Ekrem Yasar

Ekrem Yasar

Auto-generated profile, based on publicly available information

Ekrem Yasar

Department of Biophysics, Faculty of Medicine, Erzincan Binali Yildirim University, Erzincan, Turkey

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 August 2025 in Communications Chemistry
Verified
Journal Article
Open access
ASAmanda E. Sandelin
Amanda E. Sandelin
RNRicky Nencini
Ricky Nencini
EYEkrem Yasar
Ekrem Yasar
SFSatoshi Fudo
Satoshi Fudo
VSVassilis Stratoulias
Vassilis Stratoulias
SOSamuli Ollila
Samuli Ollila
Multidomain proteins containing both folded and intrinsically disordered regions are crucial for biological processes, but characterizing their conformational ensembles and dynamics remains challenging. We introduce the Quality Evaluation Based Simulation Selection (QEBSS) protocol, which combines MD simulations with NMR-derived protein backbone 15N T 1 and T 2 spin relaxation times and hetNOE values to interpret conformational ensembles and dynamics of multidomain proteins. We demonstrate the practical advantage of QEBSS by characterizing four flexible multidomain proteins: calmodulin, EN2, MANF, and CDNF. These biologically important proteins have been difficult to study due to their flexible nature. Our findings reveal new insights into their conformational landscapes and dynamics, providing mechanistic understanding of their biological functions. QEBSS offers quantitative quality evaluation of simulations and a systematic approach for resolving conformational ensembles of multidomain proteins with heterogeneous dynamics. Given the importance of such proteins in biology, biotechnology, and materials science, QEBSS should benefit fields from drug design to novel materials development.

15 March 2026 in Journal of Molecular Recognition
Verified
Journal Article
Closed access
EYEkrem Yasar
Ekrem Yasar
SDSegun Dogru
Segun Dogru
EEErol Eroglu
Erol Eroglu
NYNazmi Yaraş
Nazmi Yaraş

15 January 2023 in Turkish Computational and Theoretical Chemistry
Verified
Journal Article
Open access
PDF file
MYMehmet Murat Yaşar
Mehmet Murat Yaşar
EYEkrem Yaşar
Ekrem Yaşar
NYNuri Yorulmaz
Nuri Yorulmaz
MTMehmet Emin Tenekeci
Mehmet Emin Tenekeci
İSİsmail Hakkı SARPÜN
İsmail Hakkı SARPÜN
EEErol Eroğlu
Erol Eroğlu
Possible allosteric inhibitors of MPro were investigated using in silico methods. To this end, FDA-approved drugs in the DrugBank database were subjected to virtual screening, and drugs that strongly bind distant from the catalytic site of MPro were identified using molecular docking. Among the identified drugs, Dihydroergotamine (DHE) was chosen for further investigation due to its highest binding score against MPro in the molecular docking experiment. The allosteric inhibition potential of DHE toward MPro was demonstrated by applying some computational tools on the trajectory files which were obtained from the Molecular Dynamics Simulations. Results support that the hydrogen bonding interactions of DHE with GLU278 and THR280, located between Protomer A and Protomer B, affect the structure of the side chain of CYS145 at the catalytic site of MPro. Considering the role of CYS145 in the catalytic cycle, this structural change is likely to be a mechanism for inhibiting MPro.

27 May 2026 in bioRxiv
Verified
Preprint
Open access
PDF file
EYEkrem Yasar
Ekrem Yasar
Sonification, the systematic mapping of data to non-speech sound, has been applied with quantitative success in astronomy, seismology, and most recently materials chemistry, but has seen limited use in the analysis of biomolecular dynamics. Earlier protein-music studies have focused largely on the static amino-acid sequence, and G-protein-coupled receptor (GPCR) molecular dynamics (MD) trajectories have not previously been the subject of an auditory display framework. Here we present an end-to-end open-source sonification framework for GPCR molecular dynamics together with a quantitative cross-modal validation procedure, and we apply the framework as a proof of concept to three reference beta2-adrenergic receptor (beta2AR) trajectories from the GPCRMD repository spanning the activation continuum (inactive, active apo, active + orthosteric agonist). The framework extracts activation-related geometric features per MD frame, maps them under a single rule onto pitch, note duration, velocity, harmonic intensity, and percussive accents, and renders the result with three timbres (piano, violin, flute). The mapping was tested on two designed pairwise contrasts (activation pair; ligand pair) using Mann-Whitney U tests, Random Forest cross-modal classification with leave-one-instrument-out generalisation, and canonical correlation analysis between the MD and audio feature spaces. All four informative MD features differed between paired states at q < 1 x 10^-20. A Random Forest classifier trained on audio features alone recovered the MD state with balanced accuracy 0.995 +/- 0.003 (activation pair) and 1.000 +/- 0.000 (ligand pair), corresponding to information-retention ratios of 1.006 and 1.000 relative to the MD-feature baseline. First canonical correlations between MD and audio spaces reached r1 = 0.926 (activation) and r1 = 0.995 (ligand). The sonification framework therefore provides a quantitatively faithful auditory representation of GPCR activation dynamics, with potential applications in exploratory MD analysis, accessibility, and education. The framework is system-agnostic and transfers to other GPCRs and allosteric MD systems without code changes beyond residue selection.

