Ozlem Keskin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Ozlem Keskin has a total of 161 co-authors. Here are the top co-authors by number of shared publications:
- Attila Gursoy
Koc University
74 shared publications - Ruth Nussinov
Frederick National Laboratory for Cancer Research
19 shared publications - Nurcan Tuncbag
Koc University
15 shared publications Ruth Nussinov
Computational Structural Biology Section, Frederick National Laboratory for Cancer Research in the Laboratory of Cancer Immunometabolism, National Cancer Institute, Frederick, Maryland 21702, United States
15 shared publicationsHyunbum Jang
Computational Structural Biology Section, Frederick National Laboratory for Cancer Research in the Cancer Innovation Laboratory, National Cancer Institute
7 shared publications- E. Sila Ozdemir
Department of Chemical and Biological Engineering, Koç University, Istanbul 34450, Turkey
6 shared publications
Latest Publications
- Open Access
- 22 October 2026
- Published by American Chemical Society (ACS) in Biochemistry
Fuzzy Interactions Open Perspectives toward Drugging Conformational Heterogeneity
Fuzzy interactions enable specific biomolecular recognition while maintaining conformational freedom of the interacting partners. Here, we propose that fuzziness can be exploited for allosteric regulation to provide context-specific functional outputs. Using receptor signaling as an illustrative example, we demonstrate how heterogeneous binding modes with ligands can be fine-tuned through transient interactions. Fuzzy interactions interfere with alternative positions or modulate the mobility to mimic the landscapes of specific functional states that are linked with certain environmental conditions. Furthermore, using allosteric drugs targeting the ABL1 tyrosine kinase, KRAS oncogene, and IDH2 enzyme as examples, we demonstrate how fuzzy interaction mechanisms can affect allosteric couplings in a functionally meaningful manner through rewiring heterogeneous contact patterns. These notions could be helpful for drug development, targeting dynamic conformational ensembles and biasing them toward context-specific functional states.
- 22 October 2026
- Published by American Chemical Society (ACS) in Biochemistry
Open AccessFuzzy Interactions Open Perspectives toward Drugging Conformational Heterogeneity
Fuzzy interactions enable specific biomolecular recognition while maintaining conformational freedom of the interacting partners. Here, we propose that fuzziness can be exploited for allosteric regulation to provide context-specific functional outputs. Using receptor signaling as an illustrative example, we demonstrate how heterogeneous binding modes with ligands can be fine-tuned through transient interactions. Fuzzy interactions interfere with alternative positions or modulate the mobility to mimic the landscapes of specific functional states that are linked with certain environmental conditions. Furthermore, using allosteric drugs targeting the ABL1 tyrosine kinase, KRAS oncogene, and IDH2 enzyme as examples, we demonstrate how fuzzy interaction mechanisms can affect allosteric couplings in a functionally meaningful manner through rewiring heterogeneous contact patterns. These notions could be helpful for drug development, targeting dynamic conformational ensembles and biasing them toward context-specific functional states.
- Open Access
- 14 October 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 66 (20) , 13641-13651
PDI-Spot: A Robust Representation Learning Framework for Identifying Energetic Hotspots in Protein–DNA Complexes
The binding affinity of protein–DNA complexes is concentrated within a subset of interface residues, called “hotspots”, which are critical for maintaining stable interactions. Therefore, identifying these energetic anchors is essential for elaborating on how mutations disrupt transcriptional regulation and for identifying druggable regions within disease-associated interfaces. In this study, we introduce our novel prediction model, PDI-Spot, that leverages representation learning to capture the complex features of protein–DNA interfaces. Unlike traditional methods that rely on limited, handcrafted descriptors, PDI-Spot extracts high-dimensional, structure- and sequence-driven representations to achieve better predictive performance. We trained the PDI-Spot model on a curated protein–DNA complex data set and rigorously validated its performance on an independent test set to ensure generalizability across different structural families. Following the extraction of residue-level embeddings, an optimal feature set was identified through a combined approach of minimum redundancy maximum relevance and multilayer perceptron architectures. By training a neural network on the optimal feature set, we developed a robust predictive framework for hotspot identification. On an independent test set, PDI-Spot achieved an AUROC of 0.809, an F1-score of 0.667, and a Matthews correlation coefficient of 0.625. We also verified via permutation tests that PDI-Spot captures biologically meaningful patterns beyond random chance. PDI-Spot consistently demonstrates competitive performance compared to state-of-the-art methods, providing a reliable model for mapping the energetic landscapes of protein–DNA interfaces. The source code and pretrained models are publicly available at https://github.com/ebsrn0/pdi-spot.git.
