bioRxiv · systems biologyBuildable★ flagship
AI predicts how a cell rations its limited protein-making budget when forced to make foreign products.
Microbes are used as tiny factories to churn out useful proteins like drugs or enzymes, but a cell only has so much protein-building machinery to go around. When you force it to make a lot of a foreign product, it has to steal resources from its normal jobs, which slows everything down — this is the 'burden' problem. This work builds a tool called HyTT that predicts how the cell will reshuffle its resources from scratch, without needing expensive lab measurements for every new condition. It does this by blending a machine-learning method that learns patterns from a gene's sequence with a physics-style model of the cell's chemistry, and crucially it insists that ribosomes (the cell's protein-assembly machines) stay whole and properly counted. Getting these predictions right helps bioengineers design better production strains before running costly experiments.
Technical view
HyTT couples multivariate adaptive regression splines (MARS) — capturing sequence-derived translational costs — with an enzyme-constrained genome-scale metabolic model, enforcing an 80S ribosome integrity constraint and solving the whole thing as a mixed-integer linear program via bisection search. The novelty is ab initio prediction of proteome reallocation without condition-specific omics, linking macroscopic proteome-space bounds to microscopic per-transcript translation cost. Validation against steady-state chemostat quantitative proteomics showed it outperformed competing constraint-based approaches at predicting system-wide resource shifts. A practitioner could adapt the framework to other chassis organisms or heterologous constructs to forecast expression burden and guide strain design before wet-lab iteration.
bioRxiv · cell biologyBuildable★ flagship
Using flashes of light to switch on a cell's self-destruct button through its power plants.
Every cell has mitochondria, the little 'power plants' that generate its energy, and they also act as triggers for programmed cell death. In diseases like cancer you often want to kill specific cells on command, while in neurodegeneration you want to understand how cells wrongly die. This study engineers cells to carry light-sensitive proteins aimed at their mitochondria, so that shining light can sabotage the power plants in three different ways — making them too alkaline, draining their electrical charge, or flooding them with damaging reactive molecules. Because light can be aimed precisely in space and time, this offers a way to kill targeted cells with a level of control that drugs can't match. The hope is future therapies that flip cell death on or off exactly where needed.
Technical view
The authors deploy three optogenetic strategies to perturb mitochondria: Gloeobacter rhodopsin-mediated matrix alkalization to inhibit oxidative phosphorylation, reverse proton-pumping rhodopsins (RPPR) and anion-conducting channelrhodopsins to drive depolarization, and mitochondria-targeted miniSOG to photogenerate ROS. Each modality maps to a distinct death-relevant mechanism (bioenergetic collapse, membrane-potential loss, oxidative damage), giving a toolkit for spatiotemporally controlled induction of cell death. Practitioners could use these constructs to dissect apoptosis/necrosis pathways with light or to prototype targeted ablation; the reverse-proton-pumping and anion-channel approaches to mitochondrial depolarization are the more novel actuators worth replicating.
bioRxiv · ecologyRunnable★ flagship
Using 3D reef photos to map which coral clones are secretly copies of each other.
As reefs decline, tough weedy corals are taking over, and many of them spread by cloning themselves rather than sexual reproduction, forming dense 'carpets' of genetically identical colonies. That worries scientists because a stand of clones has little genetic diversity, making it fragile against disease or heat. The problem is these carpets blur together with no clear boundaries between individuals, so normal sampling can't tell where one genetic individual ends and another begins. The researchers photographed reefs to build detailed 3D maps, then used those maps to guide exactly where they took tissue samples, letting them see how clones are arranged in space. Studying the coral Madracis auretenra at three sites in Curaçao, they can now measure how much true genetic variety these increasingly dominant carpets actually hold.
Technical view
The study pairs Structure-from-Motion photogrammetry with spatially explicit, map-guided tissue sampling to resolve genet (genetic individual) identity and distribution in Madracis auretenra, a weedy scleractinian forming monospecific stands where colony boundaries are indistinct. Genotyping the georeferenced samples across three Curaçao reefs quantifies clonal versus sexual contribution and the spatial scale of asexual proliferation. The method overcomes the sampling-design failure inherent to boundaryless aggregations, yielding fine-scale genotypic diversity and clonal architecture estimates. Reef ecologists could replicate this photogrammetry-anchored sampling to assess intraspecific diversity and resilience in other clonal, spatially dominant taxa.
bioRxiv · ecologyRunnable★ flagship
Detecting a vanishing giant fish just from the DNA it sheds into estuary water.
The Chinese Bahaba is a huge, critically endangered fish so rare that going out to catch or even spot one for monitoring is nearly impossible. Scientists needed a way to track whether it's still present without disturbing it, so they turned to environmental DNA — the tiny traces of genetic material every animal sheds into the water around it. They designed molecular 'probes' that light up only when they encounter this specific species' DNA, targeting three distinct regions of its mitochondrial genome, and carefully checked that the probes don't accidentally react to closely related fish. After testing in the lab against real tissue and water samples, they picked the most sensitive probe for field use. This gives conservationists a cheap, non-invasive tool to find where this 'Panda of the Sea' still survives and to guide restocking efforts.
Technical view
Three TaqMan probe-based qPCR assays were designed from complete mitogenomes, targeting 12S rRNA, ND5, and the D-loop control region for species-specific detection of Bahaba taipingensis. Specificity was validated in silico against GenBank and in vitro against congeneric sciaenid tissue DNA plus positive eDNA samples, with sensitivity established via serial dilutions of synthetic target DNA. The 12S assay (Bhb_12S) was selected for field screening on superior low-copy detection and minimal non-specific by-products. The primer-probe sequences and validation pipeline are directly transferable to eDNA field surveys in the Pearl River Estuary and to designing analogous assays for other rare sciaenids.
bioRxiv · ecologyRunnable★ flagship
A crowd-built sound library of India's wildlife to teach AI to recognize species by ear.
One powerful way to track wildlife across huge areas is to put out microphones and let AI identify animals from their calls — but the AI can only learn a species if it has enough labeled example recordings, and for most tropical wildlife those examples simply don't exist. India is bursting with biodiversity yet lacks these training templates, so automatic recognition fails there. To fix this, the team recruited a crowd of researchers, conservationists, and nature lovers to listen to recordings and mark exactly when and which species vocalized, building a free open dataset spanning many animal groups. It contains thousands of minutes of precisely labeled calls plus more loosely labeled audio. This shared resource lets others train and benchmark deep-learning models for Indian fauna, filling a critical gap in global biodiversity monitoring.
Technical view
The dataset comprises 3311 minutes of strongly labelled acoustic data (species-level bounding-box annotations) and 2504 minutes of weakly labelled data across multiple taxonomic groups, assembled via a participatory crowd-sourcing pipeline. It directly addresses the training-template scarcity that blocks deep-learning bioacoustic classifiers in undersampled tropical regions. Being open-access and strongly labelled, it can serve as training and evaluation data for CNN/transformer-based species detectors and for transfer-learning from models like BirdNET. Contributors could extend the crowd-annotation protocol to new taxa or geographies, and practitioners can benchmark passive acoustic monitoring pipelines against it.
bioRxiv · geneticsConceptual★ flagship
Fourteen near-copy genes in a fungus share the work but none is fully replaceable.
