bioRxiv Subject Collection: Neuroscience's Journal

Advancing Fair and Explainable Machine Learning for Neuroimaging Dementia Pattern Classification in Multi-Ethnic Populations
Dementia, a degenerative disease affecting millions globally, is projected to triple by 2050. Early and precise diagnosis is essential for effective treatment and improved quality of life. However, current diagnostic approaches frequently demonstrate inconsistent precision and impartiality, particularly among diverse cultural groups. This study investigates performance discrepancies in dementia classification among White American, African American, and Hispanic populations. We reveal significant cross-group bias, particularly when models trained on one group are tested on another. To address this, we introduce a novel combination of few-shot learning and domain alignment to improve model adaptability across underrepresented populations. Our results show that these techniques substantially reduce inter-group performance gaps, especially between White American and Hispanic cohorts. This finding highlights the crucial need for fairness-aware strategies and the inclusion of diverse populations in training data to ensure accurate and equitable dementia diagnoses.

Validation of a combined cylindrical shield and partial-coverage mobile OPM system for detecting neuromagnetic sensorimotor responses in humans
Optically pumped magnetometers (OPMs) have emerged as a promising technology for neuromagnetic recording in humans. Current state-of-the-art OPM systems are housed in immobile magnetically-shielded rooms to reduce external electromagnetic noise, and typically comprise sensor arrays covering the entire head. Here we sought to validate a low-cost, mobile OPM system comprising a small cylindrical mu-metal shield and partial sensor coverage. Twelve participants underwent right-sided median nerve stimulation (MNS) and cued right-handed button-pressing intended to elicit ubiquitous sensorimotor responses: somatosensory-evoked fields (SEFs; comprising N20m, P35m and P60m components) and event-related (de)synchronisation (ERD/ERS) of oscillatory neuronal rhythms in the mu and beta frequency ranges. Following MNS, we observed robust N20m and P60m peaks, as well as the expected mu ERD and beta ERS effects. Moreover, we successfully localized these responses to expected cortical generators using distributed source modelling. SEFs and mu ERD were both maximal in left (i.e., contralateral to stimulation) primary somatosensory cortex (central sulcus, postcentral gyrus and sulcus), while beta ERS appeared more anteriorly, in the central sulcus and precentral gyrus. By contrast, results from the button-pressing paradigm were less conclusive--we observed beta ERS (but not mu/beta ERD), and an atypical distribution of the ERS effect over posterior ROIs. Overall, our findings provide proof-of-principle support for the use of our system in the context of passive (e.g., MNS) paradigms; its viability for cued movement tasks will require further development. Based on these results, we make recommendations for further developments in mobile and partial-coverage OPM.

Shaping a Collaborative, Sustainable, Accessible, and Reproducible Future for Computational Modeling
The o2S2PARC platform is an open-source, extensible, and scalable cloud-based platform developed in the context of the U.S. National Institutes of Health SPARC program to support collaborative, sustainable, FAIR (findable, accessible, interoperable, reusable) and reproducible computational modeling and analysis. This publication presents the main features of o2S2PARC, its underlying approaches and philosophy, innovative aspects of the developed technologies, while also drawing attention to its rapid adoption. The paper showcases a variety of applications and use cases enabled by the platform. These include hybrid electromagnetic-electrophysiology simulations of neural interfaces, personalized brain and spinal cord stimulation planning, in silico device safety assessments, the training and application of AI systems (e.g., for model-predictive control and medical image segmentation), hybridized surrogate modeling and multi-objective optimization in high-dimensional parameter spaces, sensitive and unbiased validation of measurement devices, and interactive data analysis as paper supplements.

