Most existing deep learning-based cell tracking methods rely on supervised learning, requiring large-scale annotated datasets that are often unavailable in real-world scenarios. Moreover, many approaches lack tools and methods for correcting mispredicted links or incorporating corrections through fine-tuning. These limitations contribute to the limited adoption of deep learning-based tracking methods in the life sciences, where manual tracking remains the predominant approach. To reduce the annotation burden and enable model training without extensive labeled data, we introduce a loss function for unsupervised training. Our method leverages the predictable dynamics inherent in many biological processes, providing an initialization that does not require an annotated dataset. We further investigate how minimal user-provided annotations can refine tracking accuracy. To this end, we propose an active learning framework that selectively identifies uncertain decisions within the tracking graph, allowing for efficient annotation of the most informative data points. We evaluate our approach on two microscopy datasets, demonstrating the effectiveness of both our unsupervised training strategy and active learning scheme in improving tracking performance. Our implementation and reproducible experiments are available at github.com/funkelab/attrackt and github.com/funkelab/attrackt_experiments, respectively.

Expansion microscopy (ExM) enables nanoscale imaging on standard microscopes, but combining ExM with single-molecule localization microscopy (SMLM) remains difficult, owing to the incompatibility of expanded hydrogels with photoswitching buffers. Here, we introduce a single-step expansion microscopy method that allows SMLM with spontaneously blinking dyes in 6-14× expanded samples, without re-embedding. We demonstrate nanometer-resolution imaging by resolving the organization of the nuclear pore complex (NPC) and the molecular structure of recombinant homotrimeric proliferating cell nuclear antigen (PCNA).

Many fungi utilize high turgor pressure for morphogenesis, requiring tight regulation of ionic gradients. Ion regulation is important for pathogenesis, reproduction, and general homeostasis across the fungal kingdom. In the major human fungal pathogen Candida albicans, potassium (K+) channels fine-tune ionic balance under stressful environmental conditions, contributing to colonization of the human host. Two-pore domain, outwardly rectifying potassium (TOK) channels, uniquely found in fungi, remain insufficiently characterized despite early evidence implicating them in diverse intracellular processes essential for cellular growth and viability, and their potential as antifungal targets. Here, we describe the first atomic resolution structure of a fungal potassium channel—TOK1 from C. albicans (CaTOK)—revealing a membrane topology distinct from all other known K+ channel classes. We propose that CaTOK1 utilizes two unique regions—TOK auxiliary subunit-like channel (TALC) domain and a structured c-terminal bundle—to regulate TOK1 gating. Conformational analysis of TOK1 pore features an inner helical gating mechanism with “up” and “down” conformations similar to mammalian dimeric K+ channels. These findings provide a structural framework for understanding TOK channel activity and lay the groundwork for future studies on fungal ion homeostasis, pathogenicity, and therapeutic development.

To form a blood clot, fibrinogen is converted into fibrin through the action of the enzyme thrombin. Fibrin then polymerizes longitudinally and laterally as it matures into a fiber. Polymerization results in a dense, 3-dimensional branched network. Previous research has shown the relevance of these fibrin gel structures in hemostatic conditions; however, the mechanism by which they form has not been fully resolved. Using light sheet microscopy, 3-dimensional volumes of the fibrin polymerization process were captured. Manual annotation of these microscopy videos revealed that fiber branch points occur through the collision and the binding of diffusing fibers rather than through the splitting of growing fiber termini. However, the density of fibers and amount of data greatly slows manual annotation-based analysis and limits the ability to capture important data, such as growth rates and fiber stiffness. To more quickly process these data, a computational approach was utilized. A custom tracking pipeline, suited to the networks formed by cylindrical fibrin fibers, was developed, beginning with an AI-based classifier. This custom pipeline allowed for the tracking of uniquely labeled fibers over time. Automated merge detection between linking phases further improved accuracy. Additionally, network formation was analyzed through skeletonization techniques to measure the number of branches per junction over time. Combining the skeletonization and tracking methods, single fibers were identified by their lack of branch points and tracked. The addition of branch points to previously tracked objects served as a signal for merge detection. This approach yielded measurements of single fibrin fiber diffusion rates, as well as the first volumetric and length growth rates of fibers throughout polymerization. In addition, the gel point was quantified by analyzing the span of connected objects to characterize the network consolidation over time at the level of single fibers.

Local field potentials (LFPs) reflect coordination among neural populations, yet their exact relationship to neural computation remains unknown. One exception is the theta rhythm of the rodent hippocampus, which organizes sequential firing among place cells, enabling spike timing to track the animal's path through its environment. But when the animal stops, the theta rhythm becomes irregular, which is assumed to disrupt its ability to carry spatial information. Here we challenge this assumption by developing an artificial neural network that discovers position-tuned theta rhythms (pThetas) from LFPs even in the absence of strong theta oscillations. Using recordings from male rats, we provide evidence that pTheta is distinct from the dominant theta rhythm, while reflecting rhythmic coordination among place cell populations. Our work suggests that weak and intermittent oscillations, as seen in many brain regions and species, can convey information commensurate with population spike codes when decoded using information-based rather than variance-based principles.

