Biology subjects

Yilmaz, M.

Publications and source records attributed to Yilmaz, M..

At least 19 recordsLinked to original sources

bioRxiv2026

Extracellular matrix context shapes morphogenesis and lactation-associated states in human milk-derived mammary organoids

The mammary gland relies on reciprocal interactions between epithelial cells and their surrounding extracellular matrix (ECM) to form and maintain milk-producing tissue structures. Yet these processes remain difficult to study in human model systems. Mammary epithelial cells (MECs) can be isolated noninvasively from breast milk, but whether they generate three-dimensional organoids and respond to matrix cues has been unclear. Here, human milk-derived MECs (milk MECs) spontaneously form complex organoids, including polarized acinar and terminal duct lobular unit-like structures after isolation. To investigate how matrix composition shapes these organoids, milk MECs were cultured in decellularized mammary ECM (dECM), Matrigel, and collagen I. In dECM, milk MECs formed polarized branched networks with aligned actin organization along collagen fibrils, whereas in Matrigel they adopted a more lactation-associated state, marked by {beta}-casein expression and milk fat globules. Together, these findings establish breast milk-derived MEC organoids as a human model to study how ECM context regulates mammary morphogenesis and lactation biology.

cell biology↗

bioRxiv2026

MICAL-Inspired Chiral Vanadate Nanoclusters Decelerate Actin Dynamics in Live Cells

Actin turnover is a fundamental cellular process essential for cell dynamics, whose control is critical for both medicine and biotechnology. However, conventional small molecules modifying actin turnover scramble the structure of actin filaments and display high cellular toxicity. MICAL enzymes oxidizing methionine (Met) residues in actin can potentially address this challenge, but their large size and multiple required cofactors make MICALs manufacturing and utilization difficult. Here we show that redox-active chiral decavanadate nanoclusters with tartaric acid are capable of site-selective actin modulation, mimicking MICALs, while requiring no cofactors, displaying high biocompatibility and being membrane permeable. Decavanadate nanoclusters serve as atomically precise "nano-enzymes" oxidizing three Met residues in globular actin, including Met-176; the latter inhibits the opening of the backdoor segment and prevents depolymerization of actin filaments. The structure of actin filaments formed after nanocluster treatment revealed no structural disturbances as confirmed by cryo-electron microscopy. The biocompatibility and bioactivity of chiral decavanadate nanoclusters was demonstrated by modulation of actin in living NG108-15 cells. Taking advantage of atomically precise structure of the nanoclusters, we show that their docking into actin can be predicted computationally, indicating the possibility of programmable actin modulation using the tools of nanochemistry.

biochemistry↗

bioRxiv2026

Agent-Guided De Novo Design of Nanobody Binders Against a Novel Cancer Target

Therapeutic antibody discovery remains slow and resource-intensive, with traditional methods providing limited control over epitope selection. We present a workflow for de novo nanobody design applied to a novel Desmoplastic Small Round Cell Tumor target encompassing four stages: (1) epitope identification guided by our hotspot recommendation agent using physical chemistry-based structure and sequence analysis tools with two curated databases (IEDB, PFAM), (2) de novo nanobody generation using three independent methods (RFantibody, IgGM, mBER) across multiple predicted antigen structures and nanobody frameworks, (3) multi-metric scoring including structural metrics from folding models, and in silico binding affinity from our sequence-based predictor, (4) high-throughput yeast surface display (YSD) screening followed by surface plasmon resonance (SPR) characterization of the specific binders. We generated 288,000 nanobody designs spanning eight target epitope regions and three variable domains of heavy chain-only antibody (VHH) frameworks. Multi-objective Pareto filtering with our candidate selection agent yielded 100,000 candidates for YSD screening with fluorescence-activated cell sorting (FACS). Of 116 enriched candidates advanced to SPR characterization, 46/116 (39.7%) produced reliable kinetic fits with Rmax [≥] 30 RU, yielding KD values from 0.66 nM to 305 nM (median 31.7 nM). These results show that an agent-guided computational workflow can design nanomolar to sub-nanomolar nanobody binders against a novel target without experimental structure or prior antibody information.

bioinformatics↗

bioRxiv2026

Classical enhancers couple cis-regulatory logic with transcriptional condensates and 3D genome architecture