When a gene gets accidentally duplicated, the spare copy is usually useless and gets discarded — but occasionally copies pile up into a whole family, and it's unclear how they divide labor. The fungus Candida albicans, a human pathogen, has an unusually large family of 14 related TLO genes, making it a perfect test case. Researchers deleted each of the 14 members one at a time and watched what changed, to see whether the copies are truly interchangeable or each does something slightly special. They found the genes form an overlapping network where they back each other up, but the redundancy is incomplete — losing one isn't fully covered by the others. Because these Tlo proteins plug into a master switch that controls gene expression, understanding their partial overlap sheds light on how gene families evolve new roles and how this pathogen tunes its behavior.
Technical view
A panel of single-deletion mutants for all 14 members of the lineage-specific TLO family in C. albicans was used to dissect molecular and biological redundancy among paralogs. Tlo proteins act as interchangeable subunits of the Mediator complex, so deletions probe how repeated duplication partitions function and constrains neofunctionalization. Phenotyping revealed interconnected functional networks with incomplete redundancy — paralogs buffer one another but are not fully substitutable. The mutant collection is a reusable resource for mapping paralog-specific contributions to Mediator-dependent transcriptional regulation and virulence phenotypes in this pathogen.
bioRxiv · biochemistry
bioRxiv · bioengineering
bioRxiv · bioengineering
bioRxiv · bioinformatics
bioRxiv · cancer biology
bioRxiv · cancer biology
bioRxiv · cancer biology
bioRxiv · plant biology
bioRxiv · neuroscienceConceptual
Your pupils lie about what's happening deep in a key brain 'alertness' circuit.
Deep in the brainstem sits the locus coeruleus, a tiny cluster of neurons that releases norepinephrine to control alertness and focus, shifting between a steady 'tonic' hum and sharp 'phasic' bursts. Because you can't easily peek into a living brain, scientists have used pupil size as a stand-in signal, assuming a dilated or twitchy pupil tells you which mode this circuit is in. By recording actual neuron activity and pupil size simultaneously in monkeys, researchers found the two often don't match up. That means a change in pupil size doesn't reliably tell you what the brain circuit underneath is really doing. This matters because tons of psychology and neuroscience studies use pupil size as a cheap window into brain state, and this result says that window is foggier than assumed.
Technical view
The study directly compared baseline-versus-evoked activity relationships in single-unit LC recordings against simultaneously measured pupillary dynamics in awake monkeys, testing whether the baseline-evoked coupling used to define 'tonic' (high baseline, low evoked) and 'phasic' (moderate baseline, high evoked) LC modes is mirrored in pupil metrics. They found the baseline-evoked relationship in pupil data does not predict the corresponding relationship in LC firing, and vice versa, indicating a dissociation between the two signals rather than a simple linear readout. This undercuts the common practice of inferring LC-NE activation mode purely from pupillometry in human and animal studies. Researchers using pupil-based proxies for LC state should validate against direct or converging physiological measures where possible, or restrict claims to pupil dynamics themselves rather than LC mode.
bioRxiv · neuroscienceConceptual
Serotonin doesn't just calm the brain — it flips whole regions into an off-like sleep state.
Deep in the brainstem, serotonin-releasing neurons send wires all over the front of the brain, and scientists have long argued about whether this chemical turns brain activity up or down. During light sleep or anesthesia, brain regions naturally flicker between 'UP' states (busy, active) and 'DOWN' states (quiet, resting) — and older experiments gave conflicting answers about which one serotonin promotes. This team used a precise laser-based technique to switch serotonin neurons on and off in mice, while simultaneously recording activity from thousands of neurons across the brain with high-density electrode probes. They found that releasing serotonin consistently pushed the cortex and a nearby region called the striatum into quiet DOWN states, while a deeper region, the midbrain, stayed mostly unaffected. This helps settle a long-standing debate and suggests serotonin's role in sleep and rest states is more suppressive and region-specific than previously thought.
Technical view
Using optogenetic stimulation of dorsal raphe nucleus (DRN) serotonergic neurons combined with Neuropixels recordings across cortex, striatum, and midbrain in lightly anesthetized mice, the authors show that selective 5-HT release reliably induces synchronized DOWN states in cortical and striatal networks while sparing midbrain dynamics, reconciling conflicting reports from optogenetic fMRI (suppression) versus classical electrical stimulation (UP-state induction). A multi-area computational model constrained by realistic connectivity was used to explain the regional selectivity, implicating differential local circuit architecture in how 5-HT modulates network excitability. This provides a mechanistic, cell-type-specific account of serotonergic control over cortical state transitions, with direct relevance for understanding sleep, anesthesia, and slow-oscillation pathologies. The Neuropixels + optogenetics approach offers a template for resolving other modality-dependent discrepancies in neuromodulator research.
bioRxiv · neuroscienceConceptual
Your memory center secretly has a map of what your eyes are looking at.
The hippocampus is the brain's classic memory hub, sitting at the very top of the chain of regions that process what we see, and yet nobody knew it kept track of visual space in a structured way. Earlier visual areas have 'retinotopic maps,' meaning neighboring brain cells respond to neighboring points in your field of view, like a distorted copy of the world laid out on brain tissue. Using an extremely powerful 7-Tesla MRI scanner and detailed modeling of how tiny clusters of brain tissue respond to visual stimuli, researchers scanned seven people extensively and found that the hippocampus does have such a map — and, surprisingly, it comes in two flavors: some spots respond with increased activity to a visual location, others with decreased activity. This suggests memory circuits stay in closer contact with raw visual information than scientists assumed, hinting at a deeper link between how we see and how we remember.
Technical view
Using 7T fMRI with voxel-scale population receptive field (pRF) modeling in seven densely-sampled subjects, the authors mapped retinotopic organization within the human hippocampus, finding robust pRFs distributed across subfields and along the hippocampal long axis. Roughly half the responsive voxels showed positive amplitude responses (canonical increases to a stimulus falling in their receptive field) and half showed negative amplitude responses (relative suppression), a bivalent code previously reported in high-level cortical memory networks but not shown at this resolution within the hippocampus itself. The pRFs displayed canonical visual response properties (e.g., systematic size and location tuning), arguing against these being artifacts of vasculature or eye-movement confounds. This opens a path for studying how spatial/visual coordinate frames are integrated into episodic memory formation, and could be replicated or extended with similar ultra-high-field pRF mapping protocols in clinical or memory-impaired populations.
bioRxiv · microbiologyConceptual
One parasite sneaks in with another, and the freeloader helps both survive inside your immune cells.
Kala-azar is a severe disease caused by the parasite Leishmania donovani, which hides and multiplies inside immune cells called macrophages that are normally supposed to destroy invaders. Doctors kept noticing that patients with kala-azar also often carried a second, usually harmless parasite called Leptomonas seymouri, but nobody knew if this hitchhiker mattered. Researchers infected mouse and human immune cells in lab dishes with each parasite alone and together, then tracked parasite DNA over about a week to see who was thriving. They found that when both parasites infect a cell together, the immune cells survive longer and the disease-causing parasite gets to persist and grow more successfully, meaning the 'bystander' parasite is actually an accomplice. This could reshape how doctors think about kala-azar severity and treatment, since co-infections might be quietly making the disease harder to fight.
