Biology subjects

Bao, P.

Publications and source records attributed to Bao, P..

18 recordsLinked to original sources

bioRxiv2026

Neural Representations of Visual Anagrams in Macaque Inferotemporal Cortex

Visual recognition requires stable representations of object identity despite large changes in the retinal image. Yet this stability must be selective: some transformations change not only how an image looks, but also what object it appears to depict. Visual anagrams highlight this challenge: a single image can resemble two different objects at different orientations, such as an elephant and a rabbit. Here, we asked how macaque inferotemporal (IT) cortex represents such transformations. We recorded large neuronal populations while presenting visual anagrams and unambiguous exemplars of their two interpretations across rotations. Neurons preferring one exemplar tended to prefer anagram orientations supporting the corresponding interpretation. At each of the two interpretation-relevant canonical angles, population responses more closely resembled those to the matching exemplar than those to the nonmatching exemplar. This matching advantage persisted when both exemplars were presented at the alternative canonical angle. Decoders trained exclusively on exemplars across rotations also tracked the shifts in anagram representations. Visual models showed anagram-exemplar alignment as well, with stronger alignment in models with greater relative shape preference. These findings suggest that transformation tolerance in IT is sensitive to what an image appears to depict, while relative shape preference is associated with similar representational organization in visual models.

neuroscience↗

bioRxiv2026

Shared and idiosyncratic coding regimes coexist in macaque IT

Population-level measurements portray object representations in primate inferotemporal cortex (IT) as smooth, low-dimensional, and predictable by deep neural networks (DNNs). However, it remains unclear whether this structured population-level picture is representative of the full diversity of its constituent neurons. Here, we used Neuropixels 2.0 probes to record large populations of well-isolated units from an fMRI-localized face patch in macaque anterior IT while monkeys viewed more than 3,000 natural images. Single-unit responses were substantially sparser and more heterogeneous than multi-unit activity (MUA). Sparse neurons collectively formed a higher-dimensional code, supported efficient image identification, and responded later than broadly tuned neurons. Partly independently of sparseness, some neurons exhibited reliable feature randomness: their stimulus preferences were reproducible across repeated presentations but discontinuous across DNN feature spaces, rather than reflecting trial-to-trial response variability or noise. These neurons were virtually uncorrelated with the surrounding population despite being located within the same face patch. By contrast, MUA responses were denser, more correlated, lower-dimensional, and better predicted by DNNs. Together, these findings suggest a dual coding architecture in anterior IT: shared low-dimensional structure supports category generalization, whereas sparse and reliably feature-random single-neuron responses expand the representational space and support efficient identification of individual visual inputs. Response sparseness and reliable feature randomness may therefore constitute fundamental computational resources for object recognition.

neuroscience↗

bioRxiv2026

Period five indicates autopoiesis

A key scientific challenge is to develop a universal theory of life that integrates our biological knowledge with fundamental logical principles. We propose that a "five nodes" principle may govern the origin of life and consistently exist hierarchically within living systems. In our investigation, we explored autocatalytic chemical reaction networks (CRNs) as potential origins for methanotrophy and anoxygenic phototrophy, aiming to validate the "five nodes" principle in the emergence of autopoietic systems. Our research revealed the emergence of autocatalytic peptides and weakly reversible realizations within the MSA reaction network (composed of CH4, SO42-/SO32-, and NH4+) as well as in the light-Sammox (sulfurous reduction coupled to anaerobic ammonium oxidation)-driven CRN (composed of HCO3-, SO32-, and NH4+) under hydrothermal conditions. Furthermore, we identified the possible emergence of three main interdependent components essential for life within the two reaction networks: membrane compartments, peptide nucleic acids (PNA) backbones, and catalytic capacities for energy release reactions. Our findings suggest that non-equilibrium synergy of five bioessential elements (NESFBE) can facilitate proto-energy and material metabolism, thereby enabling diverse scenarios for lifes origin. Importantly, our discovery indicates linear constraints present in these CRNs that contributed to lifes inception, which is the mathematical foundation of the "five nodes" principle. Linear constraints determine both the emergence and self-disintegration of autopoietic systems. We infer a period five existence based on hierarchical structures found within autopoietic systems, and "period five indicates autopoiesis" could be one of universal theory of life.

