Biology subjects

Stark, H.

Publications and source records attributed to Stark, H..

At least 19 recordsLinked to original sources

bioRxiv2026

BoltzProt-1: Towards Efficient De Novo Binder Design with Good Developability

Designing binders against novel protein targets remains a central challenge in computational drug discovery. Here we introduce BoltzProt-1, a pipeline for generating protein binders, including nanobodies, with improved hit rates and favorable developability properties. At its core lie a refined iteration of BoltzGens generative model and a novel protein-protein interaction prediction model, BoltzPPI. Employing BoltzPPI instead of BoltzGens standard structure-prediction confidence metrics to rank nanobody (VHH) designs increases the confirmed-binder hit rate from 3.3% to 8.0% across 10 novel targets. Assessed on 10 additional targets used in prior literature, the BoltzProt-1 pipeline obtains nanobody screening hits for 7 of 10 targets, surpassing the 6 of 10 previously reported by Chai-2. Finally, evaluating the developability of BoltzProt-1-designed nanobodies in terms of stability, aggregation, purity, polyspecificity and hydrophobicity reveals that 58% of its confirmed binders pass every criterion, exceeding both BoltzGen (40%) and clinical-stage VHH controls (21%). O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=104 SRC="FIGDIR/small/733997v1_ufig1.gif" ALT="Figure 1"> View larger version (39K): org.highwire.dtl.DTLVardef@d86de1org.highwire.dtl.DTLVardef@115da7eorg.highwire.dtl.DTLVardef@1bbb4e9org.highwire.dtl.DTLVardef@626a68_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioinformatics↗

bioRxiv2025

BoltzGen: Toward Universal Binder Design

We introduce BoltzGen, an all-atom generative model for designing proteins and peptides across all modalities to bind a wide range of biomolecular targets. BoltzGen builds strong structural reasoning capabilities about target-binder interactions into its generative design process. This is achieved by unifying design and structure prediction, resulting in a single model that also reaches state-of-the-art folding performance. BoltzGens generation process can be controlled with a flexible design specification language over covalent bonds, structure constraints, binding sites, and more. We experimentally validate these capabilities in eight diverse design campaigns with functional and affinity readouts across 26 targets. In our experiments, binder modalities span from nanobodies to disulfide-bonded peptides, and targets from disordered proteins to small molecules. In particular, we identify nanobody binders for novel targets with low similarity to proteins with already known bound structures. We release model weights, data, and both inference and training code at: https://github.com/HannesStark/boltzgen.

bioengineering↗

bioRxiv2025

Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction

Accurately modeling biomolecular interactions is a central challenge in modern biology. While recent advances, such as AlphaFold3 and Boltz-1, have substantially improved our ability to predict biomolecular complex structures, these models still fall short in predicting binding affinity, a critical property underlying molecular function and therapeutic efficacy. Here, we present Boltz-2, a new structural biology foundation model that exhibits strong performance for both structure and affinity prediction. Boltz-2 introduces controllability features including experimental method conditioning, distance constraints, and multi-chain template integration for structure prediction, and is, to our knowledge, the first AI model to approach the performance of free-energy perturbation (FEP) methods in estimating small molecule-protein binding affinity. Crucially, it achieves strong correlation with experimental readouts on many benchmarks, while being at least 1000x more computationally efficient than FEP. By coupling Boltz-2 with a generative model for small molecules, we demonstrate an effective workflow to find diverse, synthesizable, high-affinity binders, as estimated by absolute FEP simulations on the TYK2 target. To foster broad adoption and further innovation at the intersection of machine learning and biology, we are releasing Boltz-2 weights, inference, and training code 1 under a permissive open license, providing a robust and extensible foundation for both academic and industrial research.

