Biology subjects

Jansen, A.

Publications and source records attributed to Jansen, A..

At least 19 recordsLinked to original sources

bioRxiv2026

MillionTrees: A global benchmark for tree detection across geography, sensor, and annotation geometry

Airborne tree monitoring using remotely sensed data, combined with the growth of computer vision models for object detection and segmentation, can facilitate broad scale assessment of tree health, ecology and ecosystem function. However, datasets for airborne tree detection have been largely limited to a single geography, resolution, or annotation approach, making it difficult to develop general models and benchmark their utility for broad applications. We compile existing and newly published annotation datasets to create a comprehensive dataset for training and evaluating airborne tree detection in RGB imagery. The MillionTrees dataset has over a million individual trees from a broad range of forest types, tree species, and geographic locations. In addition, it has six million weakly supervised annotations from a range of sources including airborne LiDAR and photogrammetry, arborist surveys, and simulated landscapes. An evaluation standard is provided to facilitate rapid model development. We benchmark several existing tree detection algorithms and demonstrate that while current open-source models do not generalize across the diverse datasets in the benchmark, training on MillionTrees improves both within-domain and out-of-domain performance. Held-out validation datasets using ground truth derived from terrestrial LiDAR paired with overhead RGB show that overfitting to test data is common and hyperparameter tuning is an underappreciated challenge towards a global zero shot tree detection method. We discuss the limitations of the currently available data for airborne RGB tree detection, how to create more robust evaluation data and the need for better understanding of the uncertainty and utility of tree detection from airborne remote sensing.

ecology↗

bioRxiv2026

Secure Attachment despite Childhood Maltreatment: Behavioral and Neural Correlates of Interpersonal Resilience

BackgroundChildhood maltreatment is a traumatic interpersonal stressor that increases vulnerability for depression in adulthood. However, some individuals show secure attachment despite childhood maltreatment, a pattern that can be described as interpersonal resilience. The present study examined the behavioral and neural correlates of interpersonal resilience, defined as secure attachment in adults exposed to childhood maltreatment. MethodsWe analyzed structural 3T MRI data from 1,317 adults, including healthy participants and individuals with partially or fully remitted major depression. Gray matter volume was estimated from structural MRI data using voxel-based morphometry. Attachment style and childhood maltreatment were assessed using the Relationship Scales Questionnaire and the Childhood Trauma Questionnaire, respectively. A 2x2 design (childhood maltreatment by attachment style) tested main and interaction effects on behavioral outcomes and brain structure. ResultsInterpersonally resilient individuals with secure attachment and maltreatment reported significantly better mental health outcomes compared to insecurely attached adults with maltreatment. Differences included lower self-reported and rater-based depressive symptoms, lower global symptom severity, and higher global functioning. In the neuroimaging analyses, we identified a significant childhood maltreatment by attachment style interaction in the left supramarginal gyrus, with larger gray matter volume in resilient individuals compared to all other groups. This effect remained robust across multiple sensitivity analyses, controlling for medication load, antidepressant intake, diagnosis group, as well as in a complementary dimensional analysis. ConclusionsThe results identify a potential neural correlate of interpersonal resilience. Larger gray matter volume in the left supramarginal gyrus, a region previously implicated in perspective taking and self-other distinction among other functions, may be relevant to more adaptive interpersonal functioning after early adversity. Together with the robust behavioral effects, these findings are consistent with secure attachment as a protective factor that may be associated with attenuated effects of childhood maltreatment on mental health.