biophysics↗

bioRxiv2025

Spatial Reorganization of Object Representations in High-Level Visual Cortex Distinguishes Working Memory from Perception

The human visual system balances veridical object visual perception with flexible object visual working memory (VWM), both relying on high-level visual regions. However, how these competing demands shape spatial representations remains unclear. Here, we ask whether VWM inherits the spatial constraints observed in the lateral occipital complex (LOC) during perception, or instead reorganizes these representations to meet mnemonic demands. Using matched bilateral presentation paradigms and fMRI-based decoding, we systematically compared spatial representations during perception and VWM. This approach revealed a striking dissociation: during perception, object information is largely confined to the contralateral LOC, whereas during VWM, robust ipsilateral representations emerge--even when both hemifields must be remembered. Vertex-ablation analyses revealed that VWM engages 70-90% of ipsilateral LOC territories, far exceeding those recruited during unilateral perception. Neither increased attentional span nor top-down feedback from association areas fully explained this expansion; rather, ipsilateral LOC patterns closely mirrored contralateral sensory representations, implicating interhemispheric coordination in VWM. Together, these findings demonstrate that object VWM flexibly recruits distributed high-level visual cortex, with spatial reorganization distinguishing mnemonic flexibility from perceptual fidelity.

neuroscience↗

bioRxiv2025

Triple-N Dataset: Non-human Primate Neural Responses to Natural Scenes

Understanding the neural mechanisms of visual perception requires data that encompass both large-scale cortical activity and the fine-grained dynamics of individual neurons. While the Natural Scenes Dataset (NSD) has provided substantial insights into visual processing in humans (Allen et al., 2022), its reliance on functional magnetic resonance imaging (fMRI) limits the exploration of individual neuron contributions. To bridge this gap, we present a new dataset: the Triple-N (Non-human Primate Neural Responses to Natural Scenes) dataset, which extends the NSD framework to non-human primates, incorporating single-neuron activity and fMRI recorded from the inferotemporal (IT) cortex. Using Neuropixels probes, we recorded neural responses from five macaques as they passively viewed 1,000 shared NSD images. The dataset includes 59 sessions across 27 sub-regions, capturing over 30,000 visual responsive units. Many recordings were obtained from fMRI-defined category-selective regions, such as face-, body-, scene-, and color-selective areas. Our dataset enables in-depth exploration of neural responses at multiple levels - from population dynamics to single-neuron activity, providing new insights into various aspects of visual processing, including the heterogeneity of object selectivity within functional regions and the temporal dynamics of responses to natural images. Furthermore, the dataset enables joint cross-species analyses, by integrating the macaque neural recordings with human fMRI data, offering a framework for comparing and aligning visual representations across primate species. Overall, our dataset provides a valuable resource for advancing our understanding of visual perception, bridging the gap between large-scale neuroimaging and fine-grained electrophysiological signals, while also facilitating the development of computational models of the high-level visual system.

neuroscience↗

bioRxiv2025

Ether Rhodamines with Enhanced Hydrophilicity, Fluorogenicity, and Brightness for Super-Resolution Imaging