12 June 2026 in ChemRxiv
Verified
Preprint
Open access
PDF file
EYEkrem Yasar
Ekrem Yasar
SDSegun Dogru
Segun Dogru
The identification of beta-secretase 1 (BACE-1) inhibitors capable of crossing the blood–brain barrier (BBB) remains a central challenge in Alzheimer's disease drug discovery. While pretrained molecular language models such as ChemBERTa have attracted considerable attention as general-purpose feature extractors, their advantage over classical descriptors for specific therapeutic targets has not been rigorously established. Here, we report a systematic comparison of four molecular representations for binary classification of 2,299 BBB-filtered BACE-1 compounds assembled from MoleculeNet and ChEMBL. The representations evaluated were ECFP4 Morgan fingerprints (2,048 bits), MACCS structural keys (167 bits), RDKit physicochemical descriptors (168 features), and pretrained ChemBERTa embeddings (768 dimensions). Using scaffold-based splitting, data-driven model selection across 27 classifiers, Bayesian optimization via Optuna, and validation through repeated cross-validation and Y-randomization, we found that classical ECFP4 fingerprints paired with the ExtraTreesClassifier achieved the highest test performance (ROC AUC = 0.894, MCC = 0.649), followed by RDKit descriptors (0.866), MACCS keys (0.863), and ChemBERTa (0.861). All pairwise differences were statistically significant (Wilcoxon p < 0.005), yet the total performance spread was only 0.033 ROC AUC units, suggesting that representation choice may be less decisive than often assumed. SHAP-based interpretability analysis revealed markedly different encoding strategies: MACCS keys concentrated predictive importance in a small number of structural fragments, whereas ECFP4 and ChemBERTa distributed it across hundreds of features. Applicability domain analysis confirmed that prediction accuracy scaled with chemical similarity to the training set. These results indicate that pretrained embeddings do not automatically outperform classical fingerprints for target-specific QSAR, and that careful pipeline design may warrant more attention than the choice of representation alone.

27 February 2026 in bioRxiv
Verified
Preprint
Open access
PDF file
EYEkrem Yasar
Ekrem Yasar
ADAbdussamed Yasin Demir
Abdussamed Yasin Demir
SDSegun Dogru
Segun Dogru
Mitochondrial Complex III dysfunction is frequently associated with pathogenic variants in the MT-CYB gene, yet the functional consequences of many missense substitutions remain unresolved because they are classified as variants of uncertain significance (VUS). One such variant, m.14849T>C (p.Ser35Pro), has been reported in patients with multisystem mitochondrial phenotypes, including septo-optic dysplasia, cardiomyopathy, and exercise intolerance, although its structural impact on Cytochrome b function remains unclear. In this study, we employed 300 ns all-atom molecular dynamics simulations to assess structural and energetic consequences of the S35P substitution in the Cytochrome b subunit of human mitochondrial Complex III. The S35P variant did not induce global destabilization of the protein scaffold but instead promoted localized perturbations within the heme bL microenvironment. The mutation was associated with loss of a heme-proximal hydrogen-bonding network involving Ser35 and a decrease in electrostatic interaction energy between the protein matrix and the heme bL cofactor. Radial distribution function analysis further supported loosening of local packing around the prosthetic group. Consistent with these local changes, dynamics analyses indicated increased flexibility in distal transmembrane helices that form the heme-pocket scaffold and greater variability in the inter-heme Fe(bL)–Fe(bH) distance. Together, our findings suggest that S35P may exert functional effects by reorganizing the heme bL microenvironment rather than by inducing large-scale structural destabilization, underscoring the value of structure- and dynamics-based evaluation for mitochondrial VUS and suggesting a plausible mechanistic link to the pathophysiology of multisystem mitochondrial diseases.