Deciphering the regulatory logic of enhancers remains a central question in understanding cell- and tissue-specific gene expression in multicellular organisms. This is particularly pertinent at multipartite enhancer clusters such as super-enhancers, where multiple enhancers contribute to the expression of a single gene. Gene expression has been studied largely through sequence-dependent recruitment of transcription factors (TF) and co-activators, whereas 3D chromatin structure has been attributed to architectural proteins such as cohesion and CTCF1-3. However, the contribution of DNA sequence encoded in enhancers to shaping higher-order genome organization remains poorly understood. Here we show that classical enhancers, embedded within multipartite super-enhancer structures, act as determinant regulatory elements that initiate the gene regulatory cascade by linking DNA sequence recognition to 3D chromatin architecture. Classical enhancers are more evolutionarily conserved and display stronger regulatory activity than facilitator elements, which lack intrinsic enhancer activity but potentiate classical enhancer function. We show that classical enhancers are selectively bound by specific TFs with strong intrinsically disordered regions, such as NFE2L2 in liver cancer cells, capable of driving transcriptional condensate formation through phase separation. NFE2L2 depletion reduced enhancer activity and induced widespread chromatin reorganization, characterized by increased cohesin and CTCF occupancy at super-enhancer boundaries and beyond. This "cohesin clogging" impaired DNA loop extrusion, led to formation of smaller topologically associated domains, and weakened enhancer-promoter contacts. These findings highlight that sequence-specific TFs have multifaceted roles beyond transcriptional control, establishing a direct mechanistic link between enhancer sequence, TF binding, condensate formation, and 3D genome organization, with the regulatory logic being encoded in the DNA sequence itself.

genomics↗

bioRxiv2026

Context-aware Multi-Property Antibody Predictor: a Novel Framework Integrating Text and Protein Language Models

Recent advances in Machine Learning have transformed antibody development through in-silico models, accelerating therapeutic candidate identification. However, challenges persist: rapid adaptation of property predictors to laboratory-specific assays with incomplete datasets; batch effects introducing systematic bias; assay costs necessitating efficient unseen property prediction. We introduce a novel multi-modal architecture featuring specialized tokenization and embedding projection that integrates text and protein language models (pLM) and a learning strategy to enable in-context learning for multi-property prediction without learning shortcuts. Our framework enables prompting without dictionary merging across modalities, creating a compact model capable of in-context learning for multi-property prediction. The orchestrating model avoids pLM-to-text projection while enabling inference-time adaptation without retraining. Using 876,898 antibodies with batch effect simulation, our architecture achieved Spearmans {rho}>0.8 across multiple developability properties, significantly outperforming fine-tuned multimodal-LLMs and showed the ability of leveraging correlation between properties for prediction. This approach has the potential to address critical antibody development challenges.

bioinformatics↗

bioRxiv2025

Crystal Structure of 3-Hydroxypropionyl-CoA Synthetase (ADP-forming) from Nitrosopumilus maritimus

The 3-hydroxypropionate/4-hydroxybutyrate (3HP/4HB) cycle in thaumarchaeota contributes significantly to global organic carbon fixation as the most energetically efficient aerobic carbon fixation pathway. The thaumarchaeal 3-Hydroxypropionyl-CoA Synthetase (ADP-forming; Nmar_1309) is crucial to this efficiency, utilizing ATP to ADP catalysis. This first reported structure of Nmar_1309 reveals a homodimer with a unique subdomain organization ([3-4-1-2-5]) and a distinct linker between subdomains 4 and 1. The presence of bound substrates including 3-hydroxypropionate, non-hydrolyzable ATP (ADPNP), and a phosphate suggests an intermediate state mimicking a reaction step immediately preceding the formation of a 3-hydroxypropionyl-phosphohistidine. Conformational differences were observed between the two chains of the homodimer, likely influenced by the binding of a single ADPNP molecule in one chain. Phylogenetic analysis suggests that while 4HB synthetases may have evolved earlier in the evolutionary timeline, 3HP synthetases in Thaumarchaeota may have occurred after the Great Oxygenation Event. These structural data provide further characterization of the 3HP/4HB cycle and, in conjunction with the structure of 4-hydroxybutyryl-CoA synthetase, Nmar_0206, provide baseline structures of the key ADP-forming Acyl-CoA synthetases within this pathway.