Rhodamine dyes are widely used fluorophores in super-resolution fluorescence imaging due to their exceptional optical properties and "aggregation-disaggregation" induced fluorogenic activation. However, their excessive lipophilicity often reduces brightness in aqueous environments and causes off-target staining, limiting their effectiveness in high-resolution imaging. To address these challenges, we introduce an ether-decorated N-terminal modification strategy for rhodamine and silicon-rhodamine (Si-rhodamine), replacing conventional N-alkyl groups. The ether chains enhance water solubility, decrease aggregate size, and improve fluorogenicity across a wide concentration range. Their flexible, hydrophilic structure forms a protective shield around the xanthene core, minimizing dye-water interactions and reducing fluorescence quenching. Additionally, the inductive effect of the ether chains decreases the electron-donating strength of the amino groups, suppressing quenching caused by twisted intramolecular charge transfer (TICT). These modifications collectively increase the quantum yields of ER and ESiR in water from 0.35 and 0.19 (for tetraethyl-substituted analogs) to 0.70 and 0.41, respectively. Probes derived from ER and ESiR exhibit outstanding fluorogenicity, enhanced signal-to-noise ratios, and improved resolution in complex aqueous environments, demonstrating superior performance in advanced super-resolution imaging techniques such as structured illumination microscopy (SIM), stimulated emission depletion (STED) microscopy, and single-molecule localization microscopy (SMLM). This work introduces an innovative strategy for fluorophore design, offering significant advancements for super-resolution imaging applications.

biochemistry↗

bioRxiv2024

Versatile Stain Transfer in Histopathology Using a Unified Diffusion Framework

Histological staining is vital in clinical pathology for visualizing tissue structures. However, these techniques are laborious and time-consuming. Digital virtual staining offers a promising solution, but existing methods typically rely on Generative Adversarial Networks (GANs), which may suffer from artifacts and mode collapse. Motivated by the success of diffusion models, we present DUST, a novel Diffusion-based Unified framework for versatile Stain Transfer in histopathology. To enhance domain awareness and task-specific performance, we propose a dual encoding strategy that integrates the stain types of both the source and target domains. Additionally, we introduce a dynamic dual-output head to address the unstable intensity issue encountered with conventional DDPM implementations. Validated on a curated fourstain kidney histopathological dataset (H&E, MT, PAS, and PASM), DUST demonstrates superior versatile stain transfer capabilities. Our research highlights the potential of diffusion models to advance virtual staining, paving the way for more efficient digital pathology analyses.

pathology↗

bioRxiv2024

RNA-ligand interaction scoring via data perturbation and augmentation modeling

RNA-targeting drug discovery is undergoing an unprecedented revolution. Despite recent advances in this field, developing data-driven deep learning models remains challenging due to the limited availability of validated RNA-small molecule interactions and the scarcity of known RNA structures. In this context, we introduce RNAsmol, a novel sequence-based deep learning framework that incorporates data perturbation with augmentation, graph-based molecular feature representation and attention-based feature fusion modules to predict RNA-small molecule interactions. RNAsmol employs perturbation strategies to balance the bias between true negative and unknown interaction space thereby elucidating the intrinsic binding patterns between RNA and small molecules. The resulting model demonstrates accurate predictions of the binding between RNA and small molecules, outperforming other methods with average improvements of [~]8% (AUROC) in 10-fold cross-validation, [~]16% (AUROC) in cold evaluation (on unseen datasets), and [~]30% (ranking score) in decoy evaluation. Moreover, we use case studies to validate molecular binding hotspots in the prediction of RNAsmol, proving the models interpretability. In particular, we demonstrate that RNAsmol, without requiring structural input, can generate reliable predictions and be adapted to many RNA-targeting drug design scenarios.

bioinformatics↗

bioRxiv2024

De novo multi-mechanism antimicrobial peptide design via multimodal deep learning