Novel plasmalogen derivative KIT-13 restores neurological function in a mouse model of Rett syndrome by reducing neuroinflammation and restoring mitochondrial function
Neurodevelopmental disorders, including Rett Syndrome (RTT), have no functional cure and cause substantial levels of disability. Neuroinflammation is now strongly associated with both neurodevelopmental disorders and neurodegenerative diseases, providing a strong rationale for development of novel therapeutics targeting common neuroinflammatory mechanisms. RTT is caused by mutations in the methylated DNA binding factor MECP2. Mecp2-deficient (Mecp2-KO) mice, which have been extensively characterized as a mouse model of RTT, exhibit high levels of neuroinflammation, mitochondrial dysfunction, and severe neurological symptoms similar to RTT patients. KIT-13 is a novel plasmalogen derivative being developed for the treatment of neurodevelopmental disorders including RTT. This study evaluated KIT-13 in both cell-based and in vivo models for its potential to inhibit neuroinflammation and address underlying mitochondrial dysfunction, as well as effects on RTT-like neurological symptoms in the RTT mouse model. Oral administration of KIT-13 to Mecp2-KO mice significantly reduced neurological symptoms assessed by a composite score evaluating mobility, gait, hindlimb clasping, tremor, breathing, and general condition and improved the life span of the RTT model mice. In addition, KIT-13 suppressed mitochondrial DNA leakage associated with Mecp2 deficiency, and significantly suppressed neuroinflammation as measured by microglial cell morphology. These results suggest that KIT-13 may be a promising therapeutic agent for RTT and other neuroinflammation-related diseases.

Estimating individual-level changes in functional brain connectivity and correspondence with topology changes
With the rise of precision medicine and large neuroimaging datasets, measuring brain changes on an individual level becomes more possible and more important. However, functional connectivity has some mathematical and conceptual quirks that make estimating individual-level changes more complicated than, for example, estimating structural brain changes. Here, we compared six different change scores, i.e., metrics for quantifying change in functional connectivity, using a large sample with two time points, roughly 2 years apart, of low-motion data from the ABCD Study (N=2,719, ages 9-13 years). First, we compared distributions and potential interpretations of each metric. Then, we assessed how well different metrics captured topology change estimates by comparing them to individual-level changes in clustering coefficient, betweenness centrality, and network strength. As multivariate metrics may be more reliable than single connections and researchers often interpret connectivity changes in topological terms, this offers additional insight into the implications of change score choice. Overall, our analyses revealed widely different distributions between change scores that conferred vastly different results and interpretations of functional connectivity changes between time points. Thus, we provide recommendations for each change score and its optimal (or suboptimal) use cases, depending on the population, study design, and research question.

Species-specific oculomotor control tolerances predict saccadic suppression strength in macaques and zebrafish
Active sensing necessarily requires integrating information about both self-generated movements as well as past, present, and future afferent inputs1,2. However, such information is inherently variable and relies, at least in part, on uncertain extrapolations. Here, using saccadic suppression3,4 of visual sensitivity as a classic example of sensory-motor integration, we show that suppression strength in two drastically different species, macaque monkeys and zebrafish larvae, may be a direct outcome of efficient state estimation in the presence of uncontrollable sensory and motor variability. Bayesian estimator models5 suggest that optimal saccadic suppression should rely not just on the sensory-motor information being processed, but also on the time-dependent magnitude of unexplained variability in the nervous system encoding it. In both macaques and zebrafish larvae, and using matched visual stimulation regimes across the species, we experimentally measured saccadic suppression strength in the superior colliculus (SC) of the monkeys and the homologous optic tectum (OT) of the fish. We also experimentally quantified additive and multiplicative noise components in the motor systems of both species, and we furthermore estimated noise in the sensory systems. We found that inter-species differences in noise levels and their theoretically predicted impacts on visual sensitivity are qualitatively consistent with our experimentally observed differences in saccadic suppression strength between the fish and the monkeys. Because sensory and motor noise levels can reflect the amounts of available neural resources committed to a given task, our results strongly underscore the value of incorporating computational resource limits in investigating performance differences that have evolved in homologous brain areas.

