bioRxiv : the preprint server for biology最新文献

Large language models (LLMs) show promise in biomedical research, but their effectiveness for genomic inquiry remains unclear. We developed GeneTuring, a benchmark consisting of 16 genomics tasks with 1,600 curated questions, and manually evaluated 48,000 answers from ten LLM configurations, including GPT-4o (via API, ChatGPT with web access, and a custom GPT setup), GPT-3.5, Claude 3.5, Gemini Advanced, GeneGPT (both slim and full), BioGPT, and BioMedLM. A custom GPT-4o configuration integrated with NCBI APIs, developed in this study as SeqSnap, achieved the best overall performance. GPT-4o with web access and GeneGPT demonstrated complementary strengths. Our findings highlight both the promise and current limitations of LLMs in genomics, and emphasize the value of combining LLMs with domain-specific tools for robust genomic intelligence. GeneTuring offers a key resource for benchmarking and improving LLMs in biomedical research.

Primary motor cortex (M1) plays a central role in voluntary movement, but how it integrates sensory-driven corrective instructions is unclear. We analyzed population activity recorded from M1 of macaques during a sequential arm movement task with target updates requiring online adjustments to the motor plan. Using Latent Factor Analysis via Dynamical Systems (LFADS), we separated neural activity into two components: intrinsic dynamics and inferred external inputs influencing those dynamics. Inferred input timing was more strongly locked to target appearance than to movement onset, suggesting that variable reaction times reflect interactions between inputs and ongoing dynamics. Inferred inputs were tuned similarly for both initial and corrective movements, suggesting a shared input encoding across visually-instructed and corrective movements that was previously obscured by M1 dynamics. Because input inference can suffer from the challenge of nonidentifiability, where different models fit the data indistinguishably, we used ensembles of models with varied hyperparameters to diagnose when inputs are identifiable or nonidentifiable. In the monkey data, ensembles produced consistently similar results, suggesting that inputs could be meaningfully inferred and that their encoding was not simply a result of model bias. These results highlight the challenges of nonidentifiability and the potential of model ensembles to identify inputs in ongoing dynamics, at least in some cases.

Recent studies have shown that neural representation and processing are widely distributed in the brains of behaving animals [1, 2, 3, 4]. These observations challenge functional specialization as a central tenet of Neuroscience, which refers to the notion that distinct brain regions are dedicated to specific aspects of cognition such as working memory or subjective decision-making. Here we develop the concept of bifurcation in space to mechanistically account for the emergence of functional specialization that is compatible with distributed neural coding in a large-scale neocortex. Our theory starts with a departure from the canonical local circuit principle [5] by highlighting differences between cortical areas in the form of experimentally quantified heterogeneities of synaptic excitation and inhibition. We investigated connectome-based modelling of a multiregional cortex for both macaque monkeys and mice, as well as a generative model of a spatially embedded neocortex. During working memory in a simulated delayed response task, surprisingly, we found an inverted-V-shaped pattern of neuronal timescales across the cortical hierarchy as a signature of functional modularity, in sharp contrast to an increasing pattern of timescales during the resting state, as reported previously [6]. Furthermore, our model cortex simultaneously and robustly displays a plethora of bifurcations in space and their associated rich repertoire of timescale profiles across a large-scale cortex; the corresponding functionally defined modules (spatial attractors) could potentially subserve various internal mental processes. This work yields several specific experimentally testable predictions, including an inverted-V pattern of timescales, a measure of comparison between functional modules and structural modules defined by the graph theory, and a new plot for revealing bifurcation in space in neural activity recorded from animals performing different tasks that engage various functional modules. We propose that bifurcation in space, resulting from the connectome and macroscopic gradients of neurobiological properties across the cortex, represents a fundamental principle for understanding the brain's functional specialization and modular organization.

Recent work has demonstrated that both permanent lesions and acute inactivation experiments can lead to erroneous conclusions about the causal role of brain areas in specific behaviors, casting serious doubt on major avenues by which neuroscientists study the brain. To overcome this challenge, we developed a three-stage optogenetic approach which leverages the ability to precisely control the temporal period of regional inactivation with either brief or sustained illumination, enabling investigators to dissociate between putative 'permissive' and 'instructive' roles of brain areas in behavior. We applied this approach to the mouse primary visual cortex (V1) to probe whether V1 is permissive or instructive for the detection low contrast stimuli. Acute inactivation of V1 drastically suppressed performance, but during persistent inactivation, the animals' contrast detection recovered to pre-silencing levels. This recovery was itself reversible, as returning the animals to intermittent V1 inactivation reinstated the behavioral deficit. These results argue that V1 is the default circuit mice use to detect visual stimuli, but in its absence, other regions can compensate for it. This novel, temporally controllable optogenetic perturbation paradigm should be useful in other brain circuits to assess whether they are instructive or permissive in a brain function or behavior.

The striatal direct and indirect pathways constitute the core for basal ganglia function in action control. Although both striatal D1- and D2-spiny projection neurons (SPNs) receive excitatory inputs from the cerebral cortex, whether or not they share inputs from the same cortical neurons, and how pathway-specific corticostriatal projections control behavior remain largely unknown. Here using a G-deleted rabies system in mice, we found that more than two-thirds of excitatory inputs to D2-SPNs also target D1-SPNs, while only one-third do so vice versa. Optogenetic stimulation of striatal D1- vs. D2-SPN-projecting cortical neurons differently regulate locomotion, reinforcement learning and sequence behavior, implying the functional dichotomy of pathway-specific corticostriatal subcircuits. These results reveal the partially segregated yet asymmetrically overlapping cortical projections on striatal D1- vs. D2-SPNs, and that the pathway-specific corticostriatal subcircuits distinctly control behavior. It has important implications in a wide range of neurological and psychiatric diseases affecting cortico-basal ganglia circuitry.