Technical view
Using RAW 264.7 (murine) and THP-1 (human) macrophage models, the authors characterized intracellular survival and replication dynamics of Leptomonas seymouri alone versus in co-infection with Leishmania donovani, tracking parasite burden via quantitative ITS1 PCR over 48-168 hours post-infection. Co-infection significantly enhanced macrophage survival and increased intracellular L. donovani persistence relative to mono-infection, while L. seymouri itself showed active intracellular replication (rising DNA signal, reduced extracellular counts) rather than mere passive persistence. The findings implicate a previously monoxenous trypanosomatid as an active modulator of host-cell fate and co-pathogen fitness, potentially mediated by the associated Lepsey NLV1 virus, though that mechanistic link remains to be fully dissected. This motivates follow-up work on whether L. seymouri co-infection correlates with clinical kala-azar severity or drug resistance, and whether targeting the co-infection axis could improve treatment outcomes.
bioRxiv · molecular biologyConceptual
A virus edits its own gene into a cancer-driving version, and only in the right blood cells.
HTLV-1 is a virus that can cause an aggressive blood cancer called adult T-cell leukemia, and scientists have long known that a single viral gene called HBZ stays switched on throughout the disease. What wasn't clear is that this gene can be 'spliced' — edited after being copied — into different versions, and this study found that cancer cells overwhelmingly favor one particular spliced version, over 200 times more than the unedited one, while a different immune cell type keeps the unedited version instead. Using patient samples and lab-grown infected cells, researchers showed this splicing choice is controlled by a tug-of-war between two cellular proteins that promote or block the edit, and that only the edited version of the protein is actually built and drives the cancer's growth. This points to gene splicing itself, not just gene activity, as a hidden lever behind how this leukemia develops, opening a new angle for potential treatments that target the splicing machinery.
Technical view
The study quantified hbz splice isoforms in HTLV-1-infected cell lines and primary ATL patient samples, finding the spliced isoform HBZ_SP1 enriched over 200-fold in CD4 T cells from ATL patients relative to the unspliced usHBZ transcript, which predominates instead in CD8 T cells; critically, only HBZ_SP1 protein was detectable despite robust usHBZ transcription, implicating translational or stability control downstream of splicing. Mechanistically, the splicing decision is shown to be antagonistically regulated by the RNA-binding proteins hnRNPA1 and hnRNPH1, and HBZ_SP1 (but not usHBZ) was sufficient to drive oncogenic phenotypes in functional assays. This identifies a cell-type-specific alternative splicing switch as a driver of ATL leukemogenesis and nominates the hnRNPA1/hnRNPH1 splicing axis as a candidate therapeutic target, replicable via splice-isoform-specific qPCR or minigene splicing reporter assays in HTLV-1 model systems.
bioRxiv · molecular biologyBuildable
A new MRI dye lets doctors see kidney scarring building up long before damage becomes irreversible.
Polycystic kidney disease is a genetic condition where cysts and scar tissue slowly wreck the kidneys, but by the time doctors can detect the scarring (called fibrosis) with current tests, the damage is often already permanent. Researchers built a special dye, Gd-hProCA32.Collagen, designed to stick specifically to collagen, the fibrous protein that builds up as scarring progresses, and then injected it before an MRI scan so the scarred tissue would light up. In rats and mice bred to develop this kidney disease, the dye revealed early collagen buildup in kidneys and liver well before standard blood tests or imaging showed any problem, and it worked at one-tenth the dose of the current standard contrast agent while detecting nearly three times more signal. This means doctors could eventually catch this disease's damage far earlier, potentially allowing treatment before the kidneys suffer permanent, irreversible harm.
Technical view
The authors developed Gd-hProCA32.Collagen, a protein-engineered gadolinium-based MRI contrast agent targeted to type I collagen, and validated it in Pkhd1PCK/PCK rats and Pkd2-mutant mouse models of ADPKD for precision molecular MRI (pMRI) detection of early fibrotic remodeling in kidney and liver. At 10-fold lower dose than the clinical standard gadobutrol (Gadovist), the collagen-targeted agent detected approximately 2.8-fold greater signal enhancement correlating with fibrosis, and identified collagen deposition before conventional lab markers or standard imaging showed abnormalities. This establishes a noninvasive, quantitative biomarker for early-stage renal (and hepatic) fibrosis staging in ADPKD that could be used preclinically for drug efficacy studies and potentially translated to clinical trials as a surrogate endpoint for anti-fibrotic therapies. The reduced dosing requirement also suggests improved safety margins relevant to gadolinium retention concerns in repeated-imaging protocols.
bioRxiv · cell biologyBuildable
Tiny typos in one heart-muscle gene scramble its social network of protein partners, causing thickened hearts.
Hypertrophic cardiomyopathy is a genetic heart disease where the heart muscle grows abnormally thick, and it's often caused by small mutations, called missense variants, in a gene named MYBPC3 that makes a protein helping muscle fibers contract properly. Oddly, the mutated protein still ends up in the right place inside the muscle, so scientists didn't know exactly how it caused disease. This team looked at patient registries, measured how much of the mutant protein exists in actual diseased heart tissue, and used specialized lab techniques to map exactly which other proteins the mutant protein sticks to or associates with compared to the normal version. They found that these mutations change the protein's social circle — its pattern of interactions with neighboring structural proteins — even though the protein looks like it's in the right place. This suggests the disease isn't caused by the protein going missing or mislocated, but by it working with the wrong partners, a subtler and more specific mechanism that could guide future targeted therapies.
Technical view
The authors combined patient registry data, quantification of total and mutant-allele-specific MyBP-C protein levels in human left ventricular tissue, and flag-immunoprecipitation plus proximity-labeling mass spectrometry on four pathogenic MYBPC3 missense variants (R495Q, R502W in subdomain C3; W792R, R810H in subdomain C6) to profile their sarcomeric protein-protein interactomes against wild-type MyBP-C. Despite normal myofilament localization, the variants showed altered interaction profiles with other sarcomeric proteins, implicating disrupted protein-protein contacts within C3/C6 subdomains as a pathogenic mechanism distinct from mislocalization or loss of total protein. This reframes a subset of HCM-causing MYBPC3 missense mutations as interactome-disrupting rather than trafficking-defective, suggesting therapeutic strategies should target restoring specific binding interfaces rather than protein abundance or localization. The proximity-labeling MS approach offers a reusable pipeline for interrogating other localization-normal-but-pathogenic sarcomeric variants.
bioRxiv · ecologyConceptual
Seaweed doesn't even need to touch coral to poison it — and coral's tiny algae partners suffer too.