- 14 October 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Information and Modeling
- Vol. 66 (20) , 13641-13651
Open AccessPDI-Spot: A Robust Representation Learning Framework for Identifying Energetic Hotspots in Protein–DNA Complexes
The binding affinity of protein–DNA complexes is concentrated within a subset of interface residues, called “hotspots”, which are critical for maintaining stable interactions. Therefore, identifying these energetic anchors is essential for elaborating on how mutations disrupt transcriptional regulation and for identifying druggable regions within disease-associated interfaces. In this study, we introduce our novel prediction model, PDI-Spot, that leverages representation learning to capture the complex features of protein–DNA interfaces. Unlike traditional methods that rely on limited, handcrafted descriptors, PDI-Spot extracts high-dimensional, structure- and sequence-driven representations to achieve better predictive performance. We trained the PDI-Spot model on a curated protein–DNA complex data set and rigorously validated its performance on an independent test set to ensure generalizability across different structural families. Following the extraction of residue-level embeddings, an optimal feature set was identified through a combined approach of minimum redundancy maximum relevance and multilayer perceptron architectures. By training a neural network on the optimal feature set, we developed a robust predictive framework for hotspot identification. On an independent test set, PDI-Spot achieved an AUROC of 0.809, an F1-score of 0.667, and a Matthews correlation coefficient of 0.625. We also verified via permutation tests that PDI-Spot captures biologically meaningful patterns beyond random chance. PDI-Spot consistently demonstrates competitive performance compared to state-of-the-art methods, providing a reliable model for mapping the energetic landscapes of protein–DNA interfaces. The source code and pretrained models are publicly available at https://github.com/ebsrn0/pdi-spot.git.
- 18 May 2026
- Published by Elsevier in Journal of Molecular Biology
- Vol. 438 (18) , 169863
DeepAlloWeb: a web server for interactive allosteric pockets prediction using protein language model
- 18 May 2026
- Published by Elsevier in Journal of Molecular Biology
- Vol. 438 (18) , 169863
DeepAlloWeb: a web server for interactive allosteric pockets prediction using protein language model
- Open Access
- 11 March 2026
- Published by Frontiers Media SA in Frontiers in Bioinformatics
- Vol. 6, 1749317
Repurposing AI for protein interactions and dynamics: opportunities, limitations, and lessons
Understanding protein interactions and dynamics of biological systems is central in drug discovery. Advances in artificial intelligence (AI) have expanded the scope of predictive learning for complex biological systems. Repurposing current gold-standard AI algorithms for structural and biological applications illustrates how flexible and powerful these approaches can be. In this mini-review, we examine how AI models are repurposed across domains and analyze how inductive biases, learning objectives, and representation choices inherited from their original applications shape performance in protein interaction and dynamics tasks. We discuss where AI approaches succeed, where they systematically fail, and how their behavior differs from physics-based modeling. We further highlight unresolved biological challenges, data and benchmarking limitations, and emerging opportunities for hybrid AI-physics workflows that balance efficiency with physical realism. By framing recent developments through a cross-domain adaptation framework, this review aims to provide practical guidance for selecting, evaluating, and integrating AI models in protein interaction and dynamics studies, and to support more reliable and biologically meaningful applications of AI in computational protein science.
- 11 March 2026
- Published by Frontiers Media SA in Frontiers in Bioinformatics
- Vol. 6, 1749317
Open AccessRepurposing AI for protein interactions and dynamics: opportunities, limitations, and lessons
Understanding protein interactions and dynamics of biological systems is central in drug discovery. Advances in artificial intelligence (AI) have expanded the scope of predictive learning for complex biological systems. Repurposing current gold-standard AI algorithms for structural and biological applications illustrates how flexible and powerful these approaches can be. In this mini-review, we examine how AI models are repurposed across domains and analyze how inductive biases, learning objectives, and representation choices inherited from their original applications shape performance in protein interaction and dynamics tasks. We discuss where AI approaches succeed, where they systematically fail, and how their behavior differs from physics-based modeling. We further highlight unresolved biological challenges, data and benchmarking limitations, and emerging opportunities for hybrid AI-physics workflows that balance efficiency with physical realism. By framing recent developments through a cross-domain adaptation framework, this review aims to provide practical guidance for selecting, evaluating, and integrating AI models in protein interaction and dynamics studies, and to support more reliable and biologically meaningful applications of AI in computational protein science.