Artificial intelligence (AI)-driven discovery of antimicrobial peptides (AMPs) is yet to fully utilise their three-dimensional (3D) structural characteristics, microbial specie-specific antimicrobial activities and mechanisms. Here, we constructed a QLAPD database comprising the sequence, structures and antimicrobial properties of 12,914 AMPs. QLAPD underlies a multimodal, multitask, multilabel, and conditionally controlled AMP discovery (M3-CAD) pipeline, which is proposed for the de novo design of multi-mechanism AMPs to combat multidrug-resistant organisms (MDROs). This pipeline integrates the generation, regression, and classification modules, using a innovative 3D voxel coloring method to capture the nuanced physicochemical context of amino acids, significantly enhancing structural characterizations. QL-AMP-1, discovered by M3-CAD, which possesses four antimicrobial mechanisms, exhibited low toxicity and significant activity against MDROs. The skin wound infection model demonstrates its considerable antimicrobial effects and negligible toxicity. Altogether, integrating 3D features, specie-specific antimicrobial activities and mechanisms enhanced AI-driven AMP discovery, making the M3-CAD pipeline a viable tool for de novo AMP design.

bioengineering↗

bioRxiv2023

Pathformer: biological pathway informed Transformer model integrating multi-modal data of cancer

Multi-omics data provide a comprehensive view of gene regulation at multiple levels, which is helpful in achieving accurate diagnosis of complex diseases like cancer. To integrate various multi-omics data of tissue and liquid biopsies for disease diagnosis and prognosis, we developed a biological pathway informed Transformer, Pathformer. It embeds multi-omics input with a compacted multi-modal vector and a pathway-based sparse neural network. Pathformer also leverages criss-cross attention mechanism to capture the crosstalk between different pathways and modalities. We first benchmarked Pathformer with 18 comparable methods on multiple cancer datasets, where Pathformer outperformed all the other methods, with an average improvement of 6.3%-14.7% in F1 score for cancer survival prediction and 5.1%-12% for cancer stage prediction. Subsequently, for cancer prognosis prediction based on tissue multi-omics data, we used a case study to demonstrate the biological interpretability of Pathformer by identifying key pathways and their biological crosstalk. Then, for cancer early diagnosis based on liquid biopsy data, we used plasma and platelet datasets to demonstrate Pathformers potential of clinical applications in cancer screen. Moreover, we revealed deregulation of interesting pathways (e.g., scavenger receptor pathway) and their crosstalk in cancer patients blood, providing new candidate targets for cancer microenvironment study.

bioinformatics↗

bioRxiv2023

Cell-free multi-omics analysis reveals tumor status-informative signatures in gastrointestinal cancer patients' plasma

During cancer development, hosts tumorigenesis and immune signals are released to and informed by circulating molecules, like cell-free DNA (cfDNA) and RNA (cfRNA) in blood. However, these two kinds of molecules are still not systematically compared in gastrointestinal cancer. Here, we profiled 4 types of cell-free omics data from colorectal and stomach cancer patients, and assayed 15 types of genomic, epi-genomic, and transcriptomic variations. First, we demonstrated that the multi-omics data were more capable of detecting cancer genes than the single-omics data, where cfRNAs were more sensitive and informative than cfDNAs in terms of detection ratio, variation type, altered number, and enriched functional pathway. Moreover, we revealed several peripheral immune signatures that were suppressed in cancer patients and originated from specific circulating and tumor-microenvironment cells. Particularly, we defined a {gamma}{delta}-T-cell score and a cancer-associated-fibroblast (CAF) score using the cfRNA-seq data of 143 cancer patients. They were informative of clinical status like cancer stage, tumor size, and survival. In summary, our work reveals the cell-free multi-molecular landscape of colorectal and stomach cancer, and provides a potential monitoring utility in blood for the personalized cancer treatment.

systems biology↗

bioRxiv2021

Principles governing the topological organization of object selectivities in ventral temporal cortex