Chimeric antigen receptor (CAR) T cell therapy has shown remarkable efficacy in cancer treatment. Still, most patients receiving CAR T cells relapse within 5 years of treatment. CAR-mediated trogocytosis (CMT) is a potential tumor escape mechanism in which cell surface proteins transfer from tumor cells to CAR T cells. CMT results in the emergence of antigen-negative tumor cells, which can evade future CAR detection, and antigen-positive CAR T cells, which has been suggested to cause CAR T cell fratricide and exhaustion. Whether CMT indeed causes CAR T cell dysfunction and the molecular mechanisms conferring CMT remain unknown. Using a selective degrader of trogocytosed antigen in CAR T cells, we show that the presence of trogocytosed antigen on the CAR T cell surface directly causes CAR T cell fratricide and exhaustion. By performing a small molecule screening using a custom high throughput CMT-screening assay, we found that the cysteine protease cathepsin B is essential for CMT and that inhibition of cathepsin B is sufficient to prevent CAR T cell fratricide and exhaustion, leading to improved long-term CAR T cell persistence and anti-tumor activity. Our data demonstrate that it is feasible to separate CMT from cytotoxic activity, that CAR T cell persistence, a key factor associated with clinical CAR T cell efficacy, is directly linked to cathepsin B activity in CAR T cells, and that it is possible to improve CAR T cell function through selective inhibition of CMT.

We previously demonstrated that weight cycled mice have increased adipose mast cells compared to obese mice by single cell RNA-sequencing. Here, we aimed to confirm and elucidate these changes. Interestingly, we did not detect an increase in total mast cell numbers in weight cycled mice by Toluidine blue or flow cytometry, however, further subcluster analysis of our dataset showed that our initial mast cell cluster consisted of two unique populations. One population had very high expression of classical mast cell markers and another had elevated lipid handling and antigen presentation genes with a concomitant reduction in classical mast cell genes. This new "lipid-associated" mast cell cluster accounted for most of the mast cells in the weight cycled group. We induced a similar phenotype in vitro using repeated exposure to adipose tissue conditioned media to mimic weight gain and weight regain. Upon repeated exposure to adipose tissue conditioned media, bone marrow-derived mast cells had increased lipid droplets and reduced expression of cKit and FcεR1 compared to control cells. Moreover, we analyzed mast cells in a pilot study of subcutaneous adipose tissue from four obese, prediabetic women. We found two mast cell populations that appear similar to the murine populations detected by sequencing. The population with reduced cKit and FcεR1 was significantly correlated with weight variance. Together, these data suggest that weight cycling may induce a unique population of mast cells similar to lipid- associated macrophages, which have been shown to play a role in diverse diseases from obesity and atherosclerosis to Alzheimer's disease. Future studies will focus on isolation of these cells from mice and humans to better determine their lineage, differentiation, and functional roles.

The dopaminergic system has been extensively studied for its role in behavior and neurological diseases. Despite this, we still know little about how dopamine levels are regulated in vivo. To identify regulators of dopamine, we utilized Drosophila melanogaster cuticle pigmentation as a readout, where dopamine is used as a precursor to melanin. We started by measuring dopamine from known pigmentation mutants (e.g. tan, ebony, black) and then performed an RNAi-based screen to identify new regulators. We found 153 hits, which were enriched for developmental signaling pathways and mitochondria-associated proteins. From 35 prioritized candidates, 11 had an effect on head dopamine levels. Effects on brain dopamine were mild even when the rate-limiting synthesis enzyme Tyrosine hydroxylase (TH) was knocked down, suggesting changes in dopamine levels are tightly regulated in the nervous system. We pursued two of our hits that reduced brain dopamine levels, clueless and mask. Further examination suggests that mask regulates transcription of TH and affects dopamine-dependent sleep patterns. In summary, by studying genes that affect cuticle pigmentation, we were able to identify genes that affect dopamine metabolism as well as a novel regulator of behavior.

The sense of smell has potent effects on appetite, but the underlying neural pathways remain undefined. Here we investigated how olfactory signals reach two subsets of appetite-linked neurons in the hypothalamic arcuate nucleus: AgRP (agouti-related peptide) neurons, which stimulate appetite, and POMC (pro-opiomelanocortin) neurons, which suppress it. Using polysynaptic viral tracing, we show that AgRP and POMC neurons receive indirect input from partially overlapping but distinct areas of the olfactory cortex, indicating that they process different sets of olfactory information. We also identify different complements of neurons directly upstream of AgRP and POMC neurons that can relay olfactory cortical signals to the appetite neurons. Single cell transcriptomics shows heterogeneous expression of neuromodulator receptors among AgRP neurons, suggesting variations in the signals they receive. Integrated viral tracing and RNA localization further reveals selected brain areas where upstream neurons express cognate receptor ligands. Together, these findings outline multiple pathways by which distinct olfactory and modulatory signals are differentially routed to neurons that promote versus inhibit appetite.

Quantitative mass spectrometry (MS)-based proteomics has become a streamlined technology with a wide range of usage. Many emerging applications, such as single-cell proteomics, spatial proteomics of tissue sections and the profiling of low-abundant posttranslational modifications, require the analysis of minimal sample amounts and are thus constrained by the sensitivity of the workflow. Here, we present Slice-PASEF, a mass spectrometry technology that leverages trapped ion mobility separation of ions to attain the theoretical maximum of tandem MS sensitivity. We implement Slice-PASEF using a new module in our DIA-NN software and show that Slice-PASEF uniquely enables precise quantitative proteomics of low sample amounts. We further demonstrate its utility towards a range of applications, including single cell proteomics and degrader drug screens via ubiquitinomics.