- Top Cited PapersOpen Access
- 4 March 2026
- Published by Taylor & Francis in Expert Opinion on Drug Discovery
- Vol. 21 (3) , 273-287
Deep generative molecular design and its value in modern drug discovery
1. Drug discovery is undergoing a fundamental transformation as AI moves from predictive modeling to generative design [1]. For decades, computational drug discovery has focused on predicting molec...
- 4 March 2026
- Published by Taylor & Francis in Expert Opinion on Drug Discovery
- Vol. 21 (3) , 273-287
Top Cited PapersOpen AccessDeep generative molecular design and its value in modern drug discovery
1. Drug discovery is undergoing a fundamental transformation as AI moves from predictive modeling to generative design [1]. For decades, computational drug discovery has focused on predicting molec...
- Open Access
- 31 January 2026
- Published by Elsevier in Biophysical Journal
- Vol. 125 (18) , 5194-5209
G12 mutations rewire allosteric communication at the Ras–RalGDS interface
- 31 January 2026
- Published by Elsevier in Biophysical Journal
- Vol. 125 (18) , 5194-5209
Open AccessG12 mutations rewire allosteric communication at the Ras–RalGDS interface
- Open Access
- 23 December 2025
- Published by Wiley in Protein Science
- Vol. 35 (1) , e70420
The structural heterogeneity of AKT autoinhibition
AKT is key to controlling cell growth through the PI3K/AKT/mTOR pathway. In the cytosol, in the absence of stimulus, AKT is autoinhibited to prevent uncontrolled activation. Increased AKT activity contributes to tumor growth by phosphorylating numerous downstream targets. Relieving the autoinhibition is a prerequisite for full activation, which occurs through C‐terminal tail phosphorylation by mTOR, followed by activation loop phosphorylation by PDK1. However, the atomic‐level mechanisms by which AKT autoinhibition persists in the cytosol and the phosphorylation (posttranslational modifications) allosterically shift AKT to its open conformation, which may serve as drug targets, remain unclear. Here, we performed explicit molecular dynamics simulations to explore the conformational ensembles of AKT in these different states. Our unbiased results show how the variable loops of the PH domain contribute to the PH‐mediated AKT autoinhibition. Autoinhibited states are commonly only marginally stable, populating function‐related shallow metastable wells with relatively similar energies and low kinetic barriers, making them receptive to regulation. The conformational heterogeneity of AKT's autoinhibitory interface is susceptible to regulation, including by phosphorylation, but also by activating mutations and allosteric inhibitors. As to activation by phosphorylation, allosteric communication between the phosphorylated C‐terminal tail and the PH domain of AKT promotes the release of the PH domain from the kinase domain, independent of PIP 3 . Our results clarify how mutations and phosphorylation can impact autoinhibition and resistance to allosteric inhibitors, highlighting how metastable states can contribute to cellular regulation. Heterogeneous population with low stability and low kinetic barriers can be a useful attribute of living cells.
- 23 December 2025
- Published by Wiley in Protein Science
- Vol. 35 (1) , e70420
Open AccessThe structural heterogeneity of AKT autoinhibition
AKT is key to controlling cell growth through the PI3K/AKT/mTOR pathway. In the cytosol, in the absence of stimulus, AKT is autoinhibited to prevent uncontrolled activation. Increased AKT activity contributes to tumor growth by phosphorylating numerous downstream targets. Relieving the autoinhibition is a prerequisite for full activation, which occurs through C‐terminal tail phosphorylation by mTOR, followed by activation loop phosphorylation by PDK1. However, the atomic‐level mechanisms by which AKT autoinhibition persists in the cytosol and the phosphorylation (posttranslational modifications) allosterically shift AKT to its open conformation, which may serve as drug targets, remain unclear. Here, we performed explicit molecular dynamics simulations to explore the conformational ensembles of AKT in these different states. Our unbiased results show how the variable loops of the PH domain contribute to the PH‐mediated AKT autoinhibition. Autoinhibited states are commonly only marginally stable, populating function‐related shallow metastable wells with relatively similar energies and low kinetic barriers, making them receptive to regulation. The conformational heterogeneity of AKT's autoinhibitory interface is susceptible to regulation, including by phosphorylation, but also by activating mutations and allosteric inhibitors. As to activation by phosphorylation, allosteric communication between the phosphorylated C‐terminal tail and the PH domain of AKT promotes the release of the PH domain from the kinase domain, independent of PIP 3 . Our results clarify how mutations and phosphorylation can impact autoinhibition and resistance to allosteric inhibitors, highlighting how metastable states can contribute to cellular regulation. Heterogeneous population with low stability and low kinetic barriers can be a useful attribute of living cells.