Comparative Analysis of Test Tube and Volumetric Drinking Monitor Methods in Voluntary Ethanol Consumption in Female Mice for Prenatal Alcohol Exposure
ObjectiveFetal alcohol spectrum disorders affect approximately 1 in 20 school age children in the United States of America. To study fetal alcohol spectrum disorders, mouse models are commonly used. Of the many approaches of gestational exposure, voluntary drinking paradigms represent the most similar mechanism of drinking as human exposure. These exposures can be done through low-tech solutions such as test tubes (TT), or more high-tech methods such as a volumetric drinking monitor (VDM). Here were compare the TT method and the VDM directly, to evaluate their effect on female mouse drinking.

MethodWe adapted a drinking in the dark, active cycle, limited access (4 hr.) voluntary drinking paradigm first described by Brady et al. (2012) to test tubes and the volumetric drinking monitor. 8 mice were placed in either drinking method and we evaluated their drinking volume and blood alcohol concentrations (BACs). We compared the values for each group using t-tests.

ResultsAfter 2 weeks of drinking 10% ethanol with 0.4% saccharine, BACs were not significantly different [t(14)=0.2681, p=0.7935] between the VDM (81.56 {+/-} 21.16 mg/dL) vs.TT (73.14 {+/-} 23.20 mg/dL) groups. Calculated intake of ethanol (g/kg) on the day of blood draw for BAC analysis was also not significantly different [t(14)=0.4308, p=0.6732] between VDM (2.985 {+/-} 0.4127) vs.TT (3.260 {+/-} 0.4863; Fig 1B) groups.

O_FIG O_LINKSMALLFIG WIDTH=139 HEIGHT=200 SRC="FIGDIR/small/658718v1_fig1.gif" ALT="Figure 1">
View larger version (15K):
org.highwire.dtl.DTLVardef@1347429org.highwire.dtl.DTLVardef@b666bborg.highwire.dtl.DTLVardef@112a02eorg.highwire.dtl.DTLVardef@1d22fa2_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig 1.C_FLOATNO C_FIG ConclusionsTest tube or VDM resulted in similar average daily ethanol consumption and resultant BACs in female mice

Transcranial ultrasound stimulation modulates neuronal membrane potentials across broad timescales in the awake mammalian brain
BackgroundTranscranial ultrasound stimulation (TUS) offers noninvasive neuromodulation with high spatial and temporal precision, but its cellular-level effects in the awake brain remain poorly understood.

ObjectiveWe investigated how low-intensity TUS modulates membrane voltage dynamics in single cortical neurons in awake mice.

MethodsUsing the genetically encoded voltage indicator SomArchon, we performed high-speed kilohertz voltage imaging in awake head-fixed mice. TUS was delivered with a 0.35 MHz transducer at 10 or 40 Hz pulse repetition frequency with a 20% duty cycle, at intensities below the estimated threshold for auditory brainstem activation. We analyzed changes in membrane potential (Vm), spiking, and coordination across simultaneously recorded neurons.

ResultsTUS evoked rapid (<10 ms) Vm depolarizations in 42.8% of neurons, while only 20.5% showed increased spiking, highlighting a direct effect of TUS on modulating synaptic inputs. Many neurons were entrained at both PRFs (20.8% at 10 Hz; 12.7% at 40 Hz) with Vm exhibiting significant phase-locking to individual TUS pulses. Vm entrainment was accompanied by increased temporal coordination across neurons and reset network synchrony. Furthermore, TUS-evoked cellular responses adapted over time, often transitioning from membrane depolarization to hyperpolarization upon repeated exposures, demonstrating prominent response depression.

ConclusionBy resolving single-neuron responses, our results demonstrate that TUS directly activates individual cortical neuron with a latency shorter than 10 ms. TUS pulsed at physiologically relevant frequencies of 10 and 40 Hz robustly entrains neural dynamics, alters network coordination and evokes neuronal plasticity. These results highlight the therapeutic potential of designing TUS pulsing patterns to target specific neural circuit dynamics.