molecular biology↗

bioRxiv2025

Proteomic study for the prediction of μCT imaging with iodine

Iodine-based staining techniques are commonly used in histological imaging and micro-computed tomography ({micro}CT) due to iodines affinity for binding to specific molecules. However, the basis for tissue-specific contrast has not yet been sufficiently explored. In this study, we analyse the human proteome at four different levels: individual proteins, protein families, tissues with additional expression values for selected proteins, and organs as a distinct combination of different tissues. At each level, we try to identify proteins/groups with high potential for iodine binding, especially those rich in aromatic heterocyclic amino acids. Using bioinformatic methods, we evaluate the occurrence of aromatic/non-aromatic heterocyclic, carbocyclic, and the remaining 15 amino acids in 20,650 proteins, 1,487 families, 57 tissues, and 16 organs. At the protein level, structural proteins such as titin, nebulin, obscurin, mucin, filaggrin and hornerin have a high absolute number of aromatic heterocyclic amino acids, which could explain the high {micro}CT contrast in muscle, skin and mucosal tissues. At the next level, however, structural families (such as the Laminin-family) rank significantly lower in comparison. These results are reflected in tissues and organs for which protein expression is available. Here, no significant correlations between the enrichment of heterocycles and the intensity of iodine staining can be observed. Furthermore, the enrichment of amino acids in each tissue/organ is relatively similar and shows no significant difference. Our results provide a general basis for iodine-based tissue imaging and serve as a potential starting point for future research, e.g. for cross-species applications and for the structural and functional effects of iodination.

biochemistry↗

bioRxiv2025

The distribution of connective tissue in humans - Reconstruction and classification

This study aimed to measure human connective tissue distribution and characterise material properties. We used, therefore, the existing data sets of a male and female body donor from the Visible Human Project (VHP). In the first step, collagen-containing structures were segmented and reconstructed using digital image processing. In addition, several properties were characterised and analysed using computational methods. The percentage of fibrous components within the whole body was highest in the thigh. The results of the local orientation tensors showed that individual fascicles and subtendons can be resolved when isolated but not within compact regions of connective tissues.

bioengineering↗

bioRxiv2024

Control of limb loading during active horizontal perturbations at moderate and fast trots in rats

To understand how small animals cope with complex, unstructured, and unpredictable substrates, we analyzed the kinetics of rats (n = 10) moving at a fast and at a moderate trot over an unperturbed substrate and a substrate subjected to active horizontal perturbations. Perturbations were active single forwards or backwards displacements of an instrumented platform by amplitudes of 5mm or 10mm in 0.05s. Single leg ground reaction forces (SLGRF) were collected for unperturbed and perturbed locomotion (hindlimbs: 50/102, forelimbs: 45/130, respectively). When negotiating horizontal perturbations, rats displayed gait resetting (braking, accelerating) and non-resetting behaviors. Feedforward strategies differed between the fore- and hindlimbs. In circa 60% of the perturbed trials, forelimbs started the step in acceleration mode, while hindlimbs began the stance mostly in non-resetting mode ([~]45%). In about 50% of all perturbed steps, the impulse provided by the SLGRF displayed a change in behavior according to the expected response to the perturbation. The remaining 50% retained the feedforward strategy. Still, most perturbed trials displayed changes in SLGRF patterns that indicated passive and active reactions to platform shifts. Our results indicate that rats sensorimotor control system tunes fore- and hindlimbs differently in expectation of a perturbation. In addition, the tendon-muscle systems of the limbs are recruited to prevent leg collapse at the beginning and end of the stance. At lower speeds, spinal and/or higher center commands have enough time to re-adapt limb behavior. At higher speeds of locomotion, rats rely more on their limbs intrinsic stability and on feedforward control.

animal behavior and cognition↗

bioRxiv2024

PINDER: The protein interaction dataset and evaluation resource

Protein-protein interactions (PPIs) are fundamental to understanding biological processes and play a key role in therapeutic advancements. As deep-learning docking methods for PPIs gain traction, benchmarking protocols and datasets tailored for effective training and evaluation of their generalization capabilities and performance across real-world scenarios become imperative. Aiming to overcome limitations of existing approaches, we introduce PINDER, a comprehensive annotated dataset that uses structural clustering to derive non-redundant interface-based data splits and includes holo (bound), apo (unbound), and computationally predicted structures. PINDER consists of 2,319,564 dimeric PPI systems (and up to 25 million augmented PPIs) and 1,955 high-quality test PPIs with interface data leakage removed. Additionally, PINDER provides a test subset with 180 dimers for comparison to AlphaFold-Multimer without any interface leakage with respect to its training set. Unsurprisingly, the PINDER benchmark reveals that the performance of existing docking models is highly overestimated when evaluated on leaky test sets. Most importantly, by retraining DiffDock-PP on PINDER interface-clustered splits, we show that interface cluster-based sampling of the training split, along with the diverse and less leaky validation split, leads to strong generalization improvements.

