Refine
Year of publication
- 2022 (330) (remove)
Document Type
- Preprint (330) (remove)
Language
- English (330)
Has Fulltext
- yes (330)
Is part of the Bibliography
- no (330)
Keywords
Institute
- Physik (190)
- Frankfurt Institute for Advanced Studies (FIAS) (144)
- Informatik (102)
- Medizin (28)
- Ernst Strüngmann Institut (25)
- Biowissenschaften (22)
- Senckenbergische Naturforschende Gesellschaft (8)
- MPI für Hirnforschung (7)
- Psychologie (6)
- Biochemie, Chemie und Pharmazie (5)
Motivation Expert curation to differentiate between functionally diverged homologs and those that may still share a similar function routinely relies on the visual interpretation of domain architecture changes. However, the size of contemporary data sets integrating homologs from hundreds to thousands of species calls for alternate solutions. Scoring schemes to evaluate domain architecture similarities can help to automatize this procedure, in principle. But existing schemes are often too simplistic in the similarity assessment, many require an a-priori resolution of overlapping domain annotations, and those that allow overlaps to extend the set of annotations sources cannot account for redundant annotations. As a consequence, the gap between the automated similarity scoring and the similarity assessment based on visual architecture comparison is still too wide to make the integration of both approaches meaningful.
Results Here, we present FAS, a scoring system for the comparison of multi-layered feature architectures integrating information from a broad spectrum of annotation sources. Feature architectures are represented as directed acyclic graphs, and redundancies are resolved in the course of comparison using a score maximization algorithm. A benchmark using more than 10,000 human-yeast ortholog pairs reveals that FAS consistently outperforms existing scoring schemes. Using three examples, we show how automated architecture similarity assessments can be routinely applied in the benchmarking of orthology assignment software, in the identification of functionally diverged orthologs, and in the identification of entries in protein collections that most likely stem from a faulty gene prediction.
Tree bark constitutes ideal habitat for microbial communities, because it is a stable substrate, rich in micro-niches. Bacteria, fungi, and terrestrial microalgae together form microbial communities, which in turn support more bark-associated organisms, such as mosses, lichens, and invertebrates, thus contributing to forest biodiversity. We have a limited understanding of the diversity and biotic interactions of the bark-associated microbiome, as investigations have mainly focussed on agriculturally relevant systems and on single taxonomic groups. Here we implemented a multi-kingdom metabarcoding approach to analyse diversity and community structure of the green algal, bacterial, and fungal components of the bark-associated microbial communities of beech, the most common broadleaved tree of Central European forests. We identified the most abundant taxa, hub taxa, and co-occurring taxa. We found that tree size (as a proxy for age) is an important driver of community assembly, suggesting that environmental filtering leads to less diverse fungal and algal communities over time. Conversely, forest management intensity had negligible effects on microbial communities on bark. Our study suggests the presence of undescribed, yet ecologically meaningful taxa, especially in the fungi, and highlights the importance of bark surfaces as a reservoir of microbial diversity. Our results constitute a first, essential step towards an integrated framework for understanding microbial community assembly processes on bark surfaces, an understudied habitat and neglected component of terrestrial biodiversity. Finally, we propose a cost-effective sampling strategy to study bark-associated microbial communities across large spatial or environmental scales.
The pitfalls of measuring representational similarity using representational similarity analysis
(2022)
A core challenge in cognitive and brain sciences is to assess whether different biological systems represent the world in a similar manner. Representational Similarity Analysis (RSA) is an innovative approach to address this problem and has become increasingly popular across disciplines ranging from artificial intelligence to computational neuroscience. Despite these successes, RSA regularly uncovers difficult-to-reconcile and contradictory findings. Here, we demonstrate the pitfalls of using RSA and explain how contradictory findings arise due to false inferences about representational similarity based on RSA-scores. In a series of studies that capture increasingly plausible training and testing scenarios, we compare neural representations in computational models, primate cortex and human cortex. These studies reveal two problematic phenomena that are ubiquitous in current research: a “mimic” effect, where confounds in stimuli can lead to high RSA-scores between provably dissimilar systems, and a “modulation effect”, where RSA-scores become dependent on stimuli used for testing. Since our results bear on a number of influential findings and the inferences drawn by current practitioners in a wide range of disciplines, we provide recommendations to avoid these pitfalls and sketch a way forward to a more solid science of representation in cognitive systems.
The pitfalls of measuring representational similarity using representational similarity analysis
(2022)
A core challenge in neuroscience is to assess whether diverse systems represent the world similarly. Representational Similarity Analysis (RSA) is an innovative approach to address this problem and has become increasingly popular across disciplines from machine learning to computational neuroscience. Despite these successes, RSA regularly uncovers difficult-to-reconcile and contradictory findings. Here we demonstrate the pitfalls of using RSA to infer representational similarity and explain how contradictory findings arise and support false inferences when left unchecked. By comparing neural representations in primate, human and computational models, we reveal two problematic phenomena that are ubiquitous in current research: a “mimic” effect, where confounds in stimuli can lead to high RSA scores between provably dissimilar systems, and a “modulation effect”, where RSA-scores become dependent on stimuli used for testing. Since our results bear on existing findings and inferences, we provide recommendations to avoid these pitfalls and sketch a way forward.
