Biology subjects

Cheng, S.

Publications and source records attributed to Cheng, S..

At least 91 records · Page 5Linked to original sources

bioRxiv2024

Genomic and Genetic Insights into Mendel's Pea Genes

Pea, Pisum sativum, is an excellent model system through which Gregor Mendel established the foundational principles of inheritance. Surprisingly, till today, the molecular nature of the genetic differences underlying the seven pairs of contrasting traits that Mendel studied in detail remains partially understood. Here, we present a genomic and phenotypic variation map, coupled with haplotype-phenotype association analyses across a wide range of traits in a global Pisum diversity panel. We focus on a genomics-enabled genetic dissection of each of the seven traits Mendel studied, revealing many previously undescribed alleles for the four characterized genes, R, Le, I and A, and elucidating the gene identities and mutations for the remaining three uncharacterized traits. Notably, we identify: (1) a ca. 100kb deletion upstream of the Chlorophyll synthase (ChlG) gene, which generates aberrant transcripts and confers the yellow pod phenotype of gp mutants; (2) an in-frame premature stop codon mutation in a Dodeca-CLE41/44 signalling peptide which explains the parchmentless mutant phenotype corresponding to p; and (3) a 5bp in-frame deletion in a CIK-like receptor kinase gene corresponding to the fasciated stem phenotype fa, which Mendel described in terms of flower position, and we postulate the existence of a Modifier of fa (Mfa) locus that masks this meristem defect. Mendel noted the pleiotropy of the a mutation, including inhibition of axil ring anthocyanin pigmentation, a trait we found to be controlled by allelic variants of the gene D within an R2R3-MYB gene cluster. Furthermore, we characterize and validate natural variation of a quantitative genetic locus governing both pod width and seed weight, characters that Mendel deemed were not sufficiently demarcated for his analyses. This study establishes a cornerstone for fundamental research, education in biology and genetics, and pea breeding practices.

genetics↗

bioRxiv2024

Spatiotemporally resolved transcriptome atlas of developing mouse placenta reveals mechanisms of embryonic lethality.

Proper placentation is fundamental to the growth and viability of the maturing embryo. However, a thorough grasp of the spatial organization of cell types, their interactions, and gene expression patterns along the maternal-fetal function during placental development remains incomplete. Here, we utilized Stereo-seq to construct the spatiotemporal transcriptomic atlas of the mouse placenta (STAMP) spanning embryonic (E) days 9.5 to 18.5 at single cell resolution (https://db.cngb.org/stomics/stamp/). This resource delineates spatially resolved cellular dynamics and gene expression patterns across placental and maternal compartments. We identified distinct glycogen trophoblast cell (GC) subclusters, mapped their developmental trajectories, and uncovered transcriptional transitions accompanying their migration from the junctional zone to the maternal decidua from E12.5 onward. In a defective placentation model with perinatal lethality, GCs abnormally persisted at E18.5 with excessive glycogen and reduced degradation metabolites in placenta and fetal liver, indicating impaired glycogen breakdown. Maternal glucose supplementation restored glucose levels and rescued fetal survival, underscoring GC-mediated metabolic support as critical for viability. Together, this study provides a comprehensive spatiotemporal placental atlas and demonstrates that GC-mediated glycogen metabolism is essential for sustaining fetal viability.

developmental biology↗

bioRxiv2024

The cost of behavioral flexibility: reversal learning driven by a spiking neural network

To survive in a changing world, animals often need to suppress an obsolete behavior and acquire a new one. This process is known as reversal learning (RL). The neural mechanisms underlying RL in spatial navigation have received limited attention and it remains unclear what neural mechanisms maintain behavioral flexibility. We extended an existing closed-loop simulator of spatial navigation and learning, based on spiking neural networks [8]. The activity of place cells and boundary cells were fed as inputs to action selection neurons, which drove the movement of the agent. When the agent reached the goal, behavior was reinforced with spike-timing-dependent plasticity (STDP) coupled with an eligibility trace which marks synaptic connections for future reward-based updates. The modeled RL task had an ABA design, where the goal was switched between two locations A and B every 10 trials. Agents using symmetric STDP excel initially on finding target A, but fail to find target B after the goal switch, persevering on target A. Using asymmetric STDP, using many small place fields, and injecting short noise pulses to action selection neurons were effective in driving spatial exploration in the absence of rewards, which ultimately led to finding target B. However, this flexibility came at the price of slower learning and lower performance. Our work shows three examples of neural mechanisms that achieve flexibility at the behavioral level, each with different characteristic costs.

neuroscience↗

bioRxiv2024

Competition and integration of visual and goal vector signals for spatial navigation