bioinformatics↗

bioRxiv2024

Synchronization and metachronal waves of elastic cilia caused by transient viscous flow

Hydrodynamic coordination of cilia is ubiquitous in biology. It is commonly modeled using the steady Stokes equations. The flow around ciliated cells, however, exhibits finite time vorticity diffusion, requiring a dynamical description. We present a model of elastic cilia coupled by transient viscous flow in the bulk fluid. Therein, vorticity diffusion impacts cilia coordination qualitatively and quantitatively. In particular, pairs of cilia synchronize in antiphase for long diffusion times. Moreover, metachronal waves occur in cilia chains larger than the viscous penetration depth, whereas global synchronization occurs in Stokes flow.

biophysics↗

bioRxiv2024

Anomalous diffusion analysis reveals cooperative locomotion of adult parasitic nematodes in sex-mixed groups.

Parasitic worms are widespread throughout the world and causing agents of chronic infections in humans and animals. The majority of these pathogens parasitize the gut of the infected hosts, however very little is known about the locomotion of the worms dwelling the gut. We studied the movement of Heligmosomoides bakeri (previously Heligmosomoides polygyrus), a natural infection of mice and usually used as animal model to study the roundworm infections. We investigated the locomotion of H. bakeri in simplified environments mimicking key physical features of the intestinal lumen, i.e. various medium viscosities, and a periodical intestinal villi topography. The non-periodical nematode motion in these settings could be described by anomalous diffusion theory. Fascinatingly, an oriented, super-diffusive locomotion of nematodes in sex-mixed groups were detected, in the sense of a cooperative, but not collective (swarm-like) migration stimulated by mating and reproduction, while individual nematodes moved randomly, following a diffusive motion regime. High mucus-like medium viscosity and villi topography, representing physical constraints of nematode locomotion, slowed down but did not prevent this cooperative migration. Additionally, the mean displacement rate of nematodes in sex-mixed groups of 6{middle dot}10-4 mm/s in viscous mucus-like medium are in good agreement with estimates of nematode migration velocities between 10-4 to 10-3 mm/s in the gut. Thus, our data indicate the intestinal nematodes motion to be non-periodic and random but triggered to be oriented by kin of the different sex.

biophysics↗

bioRxiv2023

Discovery of a new activator of Slack potassium channels with robust efficacy in models of histamine-independent and chronic itch

Various disorders are accompanied by histamine-independent itching, which is often resistant to the currently available therapies. In this study, we hypothesized that pharmacological activation of Slack (Kcnt1, KNa1.1), a potassium channel highly expressed in itch-sensitive sensory neurons, has therapeutic potential for the treatment of itching. Based on the Slack-activating antipsychotic drug, loxapine, we designed a series of new derivatives with improved pharmacodynamic and pharmacokinetic profiles that enabled us to validate Slack as a pharmacological target in vivo. One of these new Slack activators, compound 6, exhibited negligible dopamine D2 and D3 receptor binding, unlike loxapine. We found that compound 6 displayed potent on-target antipruritic activity in multiple mouse models of acute histamine-independent and chronic itch without motor side effects. These properties make compound 6 a lead molecule for the development of new antipruritic therapies targeting Slack.

pharmacology and toxicology↗

bioRxiv2023

Evaluating interchain hydrogen bonds between collagens- A multiple alignment approach

Collagens are structural proteins that are predominantly found in the extracellular matrix of multicellular animals, where they are mainly responsible for the stability and structural integrity of various tissues. All collagens contain polypeptide strands ([a]-chains). There are several types of collagens, some of which differ significantly in form, function, and tissue specificity. Because of their importance in clinical research, they are grouped into subdivisions, the so-called collagen families, and their sequences are often analysed. However, problems arise with highly homologous sequence segments. To increase the accuracy of collagen classification and prediction of their functions, the structure of these collagens and their expression in different tissues could result in a better focus on sequence segments of interest. Here, we analyse collagen families with different levels of conservation. As a result, clusters with high interconnectivity can be found, such as the fibrillar collagens, the COL4 network-forming collagens, and the COL9 FACITs. Furthermore, a large cluster between network-forming, FACIT, and COL28a1 [a]-chains is formed with COL6a3 as a major hub node. The formation of clusters also signifies, why it is important to always analyse the [a]-chains and why structural changes can have a wide range of effects on the body.