neuroscience↗

bioRxiv2026

Stratigraphic Paleobiology of Carbonate Systems

Stratigraphic paleobiology is a newly established interdisciplinary approach, which has demonstrated that the fossil record is a joint expression of biotic and stratigraphic change, and all inferences from it must be grounded in a solid understanding of the stratigraphic context. Fossiliferous strata can be found in all depositional systems (e.g., marine, terrestrial, or lacustrine; siliciclastics or carbonates), each having a unique characteristic timescale and set of external controls, which govern the accumulation of sedimentary particles, including fossils. Consequently, the same biotic changes are preserved differently across depositional systems. While carbonate systems form a large portion of the fossil record, most studies in stratigraphic paleobiology have focused on siliciclastic systems and are not easily generalizable. As they are predominantly formed by living organisms, carbonates are both fossils and record, opening the opportunity to study the co-dependency of life and its environment. Here, we explore the stratigraphic paleobiology of carbonate systems by combining simulations of carbonate platform and ramp geometries with synthetic fossil records. We explore the preservation of extinction patterns and rates spatially and across geometries. By examining stratigraphic biases in isolation (unconformity and condensation, ecology, and abundance biases), we find characteristic differences between ramp and platform geometries due to their differential response to sea level change, spatial variability, and differences in ecological clines. Differences in the structure of the fossil record between platform geometries are traceable to the contribution and properties of the carbonate producing organisms (carbonate factories), showing that preservation of earth system data in carbonate systems will vary both latitudinally and temporally or as a result of major perturbations of the biogeosphere. Our results show that while general rules on the structure of the fossil record can be derived for entire depositional systems, accounting for the geological and ecological dynamics of a particular sedimentary basin can hugely refine interpretations of the fossil record. That is particularly true for biogenic and biologically-mediated sediments.

paleontology↗

bioRxiv2026

Individual, behavioural, and environmental determinants of personal light exposure in daily life: A multi-country wearable and experience-sampling study

Light supports circadian regulation and is associated with non-communicable diseases, yet ocular exposure patterns remain poorly understood. We recruited 191 adults across nine sites in seven countries, combining diaries and contextual reports with near-eye (141 participants; 816 participant-days) and complementary chest-level measurements (154 participants; 902 participant-days). In near-eye analyses, only 24.0% of recorded daytime minutes met the recommendation of at least 250 lx melanopic equivalent daylight illuminance; 63.3% of pre-sleep and 87.7% of bedside sleep-environment minutes met their respective limits. Variation among people and days within sites exceeded that among sites (ratio 1.99; 95%-CI, 1.29-4.77). In hourly models, light source and setting contributed the largest shares of fitted variation after time of day. Site-average outdoor exposure was 9.35 times that while awake at home (95%-CI, 6.95-12.59; 714.2 versus 76.4 lx). This baseline identifies a daytime exposure gap and supports testing interventions in everyday settings to improve light exposure and health.

physiology↗

bioRxiv2026

Interpretable Hierarchical RNNs for rs-fMRI: Promise and Limits of Individualized Brain Dynamics

Modeling individual brain dynamics from resting-state fMRI (rs-fMRI) remains challenging due to substantial inter-subject variability, noise, and limited data length per subject. Here, we systematically evaluate whether hierarchical shallow piecewise-linear recurrent neural networks (shPLRNNs), recently introduced as interpretable dynamical system reconstruction models, can generate individualized rs-fMRI time series while preserving subject-specific functional connectivity structure. We applied the framework to 1,423 rs-fMRI samples from healthy participants of the Marburg-Munster Affective Disorders Cohort Study (MACS). Simulated rs-fMRI data reproduced substantial empirical FC structure, with comparable reconstruction accuracy on the validation and held-out test sets. Generalization to unseen individuals was heterogeneous and strongly depended on how typical a subjects connectivity pattern was relative to the training cohort, with template similarity explaining 37% of variance in reconstruction accuracy. Learned subject-specific parameters exhibited significant test-retest stability and higher within-subject than between-subject similarity on longitudinal data from two different timepoints, supporting their interpretation as individualized dynamical markers. Associations between individual parameters and demographic or cognitive variables were statistically significant but modest in effect size, and predictive performance remained below that obtained using empirical rs-fMRI features directly. Empirical FC was used as a reference for static subject information rather than as a target to be outperformed. Together, these results suggest that hierarchical shPLRNNs can extract meaningful and partially stable individual-specific dynamical structure from rs-fMRI data. The findings delineate key trade-offs between model expressivity, generalization and subject specificity, and point to directions for future methodological refinement in individualized brain modeling. Graphical AbstractA hierarchical dynamical RNN captures substantial individual rs-fMRI functional connectivity structure using compact subject-specific parameters embedded in shared population dynamics. The resulting representations generalize to held-out subjects and show test-retest stability, but only modest associations with phenotypic variables. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=150 SRC="FIGDIR/small/713153v3_ufig1.gif" ALT="Figure 1"> View larger version (61K): org.highwire.dtl.DTLVardef@1b029edorg.highwire.dtl.DTLVardef@90bdd6org.highwire.dtl.DTLVardef@9f6b77org.highwire.dtl.DTLVardef@486c9e_HPS_FORMAT_FIGEXP M_FIG C_FIG

