Biology subjects

Scott, R. T.

Publications and source records attributed to Scott, R. T..

6 recordsLinked to original sources

bioRxiv2026

ABEL: an active-learning behavior estimation and labeling platform

Detailed behavior analysis is essential for thorough characterization of ethologically relevant behaviors in model organisms, yet manual annotation of the full behavioral repertoire remains subjective, time intensive, and susceptible to observer error. Advances in machine learning have enabled high-throughput pose estimation on recorded video, but tools for behavior classification from pose and video data are still developing. Instead of hand-scoring every frame of video, experimenters can instead label a small subset of video frames and software trained through machine learning makes predictions on the rest. Here, we present an Active-learning Behavior Estimation and Labeling (ABEL) platform that uses clip-level active learning (i.e., human labeling of short video snippets) with multimodal features (pose, video, context/ROI) to train robust behavior classifiers. We rigorously validated ABEL-derived behavior predictions against expert human observers and field-standard automated software, across diverse rodent behavioral assays. Across eight assays and 45 behaviors, model training required 19.5 hours of human annotation in total, with the reviewer scoring ~8% of available video. Models trained in ABEL achieved a mean precision-recall area under the curve (PR-AUC - a 0-1 score of how well a model balances missed detections against false alarms, with 1 being perfect) of 0.90 (SD 0.09, range 0.60-0.99), with no association between performance and behavior prevalence (r = 0.20). This was aided by custom tools, Essence Extractor and UMAP Interactive Selection, for targeted discovery of high probability clips which reduce the clip review needed to find a rare behavior 6-fold relative to random sampling and 10-fold relative to labeling whole videos. As a biological validation, we assessed how ABEL-derived behaviors relate to underlying neuronal calcium dynamics. Behavior labels were tightly synced with neuronal signatures distinct from ambiguous behavior and randomly chosen, behavior-unrelated time windows (shuffle control). Together, these data indicate that ABEL provides an efficient platform for frame-precise classification of distinct ethologically relevant behaviors.

neuroscience↗

bioRxiv2026

Automated behavioral segmentation and markerless pose tracking of mice during spaceflight

The NASA Rodent Habitat aboard the International Space Station enabled long-duration studies of behavioral responses to spaceflight, but video-based behavioral analysis has relied on laborious manual annotation. No study has tested whether deep learning tools can automate this analysis under the demanding imaging conditions of orbital vivaria. We applied pose estimation (SLEAP) and behavioral segmentation (DeepEthogram) to archival footage from the Rodent Research-1 mission. Nine labelers annotated 3,249 pose labels across 2,063 frames, and three behaviorists labeled 411,194 frames across 66 videos. Pose tracking accuracy approximated human inter-annotator variability despite progressive lens soiling, grid occlusions, and spherical aberration. Behavioral classification across eight categories achieved accuracy of 0.86-0.90 and suggests progressive behavioral adaptations to microgravity. Kinematic reconstruction of circling estimated centripetal accelerations periodically approaching 1g. This is the first application of deep learning-based pose estimation and behavioral segmentation to rodents in spaceflight, establishing benchmarks for future monitoring systems. One-Sentence SummaryDeep learning-based pose estimation and behavioral segmentation of ISS video reveal progressive behavioral adaptations of mice to microgravity

neuroscience↗

bioRxiv2026

Cross-Species Adaptation of RETFound for Rodent OCT Age Estimation Reveals Strong CNN Baselines in Data-Scarce Space Biology

Space-biology imaging studies are often constrained by severe data scarcity, limiting the development of robust machine-learning biomarkers. Rodent spaceflight and space-analog datasets provide an important preclinical setting for testing transfer-learning strategies, but the extent to which human retinal foundation models can generalize to rodent optical coherence tomography (OCT) remains unclear. Here, we benchmark cross-species adaptation of RETFound, a human retinal Vision Transformer pretrained on 1.6 million retinal images, for chronological age prediction from Brown Norway rat OCT B-scans in the NASA Open Science Data Repository dataset OSD-679. We adapted RETFound using Low-Rank Adaptation (LoRA) and evaluated performance on control animals under matched 3-fold rat-level cross-validation. We compared RETFound+LoRA with a strong ImageNet-pretrained Xception baseline under matched protocols and included a scratch/random ViT as a supplementary negative-control architecture check. Metrics included mean absolute error (MAE), R2, and inter-eye mean absolute difference (MAD). RETFound+LoRA achieved MAE = 26.20 {+/-} 5.03 days with R2 = 0.744 {+/-} 0.049. However, Xception performed better in the primary benchmark (MAE = 19.01 {+/-} 7.67 days, R2 = 0.853 {+/-} 0.082), and the matched-fold comparison favored Xception, although this result should be interpreted cautiously given the small number of folds. Inter-eye consistency was maintained across the matched control evaluation, and saliency maps localized model attention to anatomically plausible inner retinal regions. Together, these results show that human retinal foundation models can transfer to rodent OCT in a scientifically useful way, but also that strong CNN baselines may outperform transformer-based models in small-sample cross-species settings. This preprint provides a reproducible benchmark and baseline framework for future retinal biomarker development in space biology. Significance StatementSpace-biology imaging studies are intrinsically data-limited. This preprint provides a reproducible cross-species benchmark for adapting Earth-trained retinal models to rodent OCT under small-sample conditions, highlights the value of strong CNN baselines, and offers a reusable starting point for future retinal biomarker development in space-relevant datasets.

