durdağı, S. (2021). Binary-QSAR guided virtual screening of FDA-approved drugs and compounds in clinical investigation against SARS-CoV-2 main protease. Turkish Journal of Biology, 45(4), 459–468. https://doi.org/10.3906/biy-2106-61 PMC
-
Kanan, T., Kanan, D., Erol, İ., Yazdi, S., Stein, M., & Durdağı, S. (2019). Targeting the NF-κB/IκBα complex via fragment-based e-pharmacophore virtual screening and binary QSAR models. Journal of Molecular Graphics and Modelling, 86, 264–277. https://doi.org/10.1016/j.jmgm.2018.09.014 PubMed
-
Fidan, İ., Salmas, R. E., Arslan, M., Şentürk, M., Durdaği, S., Ekinci, D., Şentürk, E., Coşgun, S., & Supuran, C. T. (2015). Carbonic anhydrase inhibitors: Design, synthesis, kinetic, docking and molecular dynamics analysis of novel glycine and phenylalanine sulfonamide derivatives. Bioorganic & Medicinal Chemistry, 23(23), 7353–7358. https://doi.org/10.1016/j.bmc.2015.10.009 PubMed
-
Cavdar, H., Şentürk, M., Güney, M., Durdaği, S., Kayık, G., Supuran, C. T., & Ekinci, D. (2019). Inhibition of acetylcholinesterase and butyrylcholinesterase with uracil derivatives: Kinetic and computational studies. Journal of Enzyme Inhibition and Medicinal Chemistry, 34(1), 429–437. https://doi.org/10.1080/14756366.2018.1543288 PMC
-
Zengin Kurt, B., Sönmez, F., Durdaği, S., Aksoydan, B., Ekhteiari Salmas, R., Angeli, A., Küçükislamoğlu, M., & Supuran, C. T. (2017). Synthesis, biological activity and multiscale molecular modeling studies for coumaryl-carboxamide derivatives as selective carbonic anhydrase IX inhibitors. Journal of Enzyme Inhibition and Medicinal Chemistry, 32(1), 1042–1052. https://doi.org/10.1080/14756366.2017.1354857 PubMed
-
Durdagi, S., Korkmaz, N., Işık, S., Vullo, D., Astley, D., Ekinci, D., Salmas, R. E., Şentürk, M., & Supuran, C. T. (2016). Kinetic and docking studies of cytosolic/tumor-associated carbonic anhydrase isozymes I, II and IX with some hydroxylic compounds. Journal of Enzyme Inhibition and Medicinal Chemistry, 31(6), 1214–1220. https://doi.org/10.3109/14756366.2015.1114930 PubMed
At DurdağıLab, we simulate biomolecular systems across scales to capture both atomistic detail and mesoscale behavior. All-atom MD (AA) lets us resolve hydrogen-bonding, ion pairing, water networks, and side-chain rotamer dynamics that govern binding and catalysis. Coarse-grained MD (CG; e.g., Martini-style mappings) enables micro- to millisecond exploration of large assemblies—membranes, oligomers, GPCR dimers—so we can chart long-range allostery, lipid-protein coupling, and pocket opening events inaccessible to AA alone. We routinely combine the two: CG to map global conformational landscapes and identify metastable states, followed by AA “back-mapping,” enhanced sampling, and free-energy calculations (MM/GB(P)SA, FEP) to quantify mechanisms and ligand selectivity.
This multiscale workflow underpins our target-class work—kinases, proteases, ion channels, and GPCRs—and feeds directly into design: cryptic/induced pockets from CG→AA funnels guide pharmacophore hypotheses, flexible docking, and stability refinement, while in silico mutagenesis and network analysis pinpoint resistance liabilities and allosteric control points.