systems biology↗

bioRxiv2023

Lattice-Boltzmann Modelling for Inertial Particle Microfluidics Applications -A Tutorial Review

Inertial particle microfluidics (IPMF) is an emerging technology for the manipulation and separation of microparticles and biological cells. Since the flow physics of IPMF is complex and experimental studies are often time-consuming or costly, computer simulations can offer complementary insights. In this tutorial review, we provide a guide for researchers who are exploring the potential of the lattice-Boltzmann (LB) method for simulating IPMF applications. We first review the existing literature to establish the state of the art of LB-based IPMF modelling. After summarising the physics of IPMF, we then present related methods used in LB models for IPMF and show several case studies of LB simulations for a range of IPMF scenarios. Finally, we conclude with an outlook and several proposed research directions.

biophysics↗

bioRxiv2022

Structural characterization of human tryptophanhydroxylase 2 reveals L-Phe as the superior regulatorydomain ligand relevant for serotonin biosynthesis

Tryptophan hydroxylase 2 (TPH2) catalyzes the rate-limiting step in the biosynthesis of serotonin in the brain. Consequently, regulation of TPH2 is relevant for serotonin related diseases, yet, the regulatory mechanism of TPH2 is poorly understood and structural as well as dynamical insights are missing. Here, we use NMR spectroscopy to determine the structure of a 47 N-terminally truncated variant of the regulatory domain (RD) dimer of human TPH2 in complex with L-Phe, and show that L-Phe is the superior RD ligand compared to the natural substrate, L-Trp. Using cryo-EM we obtain a low-resolution structure of a similarly truncated variant of the complete tetrameric enzyme with dimerized RDs. The cryo-EM 2D class averages additionally indicate that the RDs are dynamic in the tetramer and likely exist in a monomer-dimer equilibrium. Our results provide structural information on the RD both as an isolated domain and in the TPH2 tetramer, which will facilitate future elucidation of TPH2s regulatory mechanism affecting serotonin regulation.

biophysics↗

bioRxiv2022

Limb, joint and pelvic kinematic control in the quail coping with step perturbations

Small cursorial birds display remarkable walking skills and can negotiate complex and unstructured terrains with ease. The neuromechanical control strategies necessary to adapt to these challenging terrains are still not well understood. Here, we analyzed the 2D- and 3D pelvic and leg kinematic strategies employed by the common quail to negotiate visible step-up and step-down perturbations of 1 cm, 2.5 cm, and 5 cm. We used biplanar fluoroscopy to accurately describe joint positions in three dimensions and performed semi-automatic landmark localization using deep learning. Quails negotiated vertical perturbations without major problems and rapidly regained steady-state locomotion. When coping with step-up perturbations, the quail mostly adapted the trailing limb to permit the leading leg to step on the elevated substrate in a similar way as it did during level locomotion. When the quail negotiated step-down perturbations, both legs showed significant adaptations. For small and moderate perturbations (not inducing aerial running) the quail kept the function of the distal joints (i.e., their kinematic pattern) largely unchanged during uneven locomotion, and most changes occurred in proximal joints. The hip regulated leg length, while the distal joints maintained the spring-damped limb patterns. However, to negotiate the largest visible step perturbations, more dramatic kinematic alterations were observed. For these large perturbations, all joints contributed to leg lengthening/ shortening in the trailing leg and both the trailing and leading legs stepped more vertically and less abducted. This indicates a shift from a dynamic walking program to strategies that are focused on maximizing safety.

animal behavior and cognition↗

bioRxiv2021

Conformational rearrangements upon start codon recognition in human 48S translation initiation complex

Selection of the translation start codon is a key step during protein synthesis in human cells. We obtained cryo-EM structures of human 48S initiation complexes and characterized the intermediates of codon recognition by kinetic methods using eIF1A as a reporter. Both approaches capture two distinct ribosome populations formed on an mRNA with a cognate AUG codon in the presence of eIF1, eIF1A, eIF2-GTP-Met-tRNAiMet, and eIF3. The open 40S subunit conformation differs from the human 48S scanning complex and represents an intermediate preceding the codon recognition step. The closed form is similar to reported structures of complexes from yeast and mammals formed upon codon recognition, except for the orientation of eIF1A, which is unique in our structure. Kinetic experiments show how various initiation factors mediate the population distribution of open and closed conformations until 60S subunit docking. Our results provide insights into the timing and structure of human translation initiation intermediates and suggest the differences in the mechanisms of start codon selection between mammals and yeast.