bioinformatics↗

bioRxiv2026

Machine Learning Ensemble Reveals Distinct Molecular Pathways of Retinal Damage in Spaceflown Mice

BackgroundSpaceflight-associated neuro-ocular syndrome (SANS) poses significant risks to astronaut visual health during long-duration missions, yet its underlying molecular mechanisms remain incompletely understood. Oxidative stress and apoptosis are candidate molecular drivers, but their transcriptomic signatures in spaceflight-exposed retinal tissue have not been systematically characterized. MethodsWe applied a machine learning ensemble of linear regression models to predict two ocular phenotypes: 4-hydroxynonenal (4-HNE) immunostaining as a marker of lipid peroxidation-mediated oxidative damage; and TUNEL positivity as a marker of apoptotic cell death. In this observational study, we use bulk retinal gene expression data obtained from a controlled experiment with ground control and spaceflown mice to predict these phenotypes. Gene Ontology pathway enrichment was performed on the most predictive gene sets for each phenotype. ResultsThe 4-HNE phenotype was predicted by genes that converge on membrane-associated pathways, photoreceptor protein modification, synaptic dysfunction, and extracellular matrix dysregulation, including B2m, Tf, Cnga1, mt-Nd1, Snap25, and Efemp1. The genes predicting the TUNEL phenotype revealed a distinct signature emphasizing stress-induced apoptosis, rod photoreceptor degeneration, and endoplasmic reticulum dysfunction, with Ddit4, Nrl, Rom1, Reep6, and Gabarapl1 emerging as central regulators. ConclusionsOxidative lipid peroxidation and apoptotic cell death represent complementary and molecularly distinct pathological mechanisms in spaceflight-exposed murine retinal tissue. The gene signatures provide a putative molecular framework for developing noninvasive biomarkers and therapeutic targets to monitor and protect astronaut visual health during long-duration and deep-space missions.

bioinformatics↗

bioRxiv2025

Foundational Architecture Enabling Federated Learning for Training Space Biomedical Machine Learning Models between the International Space Station and Earth

The public and commercial space industries are planning longer duration and more distant space missions, including the establishment of a habitable lunar base and crewed missions to Mars. To support Earth-independent scientific and medical operations, such missions can leverage artificial intelligence and machine learning models to assist with crew healthcare, spacecraft maintenance, and other critical tasks. However, transferring large volumes of data between Earth and space for model development consumes valuable bandwidth, is vulnerable to communication disruptions, and may compromise crew safety and data privacy. Federated learning enables model training while keeping data in situ and only transferring model parameters. In this work, we present a flexible, resilient federated learning framework that provides the secure transmission of model updates between Earth and the International Space Station. On March 15, 2024, this framework pioneered the deployment of federated learning in a spaceflight setting, training classifier models between Earth and the ISS using both real biomedical research data and synthetically generated data.

bioinformatics↗

bioRxiv2020

Mechanism of cell polarisation and first lineage segregation in the human embryo

The formation of differential cell lineages in the mammalian blastocyst from the totipotent zygote is crucial for implantation and the success of the whole pregnancy. The first lineage segregation generates the polarised trophectoderm (TE) tissue, which forms the placenta, and the apolar inner cell mass (ICM), which mainly gives rise to all foetal tissues and also the yolk sac1-3. The mechanism underlying this cell fate segregation has been extensively studied in the mouse embryo4,5. However, when and how it takes place in the human embryo remains unclear. Here, using time-lapse imaging and 325 surplus human embryos, we provide a detailed characterisation of morphological events and transcription factor expression and localisation to understand how they lead to the first lineage segregation in human embryogenesis. We show that the first lineage segregation of the human embryo is triggered by cell polarisation that occurs at the 8-cell stage in two sequential steps. In the first step, F-actin becomes apically polarised concomitantly with embryo compaction. In the second step, the Par complex becomes polarised to form the apical cellular domain. Mechanistically, we show that activation of Phospholipase C (PLC) triggers actin polarisation and is therefore essential for apical domain formation, as is the case in mouse embryos6. Finally, we show that, in contrast to the mouse embryo, the key extra-embryonic determinant GATA37,8 is expressed not only in extra-embryonic lineage precursors upon blastocyst formation. However, the cell polarity machinery enhances the expression and nuclear accumulation of GATA3. In summary, our results demonstrate for the first time that cell polarisation reinforces the first lineage segregation in the human embryo.

developmental biology↗