Selected publications:
-
Durdagi, S., Doğan, B., Erol, İ., Kayık, G., & Aksoydan, B. (2019). Current status of multiscale simulations on GPCRs. Current Opinion in Structural Biology, 55, 93–103. https://doi.org/10.1016/j.sbi.2019.02.013
-
Ekhteiari Salmas, R., Seeman, P., Stein, M., & Durdagi, S. (2018). Structural investigation of the dopamine-2 receptor agonist bromocriptine binding to dimeric D2HighR and D2LowR states. Journal of Chemical Information and Modeling, 58(4), 826–836. https://doi.org/10.1021/acs.jcim.7b00722
-
Kahveci, K., Düzgün, M. B., Atis, A. E., Yılmaz, A., Shahraki, A., Coşkun, B., Durdağı, S., & Iyison, N. B. (2024). Discovering allatostatin type-C receptor specific agonists. Nature Communications, 15, 3965. https://doi.org/10.1038/s41467-024-48156-w
At DurdağıLab, we interrogate how small molecules engage proteins and traverse/reshape membranes. Using multiscale MD (all-atom → coarse-grained), enhanced sampling, and free-energy methods (FEP, MM/GBSA/PPBSA, PMFs via umbrella sampling), we resolve hydrogen-bond networks, water/ion mediation, lipid coupling, and conformational selection that govern affinity, kinetics, and selectivity at orthosteric and allosteric sites.
Because lipid composition modulates receptor behavior, we model membrane partitioning, permeability pathways, and protein–lipid crosstalk (e.g., cholesterol/PIP2 effects) to capture phenomena that static structures miss. These insights feed our design loop—guiding e-pharmacophore hypotheses, flexible docking (DRGSCROLL), and stability filters—to prioritize chemotypes that work in the real, membrane-embedded biology.
Selected publications:
-
Rosenhouse-Dantsker, A., Noskov, S., Durdagi, S., Logothetis, D. E., & Levitan, I. (2013). Identification of novel cholesterol-binding regions in Kir2 channels. Journal of Biological Chemistry, 288(43), 31154–31164. https://doi.org/10.1074/jbc.M113.496117
-
Durdağı, S., Doğan, B., Erol, İ., Kayık, G., & Aksoydan, B. (2019). Current status of multiscale simulations on GPCRs. Current Opinion in Structural Biology, 55, 93–103. https://doi.org/10.1016/j.sbi.2019.02.013
-
Kahveci, K., Düzgün, M. B., Atis, A. E., Yılmaz, A., Shahraki, A., Coşkun, B., Durdağı, S., & Iyison, N. B. (2024). Discovering allatostatin type-C receptor specific agonists. Nature Communications, 15, 3965. https://doi.org/10.1038/s41467-024-48156-w
We translate computational discoveries into therapeutic leads by uniting AI-guided design, receptor-flexible docking, and short MD/free-energy refinement with rapid experimental triage. Our pipelines prioritize potency, selectivity, and developability from the outset, enabling fast progression from virtual hits to mechanism-anchored chemotypes for oncology, antiviral, and enzyme-targeted indications.
Representative successes include ultra-large screening to identify selective CA IX inhibitors for tumor microenvironment modulation, a hybrid ligand/structure-based strategy yielding small-molecule disruptors of the ERCC1/XPF DNA-repair interface (chemo-sensitization), and a COVID-19 repurposing workflow that nominated clinically tested compounds against Mpro and Spike/ACE2. These efforts exemplify our end-to-end path from in silico design to therapy-relevant candidates.
-
Kanan, T., Kanan, D., Erol, İ., Yazdi, S., Stein, M., & Durdaği, S. (2019). Targeting the NF-κB/IκBα complex via fragment-based e-pharmacophore virtual screening and binary QSAR models. Journal of Molecular Graphics and Modelling, 86, 264–277. PubMed
-
Zengin Kurt, B., Sönmez, F., Durdaği, S., Aksoydan, B., Ekhteiari Salmas, R., Angeli, A., Küçükislamoğlu, M., & Supuran, C. T. (2017). Synthesis, biological activity and multiscale molecular modeling studies for coumaryl-carboxamide derivatives as selective carbonic anhydrase IX inhibitors. Journal of Enzyme Inhibition and Medicinal Chemistry, 32(1), 1042–1052. PubMed
-
Cavdar, H., Şentürk, M., Güney, M., Durdaği, S., Kayık, G., Supuran, C. T., & Ekinci, D. (2019). Inhibition of acetylcholinesterase and butyrylcholinesterase with uracil derivatives: Kinetic and computational studies. Journal of Enzyme Inhibition and Medicinal Chemistry, 34(1), 429–437. PMC
-
Oktay, L., Erdemoğlu, E., Tolu, İ., Yumak, Y., Özcan, A., Acar, E., Büyükkılıç, Ş., Olkan, A., & Durdaği, S. (2021). Binary-QSAR guided virtual screening of FDA-approved drugs and compounds in clinical investigation against SARS-CoV-2 main protease. Turkish Journal of Biology, 45(4), 459–468. PMC
-
Durdagi, S., Duff, H. J., & Noskov, S. Y. (2011). Combined receptor- and ligand-based approach to a universal pharmacophore model for hERG1 pore-domain blockade. Journal of Chemical Information and Modeling, 51(2), 463–474. PubMed
-
Durdagi, S., Randall, T., Duff, H. J., Chamberlin, A., & Noskov, S. Y. (2014). Rehabilitating drug-induced long-QT promoters: In-silico design of hERG-neutral cisapride analogues with retained pharmacological activity. BMC Pharmacology and Toxicology, 15, 14. PMC
Our lab uses atomistic molecular dynamics (MD)—augmented with enhanced sampling and rigorous free-energy methods—to map conformational landscapes, allosteric communication, gating, and protonation/tautomer shifts that govern recognition and catalysis. We simulate proteins in realistic environments (explicit solvent, ions, membranes) and couple MD with docking and QM/MM to resolve state-dependent binding, transition states, and resistance-relevant mutations, converting mechanistic insight into actionable design hypotheses (Durdagi, Deshpande, Duff, & Noskov, 2012; Salmas, Yurtsever, Stein, & Durdagi, 2015).
