Frankfurt Institute for Advanced Studies

We introduce the cooperative network architecture (CNA), a model that represents sensory signals using structured, recurrently connected networks of neurons, termed "nets." Nets are dynamically assembled from overlapping net fragments, which are learned based on statistical regularities in sensory input. This architecture offers robustness to noise, deformation, and generalization to out-of-distribution data, addressing challenges in current vision systems from a novel perspective. We demonstrate that net fragments can be learned without supervision and flexibly recombined to encode novel patterns, enabling figurecompletion and resilience to noise. Our findings establish CNA as a promising paradigm for developing neural representations that integrate local feature processing with global structure formation, providing a foundation for future research on invariant object recognition.

Gravity waves (GWs) are well known for their role in shaping large-scale dynamics of the atmosphere, but they also induce strong local variability in the vertical velocity, temperature, and other fields. Such variability is often omitted when it comes to global effects due to averaging and resolution limitations. However, small-scale dynamics, such as gravity waves, have a crucial role in cirrus microphysics and life cycle. Ice clouds, on the other hand, can have a pronounced effect on the Earth’s radiation budget and global moisture distribution, making their accurate representation in climate and numerical weather prediction (NWP) models particularly important.This work investigates the effects of gravity waves on cirrus cloud microphysics using the global ICON (Icosahedral Nonhydrostatic) model. A novel, self-consistent parameterization of GW-induced homogeneous ice nucleation developed by Dolaptchiev et al. (2023) is employed, and additional GW effects on depositional ice growth are considered. The local GW field is represented using the Multi-Scale Gravity Wave Model (MS-GWaM), which supports multiple GW source types and three-dimensional wave propagation, thereby enhancing the physical realism of the parameterized GW dynamics. The full coupling of GW forcing, along with feedback from the supplemented ice scheme into the overall microphysics and radiation schemes, has been implemented and assessed within the ICON model.The results of the global test runs reveal significant GW impacts on ice formation mechanisms, leading to enhanced homogeneous nucleation in the upper troposphere–lower stratosphere (UTLS) compared to the baseline ICON configuration. Furthermore, GW-induced temperature fluctuations obtained from MS-GWaM and coupled online to depositional growth substantially increase ice growth efficiency. It results in larger ice mixing ratios in the mid-latitudes and subtropical regions. Further analyses are planned to assess the sensitivity of the coupled version to different MS-GWaM configurations, the role of lateral GW propagation, and the relative contributions of different gravity wave sources.

ABSTRACT A group of researchers from the humanities, economics, social sciences, natural and life science developed a definition of the topics complexity and emergence that can be applied across disciplines. Here, concepts of complexity and emergence in chemistry and biochemistry are discussed, to promote a discourse between the natural and life sciences and philosophy. Although chemical research often employs reductionist strategies, the properties of molecules and their linked functions exhibit emergent properties that cannot be inferred solely from their atomic constituents. Assembly theory and the work of Manfred Eigen offer ways to quantify and predict emergence in chemistry, particularly in relation to the origins and evolution of life. This review emphasizes the chemical prerequisites for life, such as the formation of natural products, the emergence of nucleic acids that carry information, and the functional roles of proteins. From a philosophical standpoint, modern ontology provides a means of understanding reality that is both process‐based and subject‐independent. By integrating chemistry, biology and philosophy, the synopsis of this review addresses the predictive, post facto and historically unique aspects of complex systems, offering a conceptual framework for comprehending the emergence of molecular function and the evolution of living systems.

BackgroundAccurate cellular models are critical for understanding tissue function and accelerating drug discovery. While in vitro systems like spheroids and organoids are widely used, they are costly, low-throughput, and often lack reproducibility. Here, we introduce Evolvoid, a computational pipeline based on genetic algorithms that generates virtual 3D spheroid-like cell constructs with optimized morphologies by simulating key biophysical principles — such as thermodynamic principles, nutrient transport and their uptake by cells.ResultsBy integrating finite element simulations with evolutionary principles, our in silico platform evolves populations of randomly generated shapes, iteratively selecting individuals that best satisfy a biophysically informed fitness function. The fitness function encodes selection pressures based on general biophysical constraints — namely, the minimization of surface energy and the maintenance of cell viability under variable oxygen conditions. The Shannon entropy is used to track genome complexity over generations. In consistence with evolutionary dynamics, the complexity increases progressively. Moreover, albeit based on basic biophysical optimization rules, the fittest individuals generated by Evolvoid closely resemble the shape and size of cellular spheroids obtained in vitro.ConclusionEvolvoid is modular, scalable, and tunable to different construct sizes, cell types, or culture conditions, and thus provides a versatile platform for designing and optimizing 3D cellular models entirely in silico. It offers a foundation for developing high-fidelity, cost-effective digital shape twins of biological systems, supporting the advancement of lab-on-a-laptop technologies and new approach methodologies, thereby reducing the reliance on in vitro and animal models.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13036-026-00642-z.