Coral reefs are increasingly losing ground to fast-growing seaweed, and when seaweed touches or grows near coral it can release chemical weapons that damage or kill the coral. Corals aren't solo organisms though — they live in partnership with communities of microbes and with tiny photosynthetic algae called Symbiodiniaceae that live inside their tissue and provide them energy, together called the coral holobiont. This study examined two coral species in Singapore to see how seaweed affects not just the coral itself but also these microbial and algal partners, and crucially, tested whether seaweed needs actual physical contact to cause harm or whether just being nearby (through diffusing chemicals in the water) is enough. The findings show seaweed can hurt corals through both direct contact and simple chemical diffusion, and that these attacks reshape the coral's microbial and algal communities. This matters because it helps explain why reefs can tip toward seaweed domination even without complete physical overgrowth, and points to the coral's tiny partners as an overlooked piece of that story.
Technical view
The study examined allelopathic (chemically-mediated antagonistic) interactions between a macroalga and two Singaporean coral species, Pocillopora acuta and a second species, distinguishing contact-based versus diffusion-based (waterborne, non-contact) mechanisms of seaweed-induced coral stress. Using microbiome profiling alongside Symbiodiniaceae community characterization, the authors tracked shifts in both bacterial associates and algal endosymbiont composition under each allelopathic exposure regime, testing whether microbiome perturbation and Symbiodiniaceae shifts are coupled outcomes of macroalgal stress. This addresses an open question about whether coral-associated microbiome changes causally mediate downstream effects on the algal symbiont community and coral health, relevant to predicting reef resilience during coral-algal phase shifts. The dual contact/diffusion experimental design offers a reusable framework for dissecting mechanism-specific holobiont responses in other coral-algal interaction studies.
bioRxiv · genomicsConceptual
Scientists sequenced the genome of a common bird malaria parasite for the first time.
Malaria isn't just a human disease — birds get their own versions too, caused by tiny blood parasites spread by biting flies. This study decoded the full genetic blueprint of one such parasite, Haemoproteus majoris, which infects songbirds. Because bird blood cells (unlike human ones) still have a nucleus full of DNA, the researchers had to specially filter out the bird's own DNA to isolate the parasite's genome. This matters because these bird parasites sit near the root of the malaria family tree, so understanding them helps scientists trace how malaria parasites evolved in the first place, including the lineage that eventually caused human malaria.
Technical view
The authors assembled the first genome for Haemoproteus majoris (lineage WW2), only the fourth avian haemosporidian genome sequenced to date, using DNA enrichment methods to overcome contamination from nucleated avian erythrocyte host DNA. Because phylogenomic analyses place Haemoproteus at the base of the haemosporidian tree, this assembly provides a critical outgroup/reference for comparative genomics against sequenced avian and mammalian Plasmodium genomes. Researchers can use this dataset to refine haemosporidian phylogenies, study gene family evolution (e.g., host-specificity and vector-adaptation genes), and calibrate divergence times for the malaria parasite lineage.
bioRxiv · genomicsConceptual
Your brain and your sperm-making cells may have evolved bigger together, through the same genetic tricks.
This research explores a surprising idea: genes that helped human brains grow larger over evolution might have first gained an advantage somewhere unexpected — in the stem cells that make sperm. The theory is that certain mutations gave those sperm-producing cells a competitive edge (a bit like how cancer cells can outcompete normal cells), got passed to offspring, and happened to also boost brain development because both tissues use similar growth signals. The tricky part is that these sperm precursor cells are rare and hard to study one by one with current lab techniques, so their genetic 'fingerprint' isn't well mapped yet. If true, this reframes brain evolution as partly a side effect of selection happening in a completely different organ.
Technical view
The study tests the hypothesis that brain-size-associated genetic variants originated via 'selfish spermatogonial selection' — an oncogenesis-like process where mutations confer a proliferative advantage to spermatogonial stem cells and become fixed in the germline, subsequently affecting neurogenesis through shared stem-cell proliferation signaling pathways. The core methodological challenge addressed is generating reliable single-cell transcriptomic signatures for spermatogonia despite their scarcity and profiling stochasticity. This work likely contributes reference spermatogonial expression profiles that others can cross-reference against known human brain-size GWAS loci and neurogenesis pathway gene sets.
bioRxiv · bioengineeringConceptual
A squishy injectable gel's stiffness alone can coax the body into building immune 'boot camps' inside tumors.
Tertiary lymphoid structures are like pop-up immune training centers that form near tumors and are linked to the body fighting cancer better. Scientists have been trying to artificially trigger these structures using injectable gel-like materials loaded with immune-signaling chemicals, but they'd mostly ignored a simpler variable: how soft or stiff the gel itself is. Here, researchers made injectable gels held together by fat-like molecules (liposomes) in soft and stiff versions that released the same chemical signals, then watched what grew inside the body. The soft gels attracted immune cells that gradually transformed the material into new, blood-vessel-rich tissue containing clusters of antibody-making B cells — showing that physical squishiness, not just chemistry, can steer immune architecture.
Technical view
The team engineered injectable, liposome-crosslinked supramolecular hydrogels with tunable viscoelasticity but matched release kinetics of ovalbumin (antigen) and LIGHT (a TNF-superfamily cytokine that promotes lymphoid neogenesis), isolating scaffold mechanics as the independent variable. Soft formulations supported distributed cellular infiltration and progressive material replacement, evolving from an early myeloid-dominated infiltrate into vascularized tissue with organized B-cell-rich aggregates characteristic of tertiary lymphoid structures (TLS). This establishes viscoelasticity as a designable parameter for in situ immune-niche engineering, with direct implications for biomaterial-based cancer immunotherapy scaffolds that could be optimized independently of soluble cue dosing.
bioRxiv · bioengineeringBuildable
Your Apple Watch could soon measure how hard your foot hits the ground, no lab required.
Scientists usually measure the force your foot exerts on the ground (useful for studying injuries, running form, or rehab) using expensive lab floor sensors called force plates. This project instead used two ordinary Apple Watches — one on the wrist, one at the waist — to stream motion data straight to an iPhone, where a compact AI model estimates that same force in real time without needing the internet or a lab. Tested on ten people, the two-watch setup got fairly close to real force-plate readings, and even a single wrist watch alone captured most of the signal. This points toward everyday gait and injury monitoring that works anywhere, not just in a research lab.
Technical view
The system streams 12-channel, 100Hz inertial data from two Apple Watch Series 6 units (wrist + waist) to an iPhone running fully on-device preprocessing and inference — no cloud roundtrip. Vertical ground reaction force (GRF) is predicted by GRFNet-MultiScale, a compact temporal convolutional network with four dilated residual blocks plus a global context branch, evaluated via leave-one-subject-out cross-validation across 539 stance windows from 10 participants, achieving Pearson r=0.798 and RMSE=257N for the dual-sensor configuration (wrist-only retained 82.5% of that correlation). The stable temporal attribution and edge-deployed architecture suggest a practical template for continuous, cloud-free biomechanical monitoring using off-the-shelf wearables, with wrist-only mode as a lower-fidelity fallback.
bioRxiv · bioinformaticsConceptual
Thousands of tiny DNA 'dimmer switches' vary between your organs — and reset when cells reprogram.