neuroscience↗

bioRxiv2026

Mechanism of Alkaline Gating in a Pentameric Ion Channel

Pentameric ligand-gated ion channels (pLGICs) are key mediators of electrochemical signal transduction in various organisms. Like many proteins involved in cellular signaling, they are modulated by a variety of environmental factors including pH and small molecules. However, the molecular mechanisms underlying pLGIC activation and modulation remain unclear. A promising model system in this family is the bacterial ion channel sTeLIC, which can be activated by alkaline pH, and for which we recently determined structures in multiple functional states. However, protonation changes and other pH-driven dynamics cannot be directly observed in these structures. Here, we used constant-pH molecular dynamics simulations and oocyte-electrophysiology recordings from engineered mutants to develop a comprehensive mechanistic model for pH sensing. Interestingly, critical residues include two Glu residues (E106, E160) located in the extracellular-vestibule and domain-interface regions of each subunit, where they mediate differential electrostatic interactions in closed versus open states. This work demonstrates the applicability of constant-pH methods to model dynamic processes in a multimeric membrane-embedded protein, and offers a detailed mechanism for pH sensing, likely extensible to human drug targets.

biophysics↗

bioRxiv2026

Single-molecule identification of full-length proteins with single-amino-acid resolution using nanopores

Nanopore-based technologies show promise in single-molecule protein sequencing. By using an unfoldase and a nanopore with enhanced electroosmotic flow, here we show the continuous identification of generic proteins during single nanopore passes. This approach enables the recording of differences in charge and size from single amino acid substitutions compared to reference signals, paving the way for single-molecule protein sequencing and high-throughput proteomics.

biophysics↗

bioRxiv2025

Dissecting heterogeneity in cortical thickness abnormalities in major depressive disorder: a large-scale ENIGMA MDD normative modelling study

ImportanceMajor depressive disorder (MDD) is highly heterogeneous, with marked individual differences in clinical presentation and neurobiology, which may obscure identification of structural brain abnormalities in MDD. To explore this, we used normative modeling to index regional patterns of variability in cortical thickness (CT) across individual patients. ObjectiveTo use normative modeling in a large dataset from the ENIGMA MDD consortium to obtain individualised CT deviations from the norm (relative to age, sex and site) and examine the relationship between these deviations and clinical characteristics. Design, setting, and participantsA normative model adjusting for age, sex and site effects was trained on 35 CT measures from FreeSurfer parcellation of 3,181 healthy controls (HC) from 34 sites (40 scanners). Individualised z-score deviations from this norm for each CT measure were calculated for a test set of 2,119 HC and 3,645 individuals with MDD. For each individual, each CT z-score was classified as being within the normal range (95% of individuals) or within the extreme range (2.5% of individuals with the thinnest or thickest cortices). Main outcome measuresZ-score deviations of CT measures of MDD individuals as estimated from a normative model based on HC. ResultsZ-score distributions of CT measures were largely overlapping between MDD and HC (minimum 92%, range 92-98%), with overall thinner cortices in MDD. 34.5% of MDD individuals, and 30% of HC individuals, showed an extreme deviation in at least one region, and these deviations were widely distributed across the brain. There was high heterogeneity in the spatial location of CT deviations across individuals with MDD: a maximum of 12% of individuals with MDD showed an extreme deviation in the same location. Extreme negative CT deviations were associated with having an earlier onset of depression and more severe depressive symptoms in the MDD group, and with higher BMI across MDD and HC groups. Extreme positive deviations were associated with being remitted, of not taking antidepressants and less severe symptoms. Conclusions and relevanceOur study illustrates a large heterogeneity in the spatial location of CT abnormalities across patients with MDD and confirms a substantial overlap of CT measures with HC. We also demonstrate that individualised extreme deviations can identify protective factors and individuals with a more severe clinical picture. Key points QuestionCan z-scores derived from normative modelling shed light on the heterogeneous group-level findings of cortical thickness abnormalities in major depression and what characterises individuals at the extreme ends of cortical thickness abnormalities? FindingWe confirmed a large overlap in z-score distributions between depressed individuals and healthy controls and a heterogeneous spatial distribution of extreme z-deviations across brain regions across individual patients. Lower z-scores for cortical thickness were related to more severe clinical characteristics. MeaningOur findings confirm the heterogeneity in individual variation in the location and extent of CT abnormalities across patients with MDD and stress the importance of individualised predictions when examining cortical thickness abnormalities.