These simulation workflows directly support discovery and optimization: we quantify binding pathways and ΔG using MM/GBSA and path-based methods, prioritize chemotypes based on stability and kinetics, and validate predictions against electrophysiology and in vitro assays. Applications span ion channels, GPCRs, and apoptosis regulators—for example, our simulations helped guide the discovery and characterization of the BH3-mimetic BCL-2 inhibitor BAU-243 with in vivo efficacy (Çalış et al., 2022).
-
Çalış, Ş., Doğan, B., Durdağı, S., Çelebi, A., Yapıcıer, O., Kılıç, T., Tahir Turanlı, E., & Avşar, T. (2022). A novel BH3 mimetic Bcl-2 inhibitor promotes autophagic cell death and reduces in vivo glioblastoma tumor growth. Cell Death Discovery, 8, 433. https://doi.org/10.1038/s41420-022-01225-9
-
Durdagi, S., Deshpande, S., Duff, H. J., & Noskov, S. Y. (2012). Modeling of open, closed, and open-inactivated states of the hERG1 channel: Structural mechanisms of state-dependent drug binding. Journal of Chemical Information and Modeling, 52(10), 2760–2774. https://doi.org/10.1021/ci300353u
-
Salmas, R. E., Yurtsever, M., Stein, M., & Durdagi, S. (2015). Modeling and protein engineering studies of active and inactive states of human dopamine D2 receptor (D2R) and investigation of drug/receptor interactions. Molecular Diversity, 19(2), 321–332. https://doi.org/10.1007/s11030-015-9598-4
At DurdağıLab, we design small molecules that disrupt disease-relevant PPIs by marrying ligand-based learning (pharmacophore/QSAR) with target-driven physics (ensemble/flexible docking, MD, free-energy calculations). We map interface “hot spots,” encode them into e-pharmacophore hypotheses, and explore chemotypes via fragment growing and scaffold hopping to mimic key secondary-structure motifs (e.g., BH3 helices) or wedge into cryptic pockets revealed by dynamics. Prioritized hits are stress-tested with MD and MM/PB(GB)SA/FEP, then validated experimentally (e.g., TR-FRET), closing the loop from interface physics to measurable inhibition.
Our pipeline has delivered PPI disruptors across oncology signaling axes—KRAS–SOS1, NF-κB/IκBα, and anti-apoptotic BCL-2—demonstrating that careful interface modeling plus active-learning triage can turn “undruggable” surfaces into tractable medicinal chemistry campaigns.