Beyond the DNA sequence itself, cells control genes using chemical tags called methylation marks, like sticky notes that turn genes up or down. This study scanned over 3,000 normal human tissue samples and found more than 12,000 spots in the genome where these tags reliably differ, even within the same tissue type — the researchers call these MeVars. Most of these differences are specific to particular tissues rather than showing up everywhere in the body, and interestingly, they largely disappear when cells are reprogrammed back into a stem-cell-like state or during sperm formation, as if reprogramming hits a 'reset' button. Some of these variable spots sit in genes linked to brain disorders and cancer, and the study found schizophrenia patients had fewer of these variants in certain genes than healthy people, hinting that this hidden layer of variation could influence disease risk.
Technical view
Using DNA methylation profiling across 3,370 normal human tissues spanning all three germ layers, the authors identified 12,587 methylation variant regions (MeVars, ~585bp average) with significant within-tissue methylation heterogeneity across 8,037 genes enriched for neurological disease and cancer associations; most MeVars showed tissue-specific rather than systemic patterns. Reprogramming to iPSCs in vitro and natural reprogramming during spermatogenesis in vivo both reduced this somatic methylation variation, and MeVars co-localized significantly with SINE1 elements, simple/low-complexity repeats, and H3K27me3/H3K9me3/H3K4me1 chromatin marks. In prefrontal cortex, MeVar incidence in candidate genes was higher in controls than schizophrenia patients with corresponding transcript-level changes, suggesting MeVars are a quantifiable epigenomic layer researchers can screen as disease-susceptibility biomarkers or mechanistic links between chromatin state and neuropsychiatric/oncologic risk.
bioRxiv · physiologyConceptual
Worms live longer when a low-oxygen alarm triggers a specific brain-to-body chemical relay.
When cells sense low oxygen (hypoxia), they can trigger a stress response that, in the roundworm C. elegans, actually extends how long the animal lives and stays healthy — but in mammals this same response can backfire in some tissues, blocking scientists from using it therapeutically. This study traces exactly which nerve cells and chemical signals in the worm's nervous system are responsible for the lifespan benefit, pinpointing a relay where oxygen-sensing triggers serotonin-producing neurons, which then signal to a different set of neurons using a specific serotonin receptor. Mapping this precise circuit is the first step toward figuring out how to get the lifespan benefits of the hypoxia response without its harmful side effects in mammals.
Technical view
The authors dissect the cell-nonautonomous circuit downstream of genetic HIF-1 (hypoxia-inducible factor) activation in C. elegans, showing that HIF-1 signaling specifically within ADF serotonergic neurons is both necessary and sufficient to drive lifespan extension. Downstream, serotonin acts through the SER-7 receptor on RIA (and related) GABAergic neurons to relay the longevity signal, defining a discrete bioaminergic-to-peptidergic neural circuit linking hypoxia sensing to organismal aging. This provides a tractable genetic entry point for identifying mammalian orthologs of this circuit, potentially enabling selective activation of pro-longevity hypoxic signaling while avoiding tissue-damaging effects of broad HIF-1 activation.
bioRxiv · physiologyConceptual
Low iron in pregnant rats seems to protect the heart from high blood pressure — but at a hidden cost.
Doctors have noticed that pregnant people with mild iron deficiency sometimes show lower blood pressure and a heart that seems to work more efficiently, especially in high-risk pregnancies — but nobody knew if that apparent benefit came with a downside inside the heart's own energy factories, the mitochondria. This study fed genetically hypertensive pregnant rats either normal or iron-restricted diets and then examined their heart cells' mitochondria in fine detail — their shape, how well they produce energy, and their chemical stress levels — near the end of pregnancy. Because mitochondria are the heart's power plants and iron is essential for that power generation, the fear is that a seemingly 'good' blood pressure effect might actually be masking cellular stress. The findings help clarify whether iron deficiency in high-risk pregnancies is truly protective or just trading one problem for another.
Technical view
Using spontaneously hypertensive rats (SHR) as a model of hypertensive pregnancy and normotensive Wistar-Kyoto (WKY) controls, the authors manipulated dietary iron before and throughout gestation and assessed cardiac mitochondrial ultrastructure (TEM), oxidative phosphorylation capacity (high-resolution respirometry), mitochondrial dynamics, and redox status at gestational day 21. The study directly tests whether iron-deficiency-associated reductions in blood pressure and improved apparent cardiac efficiency in hypertensive pregnancy are mechanistically linked to altered mitochondrial bioenergetics and stress-response remodeling. This positions cardiac mitochondrial phenotyping as a readout for evaluating whether iron restriction is a viable or risky adaptive strategy in hypertensive pregnancy, informing future work on maternal iron supplementation thresholds in preeclampsia-risk populations.
bioRxiv · plant biologyBuildable
AI-designed proteins were bolted onto plant immune sensors so rice could 'see' a banana-killing fungus.
Plants have built-in immune receptors that recognize invading pathogens and trigger a defense response, but breeding new resistance genes by hand is slow, and pathogens keep evolving to dodge detection. Here, scientists used generative AI protein-design tools (similar in spirit to AI that designs new proteins from scratch) to invent brand-new 'detector' protein modules aimed at recognizing specific molecules from crop pathogens. They then plugged these AI-designed detectors into an existing, modular rice immune receptor system, successfully creating a receptor that could recognize a disease-causing protein from the fungus behind Panama disease, a major threat to banana crops, that it couldn't normally detect. This shows a path toward rationally engineering crop disease resistance instead of waiting to discover it in nature.
Technical view
The authors used RFdiffusion (structure generation) and ProteinMPNN (sequence design) to computationally design de novo integrated domains (IDs) as bespoke binders against pathogen effector proteins, then grafted these into the modular rice blast Pik-1/Pik-2 NLR (nucleotide-binding leucine-rich repeat) immune receptor chassis. This engineered novel effector recognition against a non-cognate virulence factor from Fusarium oxysporum f. sp. cubense Tropical Race 4 (Panama disease, a major threat to Cavendish banana), demonstrating that generative protein design can be used to rationally program NLR effector specificity rather than relying on natural allelic diversity. This establishes a generalizable pipeline — design ID, integrate into a validated NLR scaffold, test recognition/HR response — that could be extended to engineer resistance against other economically important effectors across crop species.
bioRxiv · plant biologyBuildable
A cyanide-related plant chemical can trick cells into growing a whole new embryo.
Plants make a hormone called auxin that tells cells how to grow, and one flavor of it, phenylacetic acid (PAA), seems to help cells 'reset' into an embryo-like state — a process called somatic embryogenesis, basically cloning a plant from a single body cell. This study looked at benzyl cyanide, a chemical cousin/precursor to PAA, and found that even tiny doses coax Arabidopsis (a lab mustard plant) cells into starting embryo formation. They checked which genes switched on afterward and saw the plant's own 'build an embryo' toolkit light up, plus independent evidence that auxin signaling ramped up. It matters because understanding this trigger could improve how scientists clone plants or engineer crops from single cells.