GABA-independent activation of GABAB receptor by mechanical forces
The heterodimeric GABAB receptor, composed of GB1 and GB2 subunits, is a metabotropic G protein-coupled receptor (GPCR) activated by the neurotransmitter GABA. GABA binds to the extracellular domain of GB1 to activate G proteins through GB2. Here we show that GABAB receptors can be activated by mechanical forces, such as traction force and shear stress, in a GABA-independent manner. This GABA-independent mechano-activation of GABAB receptor is mediated by a direct interaction between integrins and the extracellular domain of GB1, indicating that GABAB receptor and integrin form a novel type of mechano-transduction complex. Mechanistically, shear stress promotes the binding of integrin to GB1 and induces an allosteric re-arrangement of GABAB receptor transmembrane domains towards an active conformation, culminating in receptor activation. Furthermore, we demonstrate that shear stress-induced GABAB receptor activation plays a crucial role in astrocyte remodeling. These findings reveal a previously unrecognized function of GABAB receptor in mechano-transduction, uncovering a novel ligand-independent activation mechanism for GPCRs.

Treatment of age-related decreases in GTP levels restores endocytosis and autophagy
Age-related declines in neuronal bioenergetic levels may limit vesicular trafficking and autophagic clearance of damaged organelles and proteins. Age-related ATP depletion would impact cognition dependent on ionic homeostasis, but limits on proteostasis powered by GTP are less clear. We used neurons isolated from aged 3xTg-AD Alzheimers model mice and a novel genetically encoded fluorescent GTP sensor (GEVAL) to evaluate live GTP levels in situ. We report an age-dependent reduction in ratiometric measurements of free/bound GTP levels in living hippocampal neurons. Free-GTP co-localized in the mitochondria decreased with age accompanied by the accumulation of free-GTP labeled vesicular structures. The energy dependence of autophagy was demonstrated by depletion of GTP with rapamycin stimulation, while bafilomycin inhibition of autophagy raised GTP levels. 24 hr. supplementation of aged neurons with the NAD precursor nicotinamide and the Nrf2 redox modulator EGCG restored GTP levels to youthful levels and mobilized endocytosis and lysosomal consumption for autophagy via the respective GTPases Rab7 and Arl8b. This vesicular mobilization promoted the clearance of intraneuronal A{beta} aggregates and lowered protein oxidative nitration in AD model neurons. Our results reveal age- and AD-related neuronal GTP energy deficits that impair autophagy and endocytosis. GTP deficits were remediated by an external NAD precursor together with a Nrf2 redox modulator which suggests a translational path.

Interactions across hemispheres in prefrontal cortex reflect global cognitive processing
Brain functions involve processing in local networks as well as modulation from brainwide signals, such as arousal. Dissecting the contributions of populations of neurons to these functions requires knowledge of interactions between brain areas. We investigated these interactions using dual hemisphere recordings of prefrontal cortex in monkeys performing a spatial memory task. To tease apart global processing from local interactions, we applied a novel statistical approach called pCCA-FA (a combination of probabilistic canonical correlation analysis and factor analysis) to analyze trial-to-trial variability in neuronal responses. We found substantial shared variability among neurons within each population, much of which was actually shared across populations and linked to an arousal process. Our work presents a path by which we can leverage multi-area recordings to reveal aspects of brain functions that are hidden in single-area recordings.