biophysics↗

bioRxiv2025

Improvements to Casanovo, a deep learning de novo peptidesequencer

Casanovo is a state-of-the-art deep learning model for de novo peptide sequencing from mass spectrometry proteomics data. Here we report on a series of enhancements to Casanovo, aimed at improving the interpretability of the scores assigned to predicted peptides, generalizing the software for use in database search, speeding up training and prediction runtimes, and providing workflows and visualization tools to facilitate adoption of Casanovo and interpretation of its results. Our goal is to make Casanovo accurate and easy to use for applications such as metaproteomics, antibody sequencing, immunopeptidomics, and discovery of novel peptide sequences in standard proteomics analyses. Casanovo is available as open source at https://github.com/Noble-Lab/casanovo.

molecular biology↗

bioRxiv2024

Systematic benchmarking of mass spectrometry-based antibody sequencing reveals methodological biases

The circulating antibody repertoire is crucial for immune protection, holding significant immunological and biotechnological value. While bottom-up mass spectrometry (MS) is the most widely used proteomics technique for profiling the sequence diversity of circulating antibodies (Ab-seq), it has not been thoroughly benchmarked. We quantified the replicability and robustness of Ab-seq using six monoclonal antibodies with known protein sequences in 70 different combinations of concentration and oligoclonality, both with and without polyclonal serum IgG background. Each combination underwent four protease treatments and was analyzed across four experimental and three technical replicates, totaling 3,360 LC-MS/MS runs. We quantified the dependence of MS-based Ab-seq identification on antibody sequence, concentration, protease, background signal diversity, and bioinformatics setups. Integrating the data from experimental replicates, proteases, and bioinformatics tools enhanced antibody identification. De novo peptide sequencing showed similar performance to database-dependent methods for higher antibody concentrations, but de novo antibody reconstruction remains challenging. Our work provides a foundational resource for the field of MS-based antibody profiling.

immunology↗

bioRxiv2024

LYCEUM: Learning to call copy number variants on low coverage ancient genomes

Copy number variants (CNVs) are pivotal in driving phenotypic variation that facilitates species adaptation. They are significant contributors to various disorders, making ancient genomes crucial for uncovering the genetic origins of disease susceptibility across populations. However, detecting CNVs in ancient DNA (aDNA) samples poses substantial challenges due to several factors: (i) aDNA is often highly degraded; (ii) contamination from microbial DNA and DNA from closely related species introduce additional noise into sequencing data; and finally, (iii) the typically low coverage of aDNA renders accurate CNV detection particularly difficult. Conventional CNV calling algorithms, which are optimized for high coverage read-depth signals, underperform under such conditions. To address these limitations, we introduce LYCEUM, the first machine learning-based CNV caller for aDNA. To overcome challenges related to data quality and scarcity, we employ a two-step training strategy. First, the model is pre-trained on whole genome sequencing data from the 1000 Genomes Project, teaching it CNV-calling capabilities similar to conventional methods. Next, the model is fine-tuned using high-confidence CNV calls derived from only a few existing high-coverage aDNA samples. During this stage, the model adapts to making CNV calls based on the downsampled read depth signals of the same aDNA samples. LYCEUM achieves accurate detection of CNVs even in typically low-coverage ancient genomes. We also observe that the segmental deletion calls made by LYCEUM show correlation with the demographic history of the samples and exhibit patterns of negative selection inline with natural selection. LYCEUM is available at https://github.com/ciceklab/LYCEUM.

bioinformatics↗

bioRxiv2024

Exercise-induced extracellular vesicles mediate apoptosis in human colon cancer cells in an exercise intensity-dependent manner