neuroscience↗

bioRxiv2025

Delineating transcriptomic signatures of in vitro human skeletal musclemodels in comparison to in vivo references

A pivotal question at the heart of stem cell research is how faithful cellular models recapitulate the biology of human tissues. Skeletal muscle is the largest tissue in the human body and has been extensively modelled by various in vitro systems. Here, we sought to delineate the state-of-the-art of in vitro human skeletal muscle models by performing a large-scale analysis of transcriptome datasets, covering more than 400 samples across 39 studies, including bulk and single cell RNA sequencing of 2D and 3D models and their in vivo counterparts. Our analyses highlight common discrepancies between a wide range of cellular models and human skeletal muscle with respect to their myogenic identity, transcription factors, epigenetic complexes, metabolic processes and signaling pathways. Our analyses reveal cellular processes that can be modulated to improve the in vitro models of skeletal muscle, while also paving the way for similar studies for other cell types.

cell biology↗

bioRxiv2024

Gray matter correlates of childhood maltreatment in the context of major depression: searching for replicability in a multi-cohort brain-wide association study of 3225 adults

Childhood maltreatment effects on cerebral gray matter have been frequently discussed as a neurobiological pathway for depression. However, localizations are highly heterogeneous, and recent reports have questioned the replicability of mental health neuroimaging findings. Here, we investigate the replicability of gray matter correlates of maltreatment (measured retrospectively via the Childhood Trauma Questionnaire) across three large adult cohorts (total N=3225). Pooling cohorts revealed maltreatment-related gray matter reductions, with most extensive effects when not controlling for depression diagnosis (maximum partial R2=.022). However, none of these effects significantly replicated across cohorts. Non-replicability was consistent across a variety of maltreatment subtypes and operationalizations, as well as subgroup analyses with and without depression, and stratified by sex. In this work we show that there is little evidence for the replicability of gray matter correlates of childhood maltreatment, when adequately controlling for psychopathology. This underscores the need to focus on replicability research in mental health neuroimaging.

neuroscience↗

bioRxiv2024

Brain Structural Correlates of an Impending Initial Major Depressive Episode

BackgroundNeuroimaging research has yet to elucidate, whether reported gray matter volume (GMV) alterations in major depressive disorder (MDD) exist already before the onset of the first episode. Recruitment of presently healthy individuals with a known future transition to MDD (converters) is extremely challenging but crucial to gain insights into neurobiological vulnerability. Hence, we compared converters to patients with MDD and sustained healthy controls (HC) to distinguish pre-existing neurobiological markers from those emerging later in the course of depression. MethodsCombining two clinical cohorts (n=1709), voxel-wise GMV of n=45 converters, n=748 patients with MDD, and n=916 HC were analyzed in regions-of-interest approaches. By contrasting the subgroups and considering both remission state and reported recurrence at a 2-year clinical follow-up, we stepwise disentangled effects of 1) vulnerability, 2) the acute depressive state, and 3) an initial vs. a recurrent episode. ResultsAnalyses revealed higher amygdala GMV in converters relative to HC (pTFCE-FWE=.037, d=0.447) and patients (pTFCE-FWE=.005, d=0.508), remaining significant when compared to remitted patients with imminent recurrence. Lower GMV in the dorsolateral prefrontal cortex (pTFCE-FWE<.001, d=0.188) and insula (pTFCE-FWE=.010, d=0.186) emerged in patients relative to HC but not to converters, driven by patients with acute MDD. ConclusionBy examining one of the largest available converter samples in psychiatric neuroimaging, this study allowed a first determination of neural markers for an impending initial depressive episode. Our findings suggest a temporary vulnerability, which in combination with other common risk factors might facilitate prediction and in turn improve prevention of depression.

neuroscience↗

bioRxiv2023

Constant-pH molecular dynamics simulations of progressive activation in a proton-gated ion channel