To achieve the computational goal of rapidly recognizing miscellaneous objects in the environment despite large variations in their appearance, our mind represents objects in a high-dimensional object space to provide separable category information and enable the extraction of different kinds of information necessary for various levels of the visual processing. To implement this abstract and complex object space, the ventral temporal cortex (VTC) develops different object-selective regions with certain topological organization as the physical substrate. However, the principle that governs the topological organization of object selectivities in the VTC remains unclear. Here, equipped with the wiring cost minimization principle constrained by the wiring length of neurons in human temporal lobe, we constructed a hybrid self-organizing map (SOM) model as an artificial VTC (VTC-SOM) to explain how the abstract and complex object space is faithfully implemented in the brain. In two in silico experiments with the empirical brain imaging and single-unit data, our VTC-SOM predicted the topological structure of fine-scale functional regions (face-, object-, body-, and place-selective regions) and the boundary (i.e., middle Fusiform Sulcus) in large-scale abstract functional maps (animate vs. inanimate, real-word large-size vs. small-size, central vs. peripheral), with no significant loss in functionality (e.g., categorical selectivity, a hierarchy of view-invariant representations). These findings illustrated that the simple principle utilized in our model, rather than multiple hypotheses such as temporal associations, conceptual knowledge, and computational demands together, was apparently sufficient to determine the topological organization of object-selectivities in the VTC. In this way, the high-dimensional object space is implemented in a two-dimensional cortical surface of the brain faithfully.

neuroscience↗

bioRxiv2020

Jejunal barrier dysfunction of Datong Yak with high proportion concentrate dietary during the cold season

The current study aimed to investigate the damage of long-term high concentrate diet feeding pattern on Yak jejunal structure, physiological function and protein composition during cold season. Twelve Datong male Yak (Bos grunniens) with the same age from cold season were randomly selected and slaughtered to determine Yak jejunal digestive enzyme activity, morphology and protein composition on different feeding patterns in Tibetan Plateau. The results showed that Yak jejunum digestive enzyme activity and morphology of grazing reared group were better than those in the intensively reared group. A total of 96 differentially expressed proteins were identified by label-free Mass Spectrometry (MS), which could be concluded to two predominant themes: protein structure and inflammatory response. Nine differentially expressed proteins were correlated in Yak jejunum damage in different feeding patterns. According to our study, feeding patterns resulted in the differences in Yak jejunum physiological function, morphology and protein composition. Our results confirmed that feeding of long-term high concentrate diet could damage the jejunum epithelial morphology and function.

genetics↗

bioRxiv2019

Comparison of different orders of Legendre polynomials in random regression model for estimation of genetic parameters and breeding values of milk yield in the Chinese Holstein population

Random regression test-day model has become the most commonly adopted model for routine genetic evaluations for different dairy populations, which allows accurately accounting for genetic and environmental effects at different periods during lactation. The objective of this study was to explore appropriate random regression test-day model for genetic evaluation of milk yield in Chinese Holstein population. Data included 419,567 test-day records from 54,417 cows in the first lactation. Variance components and breeding values were estimated using random regression test-day model with different order (first order to fifth order) of Legendre polynomials, and accounted for homogeneous or heterogeneous residual variance across the lactation. The goodness of fit of the models was evaluated by total residual variance (TRV) and - 2logL. Further, the predictive ability of the models was assessed by Spearmans rank correlation between estimated breeding values for 305d milk yield (EBV305) from the full data set and reduced data set in which the records from the last calving year were masked. The results showed that random regression models using third order Legendre polynomials (LP3) with heterogeneous residual variance achieved the lower TRV and - 2logL value and the highest correlation for EBV305 between full data and reduced data. Heritability estimated by this model was 0.250 for 305d milk yield and ranged from 0.163 to 0.304 for test-day milk yield. We suggest random regression model with Legendre polynomial of order 3 and accounting for heterogeneous residual variances could be an appropriate model to be used for genetic evaluation of milk yield for Chinese Holstein population.

genetics↗

bioRxiv2018

Sulfur reduction coupling with anaerobic ammonium oxidation power the origin of life