- Open Access
- 21 November 2025
- Published by Elsevier in Biophysical Journal
- Vol. 125 (18) , 4866-4884
Reimagining computational macromolecular modeling: AI-driven approaches
- 21 November 2025
- Published by Elsevier in Biophysical Journal
- Vol. 125 (18) , 4866-4884
Open AccessReimagining computational macromolecular modeling: AI-driven approaches
- 25 September 2025
- Published by Elsevier in Current Opinion in Structural Biology
- Vol. 95, 103159
Rewiring enzyme regulation: Allosteric drugs and predictive tools
- 25 September 2025
- Published by Elsevier in Current Opinion in Structural Biology
- Vol. 95, 103159
Rewiring enzyme regulation: Allosteric drugs and predictive tools
- Open Access
- 26 June 2025
- Published by Wiley in Ultrasound in Obstetrics & Gynecology
- Vol. 66 (2) , 217-225
Alterations in vaginal microbiome in women with short cervix: longitudinal study of microbial diversity and impact of vaginal progesterone treatment
Objectives: A short cervix is a known risk factor for preterm birth, and imbalances in the vaginal microbiome, such as low relative abundance of Lactobacillus , may be associated with an increased risk of preterm birth. The aim of this study was to evaluate differences in the vaginal microbiome between women with a short cervix and those with normal cervical length in the second trimester. Additionally, we aimed to assess longitudinal changes in microbial diversity during pregnancy, as well as the impact of vaginal progesterone treatment on vaginal microenvironment in women with a short cervix. Methods: This was a prospective, longitudinal study conducted at Koc University Hospital between January 2020 and May 2023, in women with a singleton pregnancy with a short cervical length (≤ 25 mm) in the second trimester (20 + 0 to 24 + 6 weeks' gestation). After diagnosis of short cervix, administration of 200 mg vaginal progesterone daily was initiated. The control group comprised women with a normal cervical length (> 25 mm) in the second trimester, matched for age and body mass index (BMI). Cervicovaginal swabs were collected from the posterior fornix at three gestational‐age ranges: in the first trimester (11 + 0 to 13 + 6 weeks), the second trimester (20 + 0 to 24 + 6 weeks) and the third trimester (28 + 0 to 34 + 6 weeks), and cervical length was measured following sample collection. DNA was extracted and the 16S rRNA bacterial gene was sequenced to analyze and compare the vaginal microbiome between women with a short cervix and controls. We also assessed the microbiome longitudinally in each group, across the first, second and third trimesters. In the short‐cervix group, we also compared the microbiome before initiation of progesterone treatment in the second trimester and 4 weeks after its initiation. Results: Among 490 pregnant women who underwent first‐trimester screening during the study period and had vaginal swabs collected, short cervical length was detected in 31 at the second‐trimester scan. These women formed the study group. A further 27 women, with a normal cervical length, were matched for BMI and age and assigned to the control group. During the second trimester, women with a short cervix exhibited greater species diversity compared with the control group; this was suggested by the higher Shannon index (0.45 vs 0.33; P = 0.135), which reflects species richness and evenness, and further demonstrated by the higher Chao index (20.2 vs 13.8; P = 0.018), which estimates species richness. In the second trimester, Lactobacillus was less abundant in women with a short cervix than in the control group, although the difference did not reach significance (86.8% vs 95.5%; P = 0.091). At the phylum level, in women with a short cervix compared to those with normal cervical length, the relative abundance of Firmicutes, to which the genus Lactobacillus belongs, was significantly lower (90.7% vs 97.6%; P = 0.041), while the relative abundances of both Bacteroidota (1.73% vs 0.4%; P = 0.004) and Proteobacteria (0.2% vs 0.01%; P = 0.007) were higher. In the second trimester, the relative abundance of Lactobacillus gasseri was significantly lower in women with a short cervix compared to controls (4.7% vs 13.8%; P = 0.023). In the longitudinal analysis of the vaginal microbiome, there were no significant differences among the trimesters in the control group. In contrast, in those with a short cervix, there was a notable decrease in the amount of Lactobacillus crispatus , from 55.0% in the first trimester to 36.1% in the second trimester ( P = 0.052). In women with a short cervix, there was no significant difference in bacterial diversity after vs before progesterone treatment (Chao index, 22.6 vs 20.5; P = 0.609). Conclusion: These findings highlight the significant alterations in the vaginal microbiome of pregnant women with a short cervix in comparison to those with normal cervical length, particularly in terms of higher species diversity and distinct community composition. The study also shows that vaginal progesterone treatment in women with a short cervix does not alter the vaginal microbiome, suggesting that it is a safe and effective intervention without disrupting the vaginal microbial balance. Understanding the relationship between cervical length and the vaginal microbiome is essential for developing strategies to reduce the risk of preterm birth in high‐risk populations. © 2025 The Author(s). Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.