Although traditional molecular dynamics simulations successfully capture a variety of different molecular interactions, the protonation states of titratable residues are kept static. A recent constant-pH molecular dynamics implementation in the GROMACS package allows pH effects to be captured dynamically, and promises to provide both the accuracy and computational performance required for studying pH-mediated conformational dynamics in large, complex systems containing hundreds of titratable residues. Here, we demonstrate the applicability of this constant-pH implementation by simulating the proton-gated ion channel GLIC at resting and activating pH, starting from closed and open structures. Our simulations identify residues E26 and E35 as especially pH-sensitive and reveal state-dependent pKa shifts at multiple residues, as well as side chain and domain rearrangements in line with the early stages of gating. Our results are consistent with several previous experimental findings, demonstrating the applicability of constant-pH simulations to elucidate pH-mediated activation mechanisms in multidomain membrane proteins, likely extensible to other complex systems. Significance statementElectrostatic interactions play important roles in protein structure and function. Since changes in pH will (de)protonate residues and thereby modify such interactions, pH itself is a critical environmental parameter. However, protonation states of titratable residues are static during classical molecular dynamics simulations. Recently, a constant-pH algorithm was implemented in the GROMACS package, allowing pH effects to be captured dynamically. Here, we used this implementation to perform constant-pH simulations of the proton-gated ion channel GLIC, providing insight into its activation mechanism by revealing state-dependent shifts in protonation as well as pH-dependent side chain and domain-level rearrangements. The results show that constant-pH simulations are accurate, efficient, and capable of modeling dozens of titratable sites, with important implications for e.g. drug design.

biophysics↗

bioRxiv2023

Cellular insights of beech leaf disease reveal abnormal ectopic cell division of symptomatic interveinal leaf areas

The beech leaf disease nematode, Litylenchus crenatae subsp. mccannii, is recognized as a newly emergent nematode species that causes beech leaf disease (BLD) in beech trees (Fagus spp.) in North America. Changes of leaf morphology induced by BLD can provoke dramatic effects into the leaf architecture and consequently to tree performance and development. The initial symptoms of BLD appear as dark green interveinal banding patterns of the leaf. Despite the fast progression of this disease, the cellular mechanisms leading to the formation of such type of aberrant leaf phenotype remains totally unknown. To understand the cellular basis of BLD, we employed several microscopy approaches to provide an exhaustive characterization of nematode-infected buds and leaves. Histological sections revealed a dramatic cell change composition of these nematode-infected tissues. Diseased bud scale cells were typically hypertrophied and showed a high variability of size. Moreover, while altered cell division had no influence on leaf organogenesis, induction of cell proliferation on young leaf primordia led to a dramatic change in cell layer architecture. Hyperplasia and hypertrophy of the different leaf cell layers, coupled with an abnormal proliferation of chloroplasts specially in the spongy mesophyll cells, resulted in the typical interveinal leaf banding. These discrepancies in leaf cell structure were depicted by an abnormal rate of cellular division of the leaf interveinal areas infected by the nematode, promoting significant increase of cell size and leaf thickness. The formation of symptomatic BLD leaves is therefore orchestrated by distinct cellular processes, to enhance the value of these feeding sites and to improve their nutrition status to the nematode. These results revealed a high specialized mode of parasitism of L. crenatae subsp. mccannii.

cell biology↗

bioRxiv2023

Fatigue during acute systemic inflammation is associated with reduced mental effort expenditure while task accuracy is preserved