The chemical reaction networks (CRNs), which led to the transition on early Earth from geochemistry to biochemistry remain unknown. We show that under mild hydrothermal circumstances, a thermodynamic chemical reaction network including sulfite/sulfate coupled with anaerobic ammonium oxidation (Sammox), might have driven prebiotic peptides synthesis. Peptides comprise 14 proteinogenic amino acids, endowed Sammox-driven CRNs with autocatalysis. The peptides exhibit both forward and reverse catalysis, with the opposite catalytic impact in sulfite- and sulfate-fueled Sammox-driven CRNs, respectively, at both a variable temperature range and a fixed temperature, resulting in seesaw-like catalytic properties. The ratio of sulfite to sulfate switches the catalytic orientation of peptides, resulting in Sammox-driven CRNs that has both anabolic and catabolic reactions at all times. Furthermore, peptides produced from sulfite-fueled Sammox-driven CRNs could catalyze both sulfite-fueled Sammox and Anammox (nitrite reduction coupled with anaerobic ammonium oxidation) reactions. We propose that Sammox-driven CRNs were critical in the creation of life and that Anammox microorganisms that have both Sammox functions are direct descendants of Sammox-driven CRNs.

microbiology↗

bioRxiv2018

Representation of multiple objects in macaque category-selective areas

Object recognition in the natural world usually occurs in the presence of multiple surrounding objects, but responses of neurons in inferotemporal (IT) cortex, the large brain area responsible for object recognition, have mostly been studied only to isolated objects. We study rules governing responses to multiple objects by cells in two category-selective regions of macaque IT cortex, the middle lateral face patch (ML) and the middle body patch (MB). We find that responses of single ML and MB cells to pairs of objects could be explained by the widely-accepted framework of normalization, with one added ingredient: homogeneous category selectivity of neighboring neurons forming the normalization pool. This rule leads to winner-take-all, contralateral-take-all, or weighted averaging behavior in single cells, depending on the category, spatial configuration, and relative contrast of the two objects. The winner-take-all behavior suggests a potential mechanism for clutter-invariant representation of face and bodies under certain conditions.

neuroscience↗

bioRxiv2017

The representation of colored objects in macaque color patches

An important question about color vision is: how does the brain represent the color of an object? The recent discovery of \"color patches\" in macaque inferotemporal (IT) cortex, the part of brain responsible for object recognition, makes this problem experimentally tractable. Here we record neurons in three color patches, middle color patch CLC (central lateral color patch), and two anterior color patches ALC (anterior lateral color patch) and AMC (anterior medial color patch), while presenting images of objects systematically varied in hue. We found that all three patches contain high concentrations of hue-selective cells, and the three patches use distinct computational strategies to represent colored objects: while all three patches multiplex hue and shape information, shape-invariant hue information is much stronger in anterior color patches ALC/AMC than CLC; furthermore, hue and object shape specifically for primate faces/bodies are over-represented in AMC but not in the other two patches.

neuroscience↗

bioRxiv2017

Direct Conversion Of Human Fibroblasts Into Osteoblasts And Osteocytes With Small Molecules And A Single Factor, Runx2

Human osteoblasts can be induced from somatic cells by introducing defined factors, however, the strategy limits cells therapeutic applications for its multi-factor and complicated genetic manipulations that may bring uncertainty into the genome. Another important cell type in bone metabolism, osteocytes, which play a central role in regulating the dynamic nature of bone in all its diverse functions, have not been obtained from transdifferetiation so far. Herein, we have established procedures to convert human fibroblast directly into osteocyte-like and osteoblast-like cells using a single transcription factor, Runx2 and chemical cocktails by activating Wnt and cAMP/PKA pathways. These induced osteoblast-like cells express osteogenic markers and generate mineralized nodule deposition. A good performance of bone formation from these cells was observed in subcutaneous site of mouse at 4 weeks post-transplantation. Moreover, further studies convert human fibroblasts into osteocyte-like cells by orchestrating timing of the aforementioned chemical cocktails exposure. These osteocyte-like cells express osteocyte-specific markers and display characteristic morphology features of osteocytes. In summary, this study provides a promising strategy for cell-based therapy in bone regenerative medicine by direct reprogramming of fibroblasts into osteocytes and osteoblasts.

cell biology↗