Multiplexed protein imaging enables spatially resolved analysis of molecular organization in tissues, but existing spatial proteomics platforms remain constrained in scalability, throughput, and integration with RNA measurements and interpretable computational analysis. Here, we present an integrated spatial omics framework that combines highly multiplexed protein and RNA imaging with explainable machine learning to map cell-type-specific molecular and structural architectures at tissue scale. Using this platform, we simultaneously profiled up to 46 proteins and 79 RNA species across \~370,000 cells in intact mouse brain tissue at diffraction-limited subcellular resolution (\~260 nm). We developed a scalable, open-source computational pipeline for large-scale image processing and analysis, and show that nuclear protein and chromatin features alone are sufficient to accurately classify brain cell types and their spatial organization. Incorporation of explainable deep learning further enabled identification of human-interpretable, cell-type-specific subnuclear structural features directly from imaging data, with independent quantitative validation. Together, this integrated experimental and computational framework enables tissue-scale spatial proteomics-based cell-type classification and structural feature discovery, providing a broadly applicable platform for mechanistic studies, high-content screening, and translational applications.

Coordinated lateralized movements are critical for natural orienting behaviors, but their neural bases remain poorly understood. The deep superior colliculus (dSC) integrates a wide range of inputs to select targets for orienting movements and coordinates downstream activity to initiate and execute movement. The substantia nigra pars reticulata (SNr) is thought to disinhibit dSC to facilitate movement, but much remains unknown about the relationship between SNr activity, dSC activity, and movement. We recorded from both regions using high-density probes in head-fixed mice performing directional orienting tasks. We found that dSC and SNr activity reflected task variables preceding and throughout movement. However, the direction-dependence of dSC activity was weaker than in other orienting behaviors, and the relationship between movement-related dSC and SNr activity was inconsistent with disinhibition of dSC determining the initiation or direction of movement. Analyses of similar data curated by the International Brain Laboratory yielded consistent results. These findings suggest diverse roles for modulatory input from SNr to dSC in shaping motor behavior.

Many animals possess mechanosensory neurons that fire when a limb nears the limit of its physical range, but the function of these proprioceptive limit detectors remains poorly understood. Here, we investigate a class of proprioceptors on the Drosophila leg called hair plates. Using calcium imaging in behaving flies, we find that a hair plate on the fly coxa (CxHP8) detects the limits of anterior leg movement. By reconstructing CxHP8 axons in an electron microscopy dataset, we found that they are wired to excite posterior leg movement and inhibit anterior leg movement. Consistent with this connectivity, optogenetic activation of CxHP8 neurons elicited posterior postural reflexes, while silencing altered the swing-to-stance transition during walking. Finally, we use comprehensive reconstruction of peripheral morphology and downstream connectivity to predict the function of other hair plates distributed across the fly leg. Our results suggest that each hair plate is specialized to control specific sensorimotor reflexes that are matched to the joint limit it detects. They also illustrate the feasibility of predicting sensorimotor reflexes from a connectome with identified proprioceptive inputs and motor outputs.

From insects to mammals, essential brain functions, such as forming long-term memories (LTMs), increase metabolic activity in stimulated neurons to meet the energetic demand associated with brain activation. However, while impairing neuronal metabolism limits brain performance, whether expanding the metabolic capacity of neurons boosts brain function remains poorly understood. Here, we show that LTM formation of flies and mice can be enhanced by increasing mitochondrial metabolism in central memory circuits. By knocking down the mitochondrial Ca exporter Letm1, we favour Ca retention in the mitochondrial matrix of neurons due to reduction of mitochondrial H/Ca exchange. The resulting increase in mitochondrial Ca over-activates mitochondrial metabolism in neurons of central memory circuits, leading to improved LTM storage in training paradigms in which wild-type counterparts of both species fail to remember. Our findings unveil an evolutionarily conserved mechanism that controls mitochondrial metabolism in neurons and indicate its involvement in shaping higher brain functions, such as LTM.

The brain’s capabilities rely on both the molecular properties of individual cells and their interactions across brain-wide networks. However, relating gene expression to activity in individual neurons across the entire brain remains elusive. Here we developed an experimental-computational platform, WARP, for whole-brain imaging of neuronal activity during behavior, expansion-assisted spatial transcriptomics, and cellular-level registration of these two modalities. Through joint analysis of whole-brain neuronal activity during multiple behaviors, cellular gene expression, and anatomy, we identified functions of molecularly defined populations — including luminance coding in a cckb-pou4f2 midbrain population and task-structured activity in pvalb7-eomesa hippocampal-like neurons — and defined over 2,000 other function-gene-anatomy subpopulations. Analysis of this unprecedented multimodal dataset also revealed that most gene-matched neurons showed stronger activity correlations, highlighting a brain-wide role for gene expression in functional organization. WARP establishes a foundational platform and open-access dataset for cross-experiment discovery, high-throughput function-to-gene mapping, unification of cell biology and systems neuroscience, and scalable circuit modeling at the whole-brain scale.