Regular exercise is known to reduce incidence rates and improve the prognosis of all cancers, but the underlying mechanisms remain elusive. Ample evidence suggests that exercise exerts therapeutic effects through extracellular vesicles (EVs), essential for cellular communication. Here, we hypothesized that exercise-induced EVs from serum of healthy young male participants will exert anti-tumorigenic effects on human colon cancer HT-29 cells, in an exercise intensity-dependent manner. 10 healthy young active males (25.4{+/-}6.2yrs, with maximal oxygen consumption (VO2max) = 45{+/-}3.7 ml.min-1.kg-1 participated in a randomized crossover trial. Participants underwent two different workload-matched, acute bouts of exercise: (1) moderate-intensity continuous exercise (MICE) at 50-55% V02max, and (2) high-intensity interval exercise (HIIE) at 90% V02max on a cycle ergometer. A control session of rest (PRE) was included. EVs were isolated from serum samples collected during PRE and immediately after each exercise session. EVs were co-incubated with colon cancer HT-29 cells (100 {micro}g EVs/ml, 48-72h), and effect on cell viability, migration, and apoptosis measured. EV treatment reduced cell viability in all groups (PRE, MICE, HIIE) by 35%, 43% and 47% respectively, vs. PBS. EVs from HIIE group showed a significantly greater reduction in cell viability vs. PRE, therefore further analysis used these groups only. PRE-EVs reduced migration by 27%, and HIIE-EVs by 39%. EV from HIIE group increased expression of pro-apoptotic markers: Bax/Bcl-2 ratio by 56% and Caspase-3 by 30% vs. PBS, with no change observed in the PRE group. Further, 16% of cells in PRE, and 28% of cells in HIIE were TUNEL-positive, indicating DNA fragmentation, a hallmark feature of apoptosis. Our data show that exercise-induced EVs reduced cell viability, in an exercise intensity-dependent manner. HIIE-derived EVs exerted the most anti-tumorigenic effects: decreased cell viability, reduced cell migration, increase in pro-apoptotic protein expression, and elevated DNA fragmentation. It is likely these changes were mediated by altered EV Cargo induced by exercise, as the amount of EVs was the same in each treatment group. To our knowledge, this is the first human study that illustrates the therapeutic potential of exercise-induced EVs in cancer treatment.

cancer biology↗

bioRxiv2024

Accounting for digestion enzyme bias in Casanovo

A key parameter of any proteomics mass spectrometry experiment is the identity of the enzyme that is used to digest proteins in the sample into peptides. The Casanovo de novo sequencing model was trained using data that was generated with trypsin digestion; consequently, the model prefers to predict peptides that end with the amino acids "K" or "R." This bias is desirable when the Casanovo is used to analyze data that was also generated using trypsin but can be problematic if the data was generated using some other digestion enzyme. In this work, we modify Casanovo to take as input the identify of the digestion enzyme, alongside each observed spectrum. We then train Casanovo with data generated using several different restriction enzymes, and we demonstrate that the resulting model successfully learns to capture enzyme-specific behavior. However, we find, surprisingly, that this new model does not yield a significant improvement in sequencing accuracy relative to a model trained without the enzyme information but using the same training set. This observation may have important implications for future attempts to make use of experimental metadata in de novo sequencing models.

bioinformatics↗

bioRxiv2024

A learned score function improves the power of mass spectrometry database search

One of the core problems in the analysis of protein tandem mass spectrometry data is the peptide assignment problem: determining, for each observed spectrum, the peptide sequence that was responsible for generating the spectrum. Two primary classes of methods are used to solve this problem: database search and de novo peptide sequencing. State-of-the-art methods for de novo sequencing employ machine learning methods, whereas most database search engines use hand-designed score functions to evaluate the quality of a match between an observed spectrum and a candidate peptide from the database. We hypothesize that machine learning models for de novo sequencing implicitly learn a score function that captures the relationship between peptides and spectra, and thus may be re-purposed as a score function for database search. Because this score function is trained from massive amounts of mass spectrometry data, it could potentially outperform existing, hand-designed database search tools. To test this hypothesis, we re-engineered Casanovo, which has been shown to provide state-of-the-art de novo sequencing capabilities, to assign scores to given peptide-spectrum pairs. We then evaluated the statistical power of this Casanovo score function, Casanovo-DB, to detect peptides on a benchmark of three mass spectrometry runs from three different species. Our results show that, at a 1% peptide-level false discovery rate threshold, Casanovo-DB outperforms existing hand-designed score functions by 35% to 88%. In addition, we show that re-scoring with the Percolator post-processor benefits Casanovo-DB more than other score functions, further increasing the number of detected peptides.