BackgroundEarlier work within the physical domain showed that acute inflammation changes motivational prioritization and effort allocation rather than physical abilities. It is currently unclear whether a similar motivational framework accounts for the mental fatigue and cognitive symptoms of acute sickness. Accordingly, this study aimed to assess the relationship between fatigue, cytokines and mental effort-based decision making during acute systemic inflammation. MethodsEighty-five participants (41 males; 18-30 years (M=23.0, SD=2.4)) performed a mental effort-based decision-making task before, 2 hours after, and 5 hours after intravenous administration of 1 ng/kg bacterial lipopolysaccharide (LPS) to induce systemic inflammation. Plasma concentrations of cytokines (interleukin (IL)-6, IL-8 and tumor necrosis factor (TNF)) and fatigue levels were assessed at similar timepoints. In the task, participants decided whether they wanted to perform (i.e., accepted) arithmetic calculations of varying difficulty (3 levels: easy, medium, hard) in order to obtain rewards (3 levels: 5, 6 or 7 points). Acceptance rates were analyzed using a binomial generalized estimated equation (GEE) approach with effort, reward and time as independent variables. Arithmetic performance was measured per effort level prior to the decisions and included as a covariate. Associations between acceptance rates, fatigue (self-reported) and cytokine concentrations levels were analyzed using partial correlation analyses. ResultsPlasma cytokine concentrations and fatigue were increased at 2 hours post-LPS compared to baseline and 5 hours post-LPS administration. Acceptance rates decreased for medium, but not for easy or hard effort levels at 2 hours post-LPS versus baseline and 5 hours post-LPS administration, irrespective of reward level. This reduction in acceptance rates occurred despite improved accuracy on the arithmetic calculations itself. Reduced acceptance rates for medium effort were associated with increased fatigue, but not with increased cytokines. ConclusionFatigue during acute systemic inflammation is associated with alterations in mental effort allocation, similarly as observed previously for physical effort-based choice. Specifically, willingness to exert mental effort depended on effort and not reward information, while task accuracy was preserved. These results extend the motivational account of inflammation to the mental domain and suggest that inflammation may not necessarily affect domain-specific mental abilities, but rather affects domain-general effort-allocation processes.

neuroscience↗

bioRxiv2023

Cross-validation for the estimation of effect size generalizability in mass-univariate brain-wide association studies

IntroductionStatistical effect sizes are systematically overestimated in small samples, leading to poor generalizability and replicability of findings in all areas of research. Due to the large number of variables, this is particularly problematic in neuroimaging research. While cross-validation is frequently used in multivariate machine learning approaches to assess model generalizability and replicability, the benefits for mass-univariate brain analysis are yet unclear. We investigated the impact of cross-validation on effect size estimation in univariate voxel-based brain-wide associations, using body mass index (BMI) as an exemplary predictor. MethodsA total of n=3401 adults were pooled from three independent cohorts. Brain-wide associations between BMI and gray matter structure were tested using a standard linear mass-univariate voxel-based approach. First, a traditional non-cross-validated analysis was conducted to identify brain-wide effect sizes in the total sample (as an estimate of a realistic reference effect size). The impact of sample size (bootstrapped samples ranging from n=25 to n=3401) and cross-validation on effect size estimates was investigated across selected voxels with differing underlying effect sizes (including the brain-wide lowest effect size). Linear effects were estimated within training sets and then applied to unseen test set data, using 5-fold cross-validation. Resulting effect sizes (explained variance) were investigated. ResultsAnalysis in the total sample (n=3401) without cross-validation yielded mainly negative correlations between BMI and gray matter density with a maximum effect size of R2p=.036 (peak voxel in the cerebellum). Effects were overestimated exponentially with decreasing sample size, with effect sizes up to R2p=.535 in samples of n=25 for the voxel with the brain-wide largest effect and up to R2p=.429 for the voxel with the brain-wide smallest effect. When applying cross-validation, linear effects estimated in small samples did not generalize to an independent test set. For the largest brain-wide effect a minimum sample size of n=100 was required to start generalizing (explained variance >0 in unseen data), while n=400 were needed for smaller effects of R2p =.005 to generalize. For a voxel with an underlying null effect, linear effects found in non-cross-validated samples did not generalize to test sets even with the maximum sample size of n=3401. Effect size estimates obtained with and without cross-validation approached convergence in large samples. DiscussionCross-validation is a useful method to counteract the overestimation of effect size particularly in small samples and to assess the generalizability of effects. Train and test set effect sizes converge in large samples which likely reflects a good generalizability for models in such samples. While linear effects start generalizing to unseen data in samples of n>100 for large effect sizes, the generalization of smaller effects requires larger samples (n>400). Cross-validation should be applied in voxel-based mass-univariate analysis to foster accurate effect size estimation and improve replicability of neuroimaging findings. We provide open-source python code for this purpose (https://osf.io/cy7fp/?view_only=a10fd0ee7b914f50820b5265f65f0cdb).