Technical view
The authors test whether benzyl cyanide/phenylacetonitrile (BnCN/PAN), a proposed CYP79A2-pathway intermediate en route to the auxin phenylacetic acid (PAA) via NITRILASE-catalyzed conversion, promotes somatic embryogenesis (SE) in Arabidopsis thaliana. Low-dose BnCN/PAN and exogenous PAA both induced embryogenic induction, and RNA-seq showed upregulation of core SE regulators and EMBRYO DEFECTIVE genes alongside enhanced auxin signaling, corroborated with a pDR5::GUS auxin-response reporter. This supports a model where BnCN/PAN acts as a NIT-dependent PAA precursor driving SE, giving researchers a genetically tractable node (NIT enzymes, CYP79A2 pathway) to manipulate for boosting in vitro regeneration efficiency.
bioRxiv · neuroscienceConceptual
Hidden wrinkles buried deep in brain folds mark invisible borders between vision regions.
Your brain's surface is folded into ridges and grooves, and scientists have long linked the big folds to brain function — but most of the cortex is actually buried inside the smaller grooves (sulci), and nobody knew if those hidden crinkles meant anything. This study zoomed into the visual part of the brain and found four small, previously unnoticed folds tucked inside one major groove. Using brain scans that measure both shape (structural MRI) and activity (functional MRI) in two separate groups of people, they showed these tiny folds reliably line up with the boundaries between different 'maps' of the visual field the brain uses to process what you see — about as accurately as current best guesses. It matters because it suggests the brain's physical wrinkling pattern could be a shortcut for finding functional boundaries without needing to scan brain activity every time.
Technical view
Using combined structural and functional MRI in observation and replication cohorts, the authors identify four consistent secondary folds within the major sulcus spanning the dorsal visual stream and show these 'hidden gyri' predict the boundaries between visual field maps with accuracy comparable to existing probabilistic atlases. This links tertiary/quaternary folding patterns, not just primary sulcal/gyral anatomy, to fine-grained functional organization. The approach offers a structural proxy for functional visual field map boundaries, potentially reducing reliance on individual fMRI retinotopic mapping and informing models of how cortical folding mechanically or developmentally tracks areal parcellation.
bioRxiv · neuroscienceRunnable
Brainwave timing from one scan makes a second brain-zapping scan more precisely targeted.
Before some brain surgeries, doctors map where speech is controlled by zapping the brain gently with magnetic pulses (TMS) and seeing what disrupts a patient's speech — but everyone's speech network sits in a slightly different spot. This study asked whether first recording a person's natural brain activity during a naming task (using MEG, a machine that reads magnetic signals from brain activity) could tell you both where and exactly when to fire the TMS pulse for better accuracy. In 13 healthy volunteers, they timed TMS pulses to match moments of real MEG-recorded brain activity and found that hitting those precise moments and locations caused more speech errors, meaning the mapping was working better. It matters because more accurate, personalized maps could help neurosurgeons avoid damaging critical speech areas during brain tumor or epilepsy surgery.
Technical view
The study combines MEG-derived spatiotemporal dynamics of picture-naming-related speech production with navigated repetitive TMS (SCM nrTMS), adjusting the picture-to-TMS interval (PTI) to coincide with individually measured MEG activation timing and location. Across 13 healthy subjects, TMS pulses timed to MEG-defined activity windows produced significantly higher speech disruption (error) rates, and pulse latency correlated with MEG activation latency at the group level. This establishes a proof-of-concept pipeline for individualizing both the spatial target and temporal window of TMS mapping using a subject's own MEG data, which could be integrated into presurgical protocols to sharpen localization of eloquent speech cortex beyond current fixed-timing TMS approaches.
bioRxiv · neuroscienceBuildable
Two different brain wiring paths make mice more, or less, likely to get hooked on cannabis-like drugs.
Addiction involves a brain region called the nucleus accumbens, which acts like a reward hub, but it gets input wires from many other brain areas, and it wasn't clear which wires matter for cannabis-type addiction specifically. Researchers used mice trained to give themselves a synthetic marijuana-like drug through a lever press, then used a technique that lets them temporarily 'switch off' one wire at a time — either the connection from the memory-related hippocampus or from the fear/emotion-related amygdala into the reward hub. By comparing what happens to drug-seeking behavior when each wire is silenced, they could tell which pathway pushes the mice toward compulsive use. This matters because it points to specific brain circuits that might one day be targeted by treatments to reduce cannabis use disorder risk in people.
Technical view
Using WIN55,212-2 intravenous self-administration in male C57BL/6J mice as a validated CUD model, the authors employ pathway-specific hM4Di DREADD chemogenetic inhibition to selectively silence dorsal hippocampus (dHPC)-to-NAc versus basolateral amygdala (BLA)-to-NAc glutamatergic projections during chronic drug exposure. Differential effects of inhibiting each pathway on self-administration behavior and addiction vulnerability implicate circuit-specific contributions of these two limbic afferents to the NAc in cannabinoid addiction. This dissociation gives a template for further circuit-mapping work (e.g., fiber photometry, optogenetic manipulation) to pinpoint which glutamatergic inputs are viable targets for circuit-based interventions in CUD.
bioRxiv · neuroscienceConceptual
A brainwave rhythm reveals whether your eyes are hunting for something new or recalling a memory.
When you look around a scene, sometimes your eyes are just exploring and sometimes they're being guided by memory of where something was before, and scientists wanted to know if the brain shows a different 'signature' for each. They focused on theta waves, a rhythmic brain signal in the hippocampus (the brain's memory hub) that's known to sync up with eye movements in monkeys, and asked whether the same thing happens in humans and whether it changes depending on memory. Using electrodes implanted in the brains of 11 patients (already there for unrelated medical reasons) plus eye-tracking, they measured theta activity right around each eye movement (fixation) and found that its patterns differed depending on whether the movement was memory-guided or not. This matters because it helps explain, at the level of brain rhythms, how memory steers where we look — relevant to understanding both normal exploration and memory disorders.
Technical view
Using intracranial EEG and eye-tracking from 11 neurosurgical patients during a naturalistic viewing paradigm (Keles et al., 2024 dataset), the authors perform fixation-locked analyses of hippocampal theta oscillations, examining slow (3–6 Hz) versus fast (6–10 Hz) sub-bands. They report that hippocampal theta dynamics reliably differentiate memory-guided from exploratory (memory-independent) saccades, extending primate findings on eye movement-theta coupling to human episodic memory-guided viewing. The dissociation between theta sub-bands offers a candidate physiological marker for oculomotor-hippocampal interaction that could be probed further with saccade-direction-specific analyses or applied to studying memory deficits via gaze/iEEG biomarkers.
bioRxiv · neuroscienceConceptual
SIDS babies' cerebellums look metabolically 'stuck,' hinting at a hidden brain failure point.
Sudden Infant Death Syndrome (SIDS) is thought to involve a failure of the brain to 'wake up' the body during dangerous drops in oxygen, and the cerebellum — a brain region involved in coordinating automatic survival responses — might be part of that failure. Researchers examined preserved brain tissue from infants who died of SIDS and compared it to similarly aged infants who didn't, looking at which genes were active, what metabolic chemicals were present, and the tissue's physical structure. They found signs of inflammation and an imbalance in two key brain chemicals (GABA and glutamate) that normally excite and calm neurons, along with signs the cerebellum hadn't matured properly. This matters because it points to a specific, testable brain circuit — the cerebellum's role in emergency breathing responses — as a possible hidden weak link behind SIDS deaths.