bioinformatics↗

bioRxiv2023

4D Crystallography Captures Transient IF1-Ribosome Dynamics in Translation Initiation

Initiation factor 1 (IF1) is one of multiple key ligands involved in the initiation of mRNA translation, a highly dynamic and carefully-orchestrated process. However, details surrounding IF1 transient interactions with the small 30S ribosomal subunit remain incompletely understood despite characterization of unbound and fully-bound 30S states. Improvements in X-ray light sources and crystallographic techniques are now enabling time-resolved structural studies at near-physiological temperature and near-atomic resolution and thus the structural investigation of such dynamic processes. Here, we employed time-resolved serial femtosecond X-ray crystallography (TR-SFX) to probe the binding of IF1 to the small 30S ribosomal subunit in real time. Our time-resolved structural data demonstrates transient cryptic short-, mid-, and long-range allostery among different regions of the small 30 ribosomal subunit during IF1 binding, revealing small- and large-scale protein-target interactions and dynamics within intermediate macromolecular states at unprecedented temporal and spatial resolution. These data represent one of the first such 4D crystallographic studies assessing protein-protein and protein-RNA interactions and could serve as the basis for subsequent studies of the ribosome and of the multitudinous dynamic processes which underpin biology, and therefore, of life.

biophysics↗

bioRxiv2023

Ambient Temperature Bacterial Large Ribosomal Subunit Structure Enabled by Serial Femtosecond X-ray Crystallography

Ribosomes are the supramolecular complexes responsible for protein synthesis. The large 50S ribosomal subunit catalyzes the peptidyl transferase reaction and peptide bond formation between amino acids. The 50S is targeted by many known clinically effective antibiotics. Available structures, obtained at cryogenic temperatures (CT), are used for drug discovery despite that active or important target sites may display a structural configuration that is CT-induced. The introduction of ultrafast and ultrabright X-ray free electron laser (XFEL) pulses has enabled the structural observation of biological macro- and supramolecules at previously unattainable, near-physiological temperatures. In this study, we use ultrafast and ultrabright XFEL pulses to solve the apo form of 50S ribosomal subunit isolated from the extremely thermophilic bacterium Thermus thermophilus at ambient temperature (AT). The dimeric structure of the 50S subunit presented in this work is among the largest ([~]3 megadalton) structures determined using an XFEL source to date. This study demonstrates the ability to obtain new information about ribosome structural dynamics at AT through serial femtosecond X-ray crystallography (SFX). This allowed us to capture previously unobserved dynamics of ribosomal protein uL23 and coordination by hexahydrated magnesium cations at a hitherto unseen resolution at near-physiological temperature. Also, residue A2602, at the core of the peptidyl transferase center (PTC), shows a rather different orientation of the sugar moiety if compared to CT structures. In addition, our structure highlights the importance of flexible residues at both the PTC and in the binding sites for antibiotics erythromycin and chloramphenicol. The method implemented here may also serve as a starting point for future structural research involving the 50S subunit complexes by employing time-resolved mix-inject and probe kineto-crystallography experiments at XFELs. Unveiling ligand-dependent 50S dynamics at physiological temperatures shall guide further development of next-generation antibiotics that target the translation machinery.

biophysics↗

bioRxiv2023

Sequence-to-sequence translation from mass spectra to peptides with a transformer model

A fundamental challenge for any mass spectrometry-based proteomics experiment is the identification of the peptide that generated each acquired tandem mass spectrum. Although approaches that leverage known peptide sequence databases are widely used and effective for well-characterized model organisms, such methods cannot detect unexpected peptides and can be impractical or impossible to apply in some settings. Thus, the ability to assign peptide sequences to the acquired tandem mass spectra without prior information--de novo peptide sequencing--is valuable for gaining biological insights for tasks including antibody sequencing, immunopeptidomics, and metaproteomics. Although many methods have been developed to address this de novo sequencing problem, it remains an outstanding challenge, in part due to the difficulty of modeling the irregular data structure of tandem mass spectra. Here, we describe Casanovo, a machine learning model that uses a transformer neural network architecture to translate the sequence of peaks in a tandem mass spectrum into the sequence of amino acids that comprise the generating peptide. We train a Casanovo model from 30 million labeled spectra and demonstrate that the model outperforms several state-of-the-art methods on a cross-species benchmark dataset. We also develop a version of Casanovo that is fine-tuned for non-enzymatic peptides. Finally, we demonstrate that Casanovos superior performance improves the analysis of immunopeptidomics and metaproteomics experiments and allows us to delve deeper into the dark proteome.