molecular biology↗

bioRxiv2020

A three-dimensional musculoskeletal model of the dog

Dogs are an interesting object of investigation because of the wide range of body size, body mass, and physique. In the last several years, the number of clinical and biomechanical studies on dog locomotion has increased. However, the relationship between body structure and joint load during locomotion, as well as between joint load and degenerative diseases of the locomotor system (e.g. dysplasia), are not sufficiently understood. In vivo measurements/records of joint forces and loads or deep/small muscles are complex, invasive, and sometimes ethically questionable. The use of detailed musculoskeletal models may help in filling that knowledge gap. We describe here the methods we used to create a detailed musculoskeletal model with 84 degrees of freedom and 134 muscles. Our model has three key-features: Three-dimensionality, scalability, and modularity. We tested the validity of the model by identifying forelimb muscle synergies of a beagle at walk. We used inverse dynamics and static optimization to estimate muscle activations based on experimental data. We identified three muscle synergy groups by using hierarchical clustering. Predicted activation patterns exhibited good agreement with experimental data for most of the forelimb muscles. We expect that our model will speed up the analysis of how body size, physique, agility, and disease influence joint neuronal control and loading in dog locomotion.

bioengineering↗

bioRxiv2020

Calculating the optimal hematocrit under the constraint of constant cardiac power

In humans and higher animals, a trade-off between sufficiently high erythrocyte concentrations to bind oxygen and sufficiently low blood viscosity to allow rapid blood flow has been achieved during evolution. Optimal hematocrit theory has been successful in predicting hematocrit values of about 0.3 - 0.5, in very good agreement with the normal values observed for humans and many animal species. However, according to those calculations, the optimal value should be independent of the mechanical load of the body. This is in contradiction to the exertional increase in hematocrit observed in some animals called natural blood dopers and to the illegal practice of blood boosting in high-performance sports. In this study, we calculate the optimal hematocrit under two different constraints - under a constant driving pressure and under constant cardiac power – and show that the optimal hematocrit under constant cardiac power is higher than the normal value, ranging from 0.5 to 0.7. We use this result to explain the tendency to better exertional performance at an increased hematocrit.Statement of Significance In humans and higher animals, erythrocytes comprise a volume fraction (hematocrit) of 30-50 % of the blood. Mathematical calculations based on the assumption of constant blood pressure show that the optimal hematocrit value is indeed in that range. However, this optimum should apply to both rest and physical exertion, which is in contradiction to the increase in hematocrit observed in some animals called natural blood dopers and to the illegal practice of blood boosting in sports. Here, we calculate the optimal value based on the alternative constraint of constant cardiac power. We show that this results in a higher optimal value, ranging from 50 to 70 %. In this way, we explain the better exertional performance at an increased hematocrit.Competing Interest StatementThe authors have declared no competing interest.View Full Text

physiology↗

bioRxiv2020

Breaking the next Cryo-EM resolution barrier - Atomic resolution determination of proteins!

Single particle cryo-EM is a powerful method to solve the three-dimensional structures of biological macromolecules. The technological development of electron microscopes, detectors, automated procedures in combination with user friendly image processing software and ever-increasing computational power have made cryo-EM a successful and largely expanding technology over the last decade. At resolutions better than 4 [A], atomic model building starts becoming possible but the direct visualization of true atomic positions in protein structure determination requires significantly higher (< 1.5 [A]) resolution, which so far could not be attained by cryo-EM. The direct visualization of atom positions is essential for understanding protein-catalyzed chemical reaction mechanisms and to study drug-binding and -interference with protein function. Here we report a 1.25 [A] resolution structure of apoferritin obtained by cryo-EM with a newly developed electron microscope providing unprecedented structural details. Our apoferritin structure has almost twice the 3D information content of the current world record reconstruction (at 1.54 [A] resolution 1). For the first time in cryo-EM we can visualize individual atoms in a protein, see density for hydrogen atoms and single atom chemical modifications. Beyond the nominal improvement in resolution we can also show a significant improvement in quality of the cryo-EM density map which is highly relevant for using cryo-EM in structure-based drug design.

biophysics↗