Some pitfalls of measuring representational similarity using Representational Similarity Analysis
(2022)
A core challenge in cognitive and brain sciences is to assess whether different biological systems represent the world in a similar manner. Representational Similarity Analysis (RSA) is an innovative approach that addresses this problem by looking for a second-order isomorphisim in neural activation patterns. This innovation makes it easy to compare latent representations across individuals, species and computational models, and accounts for its popularity across disciplines ranging from artificial intelligence to computational neuroscience. Despite these successes, using RSA has led to difficult-to-reconcile and contradictory findings, particularly when comparing primate visual representations with deep neural networks (DNNs): even though DNNs have been shown to learn and behave in vastly different ways to humans, comparisons based on RSA have shown striking similarities in some studies. Here, we demonstrate some pitfalls of using RSA and explain how contradictory findings can arise due to false inferences about representational similarity based on RSA-scores. In a series of studies that capture increasingly plausible training and testing scenarios, we compare neural representations in computational models, primate cortex and human cortex. These studies reveal two problematic phenomena that are ubiquitous in current research: a “mimic effect”, where confounds in stimuli can lead to high RSA-scores between provably dissimilar systems, and a “modulation effect”, where RSA-scores become dependent on stimuli used for testing. Since our results bear on a number of influential findings, such as comparisons made between human visual representations and those of primates and DNNs, we provide recommendations to avoid these pitfalls and sketch a way forward to a more solid science of representation in cognitive systems.
We deal with the reconstruction of inclusions in elastic bodies based on monotonicity methods and construct conditions under which a resolution for a given partition can be achieved. These conditions take into account the background error as well as the measurement noise. As a main result, this shows us that the resolution guarantees depend heavily on the Lamé parameter μ and only marginally on λ.
Holography has provided valuable insights into the time evolution of strongly coupled gauge theories in a fixed spacetime. However, this framework is insufficient if this spacetime is dynamical. We present a scheme to evolve a four-dimensional, strongly interacting gauge theory coupled to four-dimensional dynamical gravity in the semiclassical regime. As in previous work, we use holography to evolve the quantum gauge theory stress tensor, whereas the four-dimensional metric evolves according to Einstein's equations coupled to the expectation value of the stress tensor. The novelty of our approach is that both the boundary and the bulk spacetimes are constructed dynamically, one time step at a time. We focus on Friedmann-Lemaître-Robertson-Walker geometries and evolve far-from-equilibrium initial states that lead to asymptotically expanding, flat or collapsing Universes.
Quantitative MRI maps of human neocortex explored using cell type-specific gene expression analysis
(2022)
Quantitative MRI (qMRI) allows extraction of reproducible and robust parameter maps. However, the connection to underlying biological substrates remains murky, especially in the complex, densely packed cortex. We investigated associations in human neocortex between qMRI parameters and neocortical cell types by comparing the spatial distribution of the qMRI parameters longitudinal relaxation rate (R1), effective transverse relaxation rate (R2∗), and magnetization transfer saturation (MTsat) to gene expression from the Allen Human Brain Atlas, then combining this with lists of genes enriched in specific cell types found in the human brain. As qMRI parameters are magnetic field strength-dependent, the analysis was performed on MRI data at 3T and 7T. All qMRI parameters significantly covaried with genes enriched in GABA- and glutamatergic neurons, i.e. they were associated with cytoarchitecture. The qMRI parameters also significantly covaried with the distribution of genes enriched in astrocytes (R2∗ at 3T, R1 at 7T), endothelial cells (R1 and MTsat at 3T), microglia (R1 and MTsat at 3T, R1 at 7T), and oligodendrocytes (R1 at 7T). These results advance the potential use of qMRI parameters as biomarkers for specific cell types.
Neural computations emerge from recurrent neural circuits that comprise hundreds to a few thousand neurons. Continuous progress in connectomics, electrophysiology, and calcium imaging require tractable spiking network models that can consistently incorporate new information about the network structure and reproduce the recorded neural activity features. However, it is challenging to predict which spiking network connectivity configurations and neural properties can generate fundamental operational states and specific experimentally reported nonlinear cortical computations. Theoretical descriptions for the computational state of cortical spiking circuits are diverse, including the balanced state where excitatory and inhibitory inputs balance almost perfectly or the inhibition stabilized state (ISN) where the excitatory part of the circuit is unstable. It remains an open question whether these states can co-exist with experimentally reported nonlinear computations and whether they can be recovered in biologically realistic implementations of spiking networks. Here, we show how to identify spiking network connectivity patterns underlying diverse nonlinear computations such as XOR, bistability, inhibitory stabilization, supersaturation, and persistent activity. We established a mapping between the stabilized supralinear network (SSN) and spiking activity which allowed us to pinpoint the location in parameter space where these activity regimes occur. Notably, we found that biologically-sized spiking networks can have irregular asynchronous activity that does not require strong excitation-inhibition balance or large feedforward input and we showed that the dynamic firing rate trajectories in spiking networks can be precisely targeted without error-driven training algorithms.
The exploration of hot and dense nuclear matter: Introduction to relativistic heavy-ion physics
(2022)
This article summarizes our present knowledge about nuclear matter at the highest energy densities and its formation in relativistic heavy ion collisions. We review what is known about the structure and properties of the quark-gluon plasma and survey the observables that are used to glean information about it from experimental data.