bioinformatics↗

bioRxiv2022

3D Digitizing An Animal Organ Could Make It Possible To Work With Anatomy Classes In The Metaverse Or With A 3D-Printed Real-Scale Organ On Demand

Organ scanning is not a novel technology but the unique part of this study is to eliminate the need for expensive devices by using free iOS/Android applications. The apps allow users to create high-quality 3D models from photos taken with a phone or tablet. Metaverse and 3D digitization are the future of technology. Implementation of these approaches into the anatomy can change education style. Also, 3D printing technology along with the digitization of real-sized organs enables researchers and clinicians to study with 3D printed organs in class, even with augmented reality glasses. In this study, we first aimed to 3D digitized an animal organ with its real sizes to print in a 3D printer. Next, we wanted to digitize decellularized form of the organ, which suggests researchers study extracellular matrix and vascularization. This digitization of the different organ forms can improve the progression of artificial organ studies. Free and easy-to-use iOS or Android applications for 3D digitization enable medical students to study not only animal organs but cadavers. Hereby, we are suggesting organ scanning can be done on a daily basis in a standard laboratory. It is aimed to enable medical students to work effectively in anatomy lessons in Metaverse or in real cases, thanks to the anatomical library created with the accessibility brought by easy usability.

scientific communication and education↗

bioRxiv2022

Rapid and High Resolution Ambient Temperature Structure Determination at Turkish Light Source

High-resolution biomacromolecular structure determination is essential to better understand protein function and dynamics. Serial crystallography is an emerging structural biology technique which has fundamental limitations due to either sample volume requirements or immediate access to the competitive X-ray beamtime. Obtaining a high volume of well-diffracting, sufficient-size crystals while mitigating radiation damage remains a critical bottleneck of serial crystallography. As an alternative, we introduce the plate-reader module adapted for using a 72-well Terasaki plate for biomacromolecule structure determination at a convenience of a home X-ray source. We also present the first ambient temperature lysozyme structure determined at the Turkish Light Source (Turkish DeLight). The complete dataset was collected in 18.5 mins with resolution extending to 2.39 [A] and 100% completeness. Combined with our previous cryogenic structure (PDB ID: 7Y6A), the ambient temperature structure provides invaluable information about the structural dynamics of the lysozyme. Turkish DeLight provides robust and rapid ambient temperature biomacromolecular structure determination with limited radiation damage.

biophysics↗

bioRxiv2022

NUDT6, the Antisense Protein of FGF2 Gene, Plays a Depressogenic Role by Promoting Inflammation and Suppressing Neurogenesis without Altering FGF2 Signaling

Fibroblast growth factor-2 (FGF2) is involved in the regulation of affective behavior and shows antidepressant effects through Akt and ERK1/2 pathways. NUDT6 is a protein encoded from FGF2 genes antisense strand and its role in the regulation of affective behavior is unclear. Here, we show that increasing NUDT6 expression in the hippocampus results in depression-like behavior in rats without changing FGF2 levels or activating its downstream effectors, Akt and ERK1/2. Instead, NUDT6 acts by inducing inflammatory signaling, specifically by increasing S100A9 levels, activating NF-{kappa}B and rising microglia number along with a reduction in neurogenesis. Conversely, inhibition of hippocampal NUDT6 expression by shRNA results in antidepressant effects and increases neurogenesis without altering FGF2 levels. Together these findings suggest that NUDT6 may play a role in major depression by inducing a proinflammatory state and serve as a novel therapeutic target for antidepressant development. This is the first report of an antisense protein acting through a different mechanism of action than regulation of its sense protein. The opposite effects of NUDT6 and FGF2 on depression-like behavior may serve as a mechanism to fine-tune affective behavior. Our findings open up new venues for studying the differential regulation and functional interactions of sense and antisense proteins in neural function and behavior as well as in neuropsychiatric disorders.

neuroscience↗