Technical view
The study performs multi-omic integration (RNA-seq, targeted metabolomics, quantitative histology) on postmortem cerebellar tissue from SIDS cases versus age-matched controls, with cross-omic concordance analysis linking metabolite abundance to transcriptional direction of the corresponding metabolic enzymes. Results reveal a distinct SIDS cerebellar signature marked by neuroinflammatory gene expression and imbalanced GABAergic/glutamatergic metabolism, alongside evidence of developmental immaturity, consistent with a model of arrested cerebellar maturation compromising autoresuscitation circuitry. This multi-omic concordance framework is replicable on other postmortem SIDS cohorts or brain regions to test whether the identified GABA/glutamate and inflammatory signature is cerebellum-specific or part of a broader brainstem-cerebellar respiratory network failure.
bioRxiv · neuroscienceConceptual
Two proteins linked to ALS team up inside neurons to help build motor neurons correctly.
TDP-43 is a protein famous for going wrong in ALS (a disease that destroys motor neurons controlling movement), and SNRNP70 is another protein normally known for helping process genetic instructions (splicing) inside the cell nucleus, but it also does mysterious jobs out in the rest of the cell. This study found that SNRNP70 and TDP-43 show up in the same places inside developing neurons — including out in the long nerve fibers (axons) — and that they physically team up. It turns out SNRNP70 needs TDP-43's help to get recruited into cellular 'freight containers' (RNP granules) that transport genetic material, and this partnership is essential for motor neurons to develop properly. This matters because it uncovers a new working relationship between two proteins tied to motor neuron disease, potentially explaining part of how ALS damages cells.
Technical view
The authors show that SNRNP70, a core spliceosomal U1 snRNP component previously implicated in axonal mRNA transport during zebrafish motor neuron development, colocalizes with TDP-43 in both nuclear and cytoplasmic compartments, including axonal RNP granules, and that this co-association is functionally required for motor neuron development. Critically, TDP-43 is shown to be necessary for recruiting SNRNP70 to cytoplasmic RNP granules, establishing a hierarchical, TDP-43-dependent mechanism for SNRNP70's non-canonical cytoplasmic function. This defines a new protein-protein interaction axis relevant to ALS/FTD pathobiology, suggesting that disruption of TDP-43-SNRNP70 granule recruitment (rather than splicing dysfunction alone) could be a route to motor neuron pathology, and provides a system (zebrafish motor neuron development) for further dissecting granule assembly determinants.
bioRxiv · neuroscienceConceptual
Neurons arrange ion channels into a precise honeycomb grid to shape brain cell firing patterns.
Neurons fire electrical spikes, and after firing, some neurons pause briefly before they can fire again — a kind of built-in cooldown called an afterhyperpolarization — controlled by a team of calcium and potassium channels. Scientists already knew this cooldown happens in hippocampal neurons (key for memory), but not how those channels are physically arranged on the cell body to make it work. Using super-resolution microscopy (imaging fine enough to see structures far smaller than a wavelength of light) and computational pattern analysis, they discovered the channels aren't scattered randomly but form a striking repeating honeycomb-like lattice, spaced with clock-like regularity, that overlaps with the cell's internal scaffolding protein network (spectrin). This matters because it shows the cell's structural skeleton isn't just architecture — it actively organizes electrical machinery, a principle that could apply to how neurons fine-tune their firing patterns more broadly.
Technical view
The authors characterize the sub-cellular organization of the CaRyK complex (Cav1.3 calcium channels, RyR2 ryanodine receptors, and IK potassium channels) at ER-plasma membrane junctions in hippocampal pyramidal cell somata, which generates the slow afterhyperpolarization (sAHP) governing spike-frequency adaptation. Super-resolution imaging combined with dimensionality-reduction analysis reveals CaRyK clusters arranged in rows with ~155 nm periodicity, forming a non-rigid polygonal lattice extending to dendritic branchpoints, and this lattice colocalizes precisely with the βII-spectrin cytoskeleton. This establishes the spectrin-actin cytoskeleton as a structural scaffold organizing somatic ER-PM junction channel complexes, giving a concrete nanoscale architecture that could be probed for how cytoskeletal disruption (e.g., in disease models) might alter sAHP-dependent excitability.
bioRxiv · neuroscienceConceptual
Monkey brains track both 'where I am' and 'where I'm looking' with separate cell types.
For decades neuroscientists have debated whether monkeys and rodents navigate space using the same brain trick. Rats rely on 'place cells' that fire when the animal is in a specific spot, while earlier monkey studies found 'view cells' that fire based on where the animal is looking, even from a distance. By recording individual neurons in macaque hippocampus while the animals walked down tracks and made choices, researchers found the brain actually keeps three separate populations: cells for location, cells for gaze direction, and cells that do both at once. This matters because it suggests primates didn't replace the rodent-style map, they layered a new one on top of it, coexisting side by side.
Technical view
Using single-unit recordings in macaque hippocampus during linear-track navigation across two visually distinct contexts, the authors applied a generalized linear model to statistically dissociate position coding from spatial-view coding in the same neural population. They identified three coexisting classes: Position cells (rodent-like, with localized firing, directional tuning, and contextual remapping), View cells, and Conjunctive cells encoding both. This resolves the rodent-vs-primate spatial coding debate by showing dissociable-but-coexisting codes rather than a wholesale species difference, and provides a GLM-based decomposition method others can apply to disentangle egocentric/allocentric signals in mixed-selectivity hippocampal data.
bioRxiv · neuroscienceConceptual
Aging brains show wiring and activity patterns that drift further apart over the decades.
Your brain has a physical wiring diagram (which regions connect to which) and a pattern of actual activity that flows over those wires, and normally these two are tightly coupled. This study asks what happens to that coupling as people age, using brain scans from nearly 600 healthy adults aged 18 to 88. By treating the brain as a 'multiplex network,' essentially two overlapping maps, structural and functional, layered on the same nodes, and measuring how statistically different the two maps become using information-theory tools, the researchers found that structure and function increasingly diverge as people get older. This gives scientists a new numerical signature of healthy brain aging that goes beyond just seeing individual regions shrink or dim.
Technical view
The authors model structural connectivity (SC) and functional connectivity (FC) from diffusion and resting-state MRI as two layers of a multiplex network in 589 subjects (ages 18-88, Cam-CAN dataset), then quantify inter-layer coupling using Jensen-Shannon Divergence and relative entropy rather than standard correlation-based SC-FC coupling metrics. They report a progressive, age-related increase in information divergence between the layers, indicating the functional layer decouples from its structural scaffold with age. This information-theoretic framework offers a replicable, cross-cohort applicable metric for tracking SC-FC decoupling trajectories, potentially useful as a normative baseline against which neurodegenerative divergence patterns could be compared.
bioRxiv · neuroscienceBuildable
A math trick unrolls the brain's donut-shaped map of space back into a real walking path.