Auditory Gamma-Frequency Entrainment Abolishes Working Memory Deficits in a Rodent Model of Autism
Individuals on the autism spectrum often exhibit atypical responses to sensory stimuli and difficulties with behavioral regulation, reflecting altered functional patterns in core sensory-motor circuits. These sensory processing differences suggest the potential for using patterned auditory stimulation to modulate neural activity and mitigate symptoms. While impairments in higher cognitive functions are a hallmark of autism, they remain underexplored in rodent models compared to basic sensory-motor deficits. In this study, we investigated higher cognitive function in the valproic acid (VPA) rodent model of autism and evaluated whether auditory entrainment could ameliorate associated impairments. Pregnant dams received an intraperitoneal injection of VPA on embryonic day 12.5 (E12.5), and offspring were subsequently tested on a delayed non-match to place (DNMP) task to assess working memory. Following 40Hz sensory entrainment, VPA-exposed animals showed significantly impaired performance and altered trial-by-trial learning dynamics, indicating deficits in working memory-based decision making. More importantly, auditory stimulation led to increased gamma-frequency oscillations and enhanced power in other bands (theta, beta) across brain regions, as assessed by LFP recordings in freely moving animals. These neural effects were sustained even after stimulation ended, indicating robust and lasting modulation of brain dynamics.

This provides strong evidence for the therapeutic potential of sensory entrainment in reversing autism-related cognitive deficits.

The Art of Not Knowing: Accommodating Structured Missingness in Biomedical Research
Missing data remain a ubiquitous and critical challenge in large-scale clinical studies. Despite advances in imputation, most existing methods fail to address structured missingness, where data are missing according a deterministic pattern and which arise due to systematic patterns introduced by experimental design, site protocols, or cohort differences. These patterns violate key assumptions of most imputation algorithms, yet their impact is rarely evaluated. We demonstrate that structured missingness is a fundamental challenge to drawing valid inferences from standard imputation techniques. First, we present a comprehensive framework for understanding and accommodating the effects of structured missingness. Next, we show through simulations and real-world psychometric data with structured missingness, that widely used algorithms optimised for numerical precision (e.g., Extra Trees, AutoComplete) underperform due to site effects, while donor-based methods (e.g., MICE, hierarchical MICE) better preserve multivariate structure. We propose a novel hierarchical approach that provides optimal performance in simulated and experimental data. Finally, we show that commonly used accuracy metrics, such as mean squared error can obscure these failures, and are therefore inadequate for the evaluation of structured missingness. In contrast, other divergence-based metrics offer a more sensitive and interpretable alternative. We apply this approach to harmonising psychometric data across cohorts, which provides excellent item-level alignment across different instruments. Our study highlights the need for a paradigm shift in handling missing data within biomedical research, moving beyond conventional imputation frameworks to develop tools that can account for structured missingness. This shift is essential for ensuring reliable inference in multi-site clinical studies, precision medicine, and large-scale population analyses.

Boolean Network Modeling Identifies Cognitive Resilience in the First Murine Model of Asymptomatic Alzheimer Disease
Alzheimer disease (AD) is a progressive neurodegenerative disorder defined by amyloid beta plaques and neurofibrillary tangles (NFTs), yet approximately 30% of aged individuals exhibit these hallmark lesions without developing cognitive impairment, a clinically silent condition termed asymptomatic AD (AsymAD). The molecular basis of this cognitive resilience remains poorly understood due to a lack of mechanistic models. Here, we integrate systems level Boolean network modeling with in vivo validation to define the transcriptomic logic of AsymAD and uncover a novel preclinical model. Using Boolean implication networks trained on large-scale human cortical RNA seq datasets, we identified a robust and invariant AD gene signature that accurately stratifies disease states across independent datasets. Application of this signature to Chromogranin A deficient PS19 mice (CgA-KO/PS19) revealed a unique resilience phenotype: male mice developed AD like molecular and neuropathological profiles in the pre-frontal cortex yet retained intact learning and memory. Female CgA-KO/PS19 mice displayed even greater protection, including reduced Tau phosphorylation and preserved synaptic ultrastructure. These findings establish the first validated murine model of AsymAD and identify CgA as a modifiable node linking neuroendocrine signaling, Tauopathy, and cognitive preservation. This work provides a scalable platform to probe sex-specific resilience, uncover early-stage biomarkers, and accelerate preventive therapeutic development in AD.