neuroscience↗

bioRxiv2022

Network controllability measures of subnetworks: implications for neurosciences

Recent progress in network sciences has made it possible to apply key findings from control theory to the study of networks. Referred to as network control theory, this framework describes how the interactions between interconnected system elements and external energy sources, potentially constrained by different optimality criteria, result in complex network behavior. A typical example is the quantification of the functional role certain brain regions or symptoms play in shaping the temporal dynamics of brain activity or the clinical course of a disease, a property that is quantified in terms of the so-called controllability metrics. Critically though, contrary to the engineering context in which control theory was originally developed, a mathematical understanding of the network nodes and connections in neurosciences cannot be assumed. For instance, in the case of psychological systems such as those studied to understand the psychiatric disorders, a potentially large set variables are unknown. As such, while the measures offered by network control theory would be mathematically correct, in that they can be calculated with high precision, they could have little translational values with respect to their putative role suggested by controllability metrics. It is therefore critical to understand if and how the controllability metrics computer over subnetworks would deviate, if access to the complete set of variables, as in neurosciences, cannot be taken for granted. In this paper, we use a host of simulations based on synthetic as well as structural MRI data to study the potential deviation of controllability metrics in sub-compared to the full networks. Specifically, we estimate average- and modal-controllability, two of the most widely used controllability measures in neurosciences, in a large number of settings where we systematically vary network type, network size, and edge density. We find out, across all network types we test, that average and modal controllability are systematically, either over- or underestimated depending on the number of nodes in the sub- and full network and the edge density. Finally, we provide a formal theoretical proof that our observations generalize to any network type and discuss the ramifications of this systematic bias and potential solutions to alleviate the problem.

neuroscience↗

bioRxiv2022

Let's face it: The lateralization of the face perception network as measured with fMRI is not clearly right dominant

The neural face perception network is distributed across both hemispheres. However, the dominant role in humans is virtually unanimously attributed to the right hemisphere. Interestingly, there are, to our knowledge, no imaging studies that systematically describe the distribution of hemispheric lateralization in the core system of face perception across subjects in large cohorts so far. To address this, we determined the hemispheric lateralization of all core system regions (i.e., occipital face area (OFA), fusiform face area (FFA), posterior superior temporal sulcus (pSTS)) in 108 healthy subjects using functional magnetic resonance imaging (fMRI). We were particularly interested in the variability of hemispheric lateralization across subjects and explored how many subjects can be classified as right-dominant based on the fMRI activation pattern. We further assessed lateralization differences between different regions of the core system and analyzed the influence of handedness and sex on the lateralization with a generalized mixed effects regression model. As expected, brain activity was on average stronger in right-hemispheric brain regions than in their left-hemispheric homologues. This asymmetry was, however, only weakly pronounced in comparison to other lateralized brain functions (such as language and spatial attention) and strongly varied between individuals. Only half of the subjects in the present study could be classified as right-hemispheric dominant. Additionally, we did not detect significant lateralization differences between core system regions. Our data did also not support a general leftward shift of hemispheric lateralization in left-handers. Only the interaction of handedness and sex in the FFA revealed that specifically left-handed men were significantly more left-lateralized compared to right-handed males. In essence, our fMRI data did not support a clear right-hemispheric dominance of the face perception network. Our findings thus ultimately question the dogma that the face perception network - as measured with fMRI - can be characterized as "typically right lateralized".

neuroscience↗

bioRxiv2020

Thinking ahead: prediction in context as a keystone of language in humans and machines

Departing from traditional linguistic models, advances in deep learning have resulted in a new type of predictive (autoregressive) deep language models (DLMs). Using a self-supervised next-word prediction task, these models are trained to generate appropriate linguistic responses in a given context. We provide empirical evidence that the human brain and autoregressive DLMs share three fundamental computational principles as they process natural language: 1) both are engaged in continuous next-word prediction before word-onset; 2) both match their pre-onset predictions to the incoming word to calculate post-onset surprise (i.e., prediction error signals); 3) both represent words as a function of the previous context. In support of these three principles, our findings indicate that: a) the neural activity before word-onset contains context-dependent predictive information about forthcoming words, even hundreds of milliseconds before the words are perceived; b) the neural activity after word-onset reflects the surprise level and prediction error; and c) autoregressive DLM contextual embeddings capture the neural representation of context-specific word meaning better than arbitrary or static semantic embeddings. Together, our findings suggest that autoregressive DLMs provide a novel and biologically feasible computational framework for studying the neural basis of language.

neuroscience↗