Deep in the brain, grid cells create a mental map of space that, weirdly, forms a shape like a donut (a torus) when you plot their activity, because the pattern repeats periodically as you move. The catch is that a repeating, looping code is ambiguous: many different real-world locations could produce the same donut-coordinate. This paper borrows a tool from topology (the math of shapes) to 'unroll' the looping donut coordinates back into a straight, sensible path through actual space, similar to how you'd unroll a spiral staircase into a straight ramp. Being able to reconstruct the animal's true trajectory from grid cell firing alone would be a major step toward reading spatial thoughts directly from neural activity.
Technical view
Grid cell population activity is known to lie near a toroidal manifold; the authors use topological data analysis to extract toroidal (periodic) coordinates from this activity and then apply 'path lifting' through covering-space theory to unwrap the periodic torus coordinates into a continuous trajectory in physical space, recovering the true path up to an affine transformation. They validate the method on both continuous attractor network simulations and experimental recordings, demonstrating it can decode spatial trajectories without needing environment-specific calibration. This provides a principled, topology-based decoding pipeline that practitioners could adapt for any periodic neural code (not just grid cells) where standard linear decoders struggle with the code's inherent ambiguity.
bioRxiv · neuroscienceRunnable
A new push-button toolkit finally makes monkey brain scans as easy to process as human ones.
Scientists who study monkey brains with MRI face a headache human neuroimagers don't: nearly all the standard software was built and tuned for human brains, so it often breaks or gives unreliable results on macaques. Brainana is a new all-in-one software package that automatically cleans, aligns, and organizes macaque brain scans, handling everything from stripping out the skull to reconstructing the wrinkly surface of the cortex, using deep learning models specifically trained on monkey brains. It even includes a viewer so researchers can visually check their results, and it can run in the cloud for labs without powerful computers. This kind of standardized, reproducible pipeline matters because it lets different labs' monkey imaging results actually be compared apples-to-apples.
Technical view
Brainana is a containerized, BIDS-compatible preprocessing framework for macaque MRI that integrates structural and functional preprocessing, cortical surface reconstruction, quality control, transform tracking, and atlas projection into a single automated pipeline, with cloud deployment for compute-limited users. It uses macaque-trained deep learning models for brain extraction and tissue segmentation and includes macaque-specific optimizations for orientation standardization and surface reconstruction to handle acquisition variability across sites. A built-in viewer links volumetric and surface derivatives for QC inspection, positioning Brainana as a reproducibility-focused analog to human pipelines like fMRIPrep/HCP for the nonhuman primate imaging community.
bioRxiv · neuroscienceConceptual
Where your attention drifts, unconsciously, decides which musical line becomes 'the melody' you hear.
When you listen to music with multiple instruments or voices playing at once, like a Bach fugue, your brain somehow picks out one strand as 'the melody' even though nobody told it which one to follow. This study used brain-wave recordings (EEG) plus behavioral tests to find out how that happens, and discovered that it's driven by spontaneous, involuntary shifts in attention rather than a fixed rule like 'always follow the highest note.' In other words, your brain isn't passively receiving a melody, it's actively constructing one by unconsciously spotlighting different voices, informed by how well your brain can predict what note comes next in each line. This helps explain a basic mystery of music perception: how we turn a wall of simultaneous sound into a single coherent tune.
Technical view
The authors combined behavioral melody-identification measures with EEG to test how spontaneous, stimulus-driven attentional allocation and predictive/melodic-expectation mechanisms jointly determine which voice in polyphonic (multi-voice) music is perceived as the foreground melody. Their results argue against existing single-mechanism accounts (divided attention, figure-ground, stream integration alone) and instead support an interaction between spontaneous attentional bias and predictive processing, extending prior monophonic-melody predictive-coding findings to the polyphonic case. This offers a testable neural framework (likely via evoked/entrained EEG responses to expectancy violations per voice) for researchers studying auditory scene analysis, with potential application to modeling voice-leading salience in music cognition or algorithmic composition.
bioRxiv · neuroscienceBuildable
How well individual brain wiring links steer neural states may predict who responds to depression zaps.
Repetitive transcranial magnetic stimulation (rTMS) is a real, FDA-cleared treatment that uses magnetic pulses to shift brain activity and treat depression that hasn't responded to medication, but doctors currently can't predict in advance who it will actually help. This study looks at the brain's wiring diagram (from diffusion MRI scans) in 25 depression patients before treatment, and instead of just asking how important individual brain regions are for controlling brain-state changes, it examines how important individual connections (edges) between regions are, especially ones tied to a region called the middle frontal gyrus, near where rTMS is typically applied. They found that patients whose specific connections had higher 'controllability' (a measure of how much that wire can steer the brain into new states) tended to respond better to treatment. This could eventually let doctors use a brain scan to personalize who gets rTMS and where to target it.
Technical view
In 25 treatment-resistant MDD patients undergoing 5 weeks of high-frequency dlPFC-targeted rTMS, the author constructed structural connectomes from diffusion MRI (MRtrix3, Destrieux atlas) and computed edge-level (rather than the more common node-level) controllability metrics from network control theory at baseline. Controllability of specific edges centered on the middle frontal gyrus was significantly associated with clinical treatment response, suggesting edge-based metrics carry predictive signal that node-level summaries may miss. This provides a candidate baseline biomarker and an edge-controllability computational pipeline that could be validated in larger cohorts or used to personalize rTMS coil targeting based on individual connectome control properties.
bioRxiv · neuroscienceConceptual
General intelligence isn't in one brain spot, it rides on wiring spanning almost the entire brain.
Psychologists have long measured a single underlying factor called 'g,' or general cognitive ability, that predicts performance across many different mental tasks, but where in the brain's wiring this ability actually lives has been unclear. This massive study looked at brain scan data from nearly 39,000 people aged 26 to 84 across three separate datasets, measuring the white matter 'cables' that connect different brain regions using several different scan-based measures of connection strength and health. They found that general cognitive ability isn't tied to any one hotspot, it's supported by connections spread across almost the entire brain, especially long-distance connections within the same hemisphere, and this wiring pattern lines up with independent measurements of gray matter thickness. Notably, the link between these wiring health measures and cognitive ability actually gets stronger as people age, hinting that maintaining broad brain connectivity becomes more important, not less, later in life.
Technical view
Across 38,824 individuals (26-84 years) from three cohorts, the authors performed meta-analytic association of general cognitive function (g) with structural connectome measures, streamline count, fractional anisotropy (FA), and mean diffusivity (MD), at global, nodal, and edge levels. They report widespread significant associations spanning all cerebral lobes and key subcortical structures, driven by both inter- and especially intra-hemispheric long-range inter-lobar connections, with white matter node-g associations spatially converging with independent cortical morphometry-g associations. Effect sizes strengthen with age (MD associations more negative, FA more positive), suggesting white matter integrity becomes an increasingly important substrate of g across the lifespan; this large-scale, multi-cohort meta-analytic approach offers a reference map that could be used to benchmark connectome-based cognitive biomarkers or study divergence in clinical populations.