Selected publications:
-
Ikram, S., Sayyah, E., & Durdağı, S. (2024). Identifying potential SOS1 inhibitors via virtual screening of multiple small-molecule libraries against the KRAS–SOS1 interaction. ChemBioChem, 25(12), e202400008. https://doi.org/10.1002/cbic.202400008
-
Kanan, T., Kanan, D., Erol, İ., Yazdi, S., Stein, M., & Durdağı, S. (2019). Targeting the NF-κB/IκBα complex via fragment-based e-pharmacophore screening and binary QSAR models. Journal of Molecular Graphics and Modelling, 86, 264–277. https://doi.org/10.1016/j.jmgm.2018.09.014
-
Şahin, K., Orhan, M. D., Avşar, T., & Durdağı, S. (2021). Hybrid in silico and TR-FRET-guided discovery of novel BCL-2 inhibitors. ACS Pharmacology & Translational Science, 4(3), 1111–1123. https://doi.org/10.1021/acsptsci.0c00210
Our lab delivers end-to-end CADD by integrating structure-based and ligand-based pipelines. We curate biology-faithful targets, generate ensemble or induced-fit conformations, and perform flexible/ensemble docking followed by atomistic MD and free-energy ranking to capture state-dependent recognition and prioritize chemotypes. In parallel, we train binary/continuous QSAR and ML models to predict potency, selectivity, and ADME/Tox, then fuse these readouts into multi-parameter optimization for decision-grade triage.
This workflow has translated to experimentally validated outcomes across indications. Examples include binary-QSAR–guided discovery for CNS targets, pandemic-scale repurposing against viral proteases, and multistep virtual screening that yielded the BH3-mimetic BCL-2 inhibitor BAU-243 with in vivo efficacy—illustrating how our computational hypotheses progress to mechanistic and therapeutic impact.
Selected Publications:
-
Durdagi, S., Deshpande, S., Duff, H. J., & Noskov, S. Y. (2012). Modeling of open, closed, and open-inactivated states of the hERG1 channel: Structural mechanisms of state-dependent drug binding. Journal of Chemical Information and Modeling, 52(10), 2760–2774. PubMed
-
Wang, Y., Guo, J., Perissinotti, L. L., Lees-Miller, J., Teng, G., Durdagi, S., Duff, H. J., & Noskov, S. Y. (2016). Role of the pH in state-dependent blockade of hERG currents. Scientific Reports, 6, 32536. Nature
-
Salmas, R. E., Yurtsever, M., Stein, M., & Durdagi, S. (2015). Modeling and protein engineering studies of active and inactive states of human dopamine D2 receptor (D2R) and investigation of drug/receptor interactions. Molecular Diversity, 19(2), 321–332. SpringerLink
-
Durdagi, S., Mavromoustakos, T., Chronakis, N., & Papadopoulos, M. G. (2008). Computational design of novel fullerene analogues as potential HIV-1 PR inhibitors: 3D QSAR, docking and MD analyses. Bioorganic & Medicinal Chemistry, 16(23), 9957–9974. PubMed
-
Durdagi, S., Kapou, A., Kourouli, T., Andreou, T., Nikas, S. P., Nahmias, V. R., Papahatjis, D. P., Papadopoulos, M. G., & Mavromoustakos, T. (2007). The application of 3D-QSAR studies for novel cannabinoid ligands at CB1/CB2 receptors. Journal of Medicinal Chemistry, 50(12), 2875–2885. PubMed
-
Is, Y. S., Durdagi, S., Aksoydan, B., & Yurtsever, M. (2018). Proposing novel MAO-B hit inhibitors using multidimensional molecular modeling and binary QSAR models. ACS Chemical Neuroscience, 9(7), 1768–1782. PubMed
-
Kanan, D., Kanan, T., Dogan, B., Orhan, M. D., Avsar, T., & Durdagi, S. (2021). An integrated in silico approach and in vitro study for the discovery of small-molecule USP7 inhibitors. ChemMedChem, 16(3), 555–567. PubMed
-
Tutumlu, G., Dogan, B., Avsar, T., Orhan, M. D., Calis, Ş., & Durdagi, S. (2020). Integrating ligand- and target-driven virtual screening with TR-FRET assays to identify BCL-2 hits. Frontiers in Chemistry, 8, 167. PMC
-
Dogan, B., & Durdagi, S. (2021). Drug re-positioning studies for novel HIV-1 inhibitors using binary QSAR models and multi-target-driven in silico studies. Molecular Informatics, 40(2), 2000012. PubMed
-
Çalış, Ş., Doğan, B., Durdağı, S., Çelebi, A., Yapıcıer, O., Kılıç, T., Tahir Turanlı, E., & Avşar, T. (2022). A novel BH3-mimetic BCL-2 inhibitor (BAU-243) promotes autophagic cell death and reduces in vivo glioblastoma growth. Cell Death Discovery, 8, 433. Nature