- 26 June 2025
- Published by Wiley in Ultrasound in Obstetrics & Gynecology
- Vol. 66 (2) , 217-225
Open AccessAlterations in vaginal microbiome in women with short cervix: longitudinal study of microbial diversity and impact of vaginal progesterone treatment
Objectives: A short cervix is a known risk factor for preterm birth, and imbalances in the vaginal microbiome, such as low relative abundance of Lactobacillus , may be associated with an increased risk of preterm birth. The aim of this study was to evaluate differences in the vaginal microbiome between women with a short cervix and those with normal cervical length in the second trimester. Additionally, we aimed to assess longitudinal changes in microbial diversity during pregnancy, as well as the impact of vaginal progesterone treatment on vaginal microenvironment in women with a short cervix. Methods: This was a prospective, longitudinal study conducted at Koc University Hospital between January 2020 and May 2023, in women with a singleton pregnancy with a short cervical length (≤ 25 mm) in the second trimester (20 + 0 to 24 + 6 weeks' gestation). After diagnosis of short cervix, administration of 200 mg vaginal progesterone daily was initiated. The control group comprised women with a normal cervical length (> 25 mm) in the second trimester, matched for age and body mass index (BMI). Cervicovaginal swabs were collected from the posterior fornix at three gestational‐age ranges: in the first trimester (11 + 0 to 13 + 6 weeks), the second trimester (20 + 0 to 24 + 6 weeks) and the third trimester (28 + 0 to 34 + 6 weeks), and cervical length was measured following sample collection. DNA was extracted and the 16S rRNA bacterial gene was sequenced to analyze and compare the vaginal microbiome between women with a short cervix and controls. We also assessed the microbiome longitudinally in each group, across the first, second and third trimesters. In the short‐cervix group, we also compared the microbiome before initiation of progesterone treatment in the second trimester and 4 weeks after its initiation. Results: Among 490 pregnant women who underwent first‐trimester screening during the study period and had vaginal swabs collected, short cervical length was detected in 31 at the second‐trimester scan. These women formed the study group. A further 27 women, with a normal cervical length, were matched for BMI and age and assigned to the control group. During the second trimester, women with a short cervix exhibited greater species diversity compared with the control group; this was suggested by the higher Shannon index (0.45 vs 0.33; P = 0.135), which reflects species richness and evenness, and further demonstrated by the higher Chao index (20.2 vs 13.8; P = 0.018), which estimates species richness. In the second trimester, Lactobacillus was less abundant in women with a short cervix than in the control group, although the difference did not reach significance (86.8% vs 95.5%; P = 0.091). At the phylum level, in women with a short cervix compared to those with normal cervical length, the relative abundance of Firmicutes, to which the genus Lactobacillus belongs, was significantly lower (90.7% vs 97.6%; P = 0.041), while the relative abundances of both Bacteroidota (1.73% vs 0.4%; P = 0.004) and Proteobacteria (0.2% vs 0.01%; P = 0.007) were higher. In the second trimester, the relative abundance of Lactobacillus gasseri was significantly lower in women with a short cervix compared to controls (4.7% vs 13.8%; P = 0.023). In the longitudinal analysis of the vaginal microbiome, there were no significant differences among the trimesters in the control group. In contrast, in those with a short cervix, there was a notable decrease in the amount of Lactobacillus crispatus , from 55.0% in the first trimester to 36.1% in the second trimester ( P = 0.052). In women with a short cervix, there was no significant difference in bacterial diversity after vs before progesterone treatment (Chao index, 22.6 vs 20.5; P = 0.609). Conclusion: These findings highlight the significant alterations in the vaginal microbiome of pregnant women with a short cervix in comparison to those with normal cervical length, particularly in terms of higher species diversity and distinct community composition. The study also shows that vaginal progesterone treatment in women with a short cervix does not alter the vaginal microbiome, suggesting that it is a safe and effective intervention without disrupting the vaginal microbial balance. Understanding the relationship between cervical length and the vaginal microbiome is essential for developing strategies to reduce the risk of preterm birth in high‐risk populations. © 2025 The Author(s). Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.