Diffuse large B‑cell lymphomas are highlighted within the group of high-gradeB-cell lymphomas, as they represent one of the most common lymphoma types in our region. In addition, there are numerous rare entities that can often be suspected morphologically and clinically but need to be confirmed molecularly or through extended immunophenotyping panels. Hodgkin lymphomas are also common, characterized by afew typical tumor cells and apronounced variability in their microenvironment, leading to classification into characteristic subtypes.

Spatiotemporal changes in the nuclear lamina and cell metabolism shape cell fate, yet their interplay is poorly understood. Here we identify lamin A/C as a key regulator of cysteine catabolic flux essential for proper cell fate and longevity. Its loss in naive mouse pluripotent stem cells leads to upregulation of the cysteine-generating and catabolizing enzymes, cystathionine γ-lyase (CTH) and cystathionine β-synthase (CBS), thereby promoting de novo cysteine synthesis. Increased cysteine flux into acetyl-CoA fosters histone H3K9 and H3K27 acetylation, triggering a transition from naive to primed pluripotency and abnormal cell fate and function. Conversely, the toxic gain-of-function mutation of Lmna, encoding lamin A/C and associated with premature ageing, reduces CTH and CBS levels. This reroutes cysteine catabolic flux and alters the balance between H3K9 acetylation and methylation, crucially impacting germ layer formation and genome stability. Notably, modulation of Cth and Cbs rescues the abnormal cell fate and function, restores the DNA damage repair capacity and alleviates the senescent phenotype caused by lamin A/C mutations, highlighting the potential of modulating cell metabolism to mitigate epigenetic diseases.

Metazoan fatty acid (FA) synthases (mFASs) facilitate the de novo synthesis of C16- and C18-FAs through iterative extensions within the FA cycle and hydrolytic release. Here we re-engineer mFAS to fine-tune the interplay between FA extension and FA hydrolytic release for the targeted production of short- and medium-chain fatty acids. Single amino acid exchanges in the ketosynthase domain can redirect FA product profiles from predominantly C8 (G113W) to C8/C10 (G113F) and C12/C14 (G113M). Integration of a thioreductase domain enables the production of medium-chain fatty aldehydes and alcohols. We apply our approach for controlling chain length in FA biosynthesis to the microbial production of C10- and C12-FAs, translate it into a yeast cell factory and achieve C10/C12-FAs titers of 674 mg l-1 and 67% purity of total free FAs. Our work demonstrates a modular platform for programmable FA synthesis and paves the way toward sustainable bioproduction of valuable oleochemicals.

Polyketides constitute a large class of natural products with important biological activities and applications such as antibiotics, antitumor agents, pesticides, and pigments. Their biosynthesis is catalyzed by polyketide synthases (PKSs) which are multidomain enzymes evolutionarily related to fatty acid synthases (FASs). Despite their close homology in structure and the chemistry they perform, FASs and PKSs differ fundamentally in their catalytic programming: FASs run fully reducing elongation reactions to yield saturated fatty acids, while iterative PKSs execute reductions just in selected cycles, generating complex oxidized compounds. In this study, we aimed at engineering the metazoan FAS in its ketoreduction (KR) domain to switch from fully reducing to a nonreducing mode during chain elongation. Guided by recent insights into KR programming, we incorporated a helix into metazoan FAS, which is found in KRs from iterative PKSs and type II FASs with chain length programming. These FAS variants initially catalyze complete fatty acid cycles but lose the ability of β-keto reduction in later elongation rounds, producing intermediates that spontaneously cyclize to pyrone products. Finally, our study provides valuable insight into the mechanism of KR catalysis identifying another amino acid next to the active tyrosine which is capable for intermediate protonation.

Novel variants of known natural product (NP) classes can provide valuable insights into their biosynthesis, mechanisms of action, and potential as drug leads across the entire class. Here, we describe a novel member of the widespread detoxine/rimosamide‐like (DRL) natural products, named pseudotetraivprolide, produced by Pseudomonas strains. Pseudotetraivprolide exhibits the characteristic DRL‐activity of protecting Bacillus cereus against the antibiotic blasticidin S. Through the generation of multiple deletion and complementation mutants, heterologous expression experiments, identification and structure elucidation of several derivatives, chemical synthesis of main derivatives, enzymatic characterization of individual biochemical steps, and detailed homology modelling of enzyme complexes, we elucidated key aspects of its biosynthesis. Our findings demonstrate that the primary metabolism‐derived malonyl CoA:ACP transacylase (FabD) functions as a trans‐AT in the biosynthesis pathway. Furthermore, we suggest an order for all late‐stage modifications and assign a function for the three conserved hypothetical proteins PipDFG acting as last‐step acetylation complex responsible for stabilization and activation of the final product.