AO_SCPLOWBSTRACTC_SCPLOWIntegrating different sources of information is essential to successful spatial navigation. For instance, animals must often rely on a combination of visual impressions, self-motion, olfaction, and other signals to navigate to a goal. This is especially important when navigating in uncertain environments, where switching from one source of information to another or integrating multiple sources of information may be required to make navigation decisions. We propose a computational model of the interaction of visual and goal-vector signals based on reinforcement learning and use it to study behavior and spatial representations. Our model demonstrates that the ability to navigate using each information source independently, in addition to integrating them, is crucial to successfully navigating in uncertain environments. Counterintuitively, our model also shows that when one of the signals is removed, navigation may be improved if the remaining signal is reliable and sufficient to navigate, however, this improvement comes at the expense of robustness.

neuroscience↗

bioRxiv2024

Accurate estimation of SNP genotypes and genetic relatedness from DNA methylation data

Epigenome-wide association studies (EWAS) are susceptible to widespread confounding caused by population structure and genetic relatedness. Nevertheless, kinship estimation is challenging in EWAS without genotyping data. We propose MethylGenotyper, a method that for the first time enables accurate genotyping at thousands of SNPs directly from commercial DNA methylation microarrays. We model the intensities of methylation probes near SNPs with a mixture of three beta distributions corresponding to different genotypes and estimate parameters with an expectation-maximization algorithm. We conduct extensive simulations to demonstrate the performance of the method. When applying MethylGenotyper to Infinium EPIC array data of 4,662 Chinese, we obtain genotypes at 4,319 SNPs with a concordance rate of 98.26%, enabling the identification of 255 pairs of close relatedness. Furthermore, we show that MethylGenotyper allows for the estimation of both population structure and cryptic relatedness among 702 Australians of diverse ancestry. We have implemented MethylGenotyper in a publicly available R package to facilitate future large-scale EWAS.

bioinformatics↗

bioRxiv2024

Paradigm shift in biomarker translation: a pipeline to generate clinical grade biomarker candidates from DIA-MS discovery

Clinical biomarker development has been stymied by inaccurate protein quantification from mass spectrometry (MS) discovery data and a prolonged validation process. To mitigate these issues, we created the Targeted Extraction Assessment of Quantification (TEAQ) software package. This innovative tool uses the discovery cohort analysis to select precursors, peptides, and proteins that adhere to established targeted assay criteria. TEAQ was applied to Data-Independent Acquisition MS data from plasma samples acquired on an Orbitrap Astral MS. Identified precursors were evaluated for linearity, specificity, repeatability, reproducibility, and intra-protein correlation from 11-point loading curves under three throughputs, to develop a resource for clinical-grade targeted assays. From a clinical cohort of individuals with inflammatory bowel disease (n=492), TEAQ successfully identified 1116 signature peptides for 327 quantifiable proteins from 1180 identified proteins. Embedding stringent selection criteria adaptable to targeted assay development into the analysis of discovery data will streamline the transition to validation and clinical studies.

biochemistry↗

bioRxiv2024

Unraveling the Time-Frequency Features of Emotional Regulation: Based on an Interpretable XGBoost-SHAP Analytical Framework

Negative emotions, while crucial for survival, can lead to adverse health effects if not managed properly. Our understanding of temporal EEG changes during emotion regulation is limited. To address this gap, this study employs interpretable machine learning techniques, XGBoost-SHAP model, to analyze EEG data. This study investigates the neural mechanisms underlying emotion regulation, with a focus on EEG oscillations in the lateral prefrontal area channels (F3, F4, F7, F8) across four specific frequency bands (Alpha, Beta, Theta, Delta). By identifying predictive features and patterns, this approach offers insights into the temporal dynamics of emotion regulation and the involvement of specific brain regions, enhancing our understanding of emotional processing and providing avenues for effective interventions. The findings reveal a significant relationship between specific EEG feature changes and emotional ratings during the emotion regulation process. The LPFC emerges as central in cognitive control and emotional regulation. These results highlight the LPFCs rapid and effective role in regulating complex emotional dynamics, crucial for understanding and treating emotional disorders. The study underscores the importance of machine learning in elucidating neural mechanisms and guiding personalized interventions for emotional well-being.

neuroscience↗

bioRxiv2024

Genetic regulation and targeted reversal of lysosomal dysfunction and inflammatory sterol metabolism in pulmonary arterial hypertension

Vascular inflammation critically regulates endothelial cell (EC) pathophenotypes, particularly in pulmonary arterial hypertension (PAH). Dysregulation of lysosomal activity and cholesterol metabolism have known inflammatory roles in disease, but their relevance to PAH is unclear. In human pulmonary arterial ECs and in PAH, we found that inflammatory cytokine induction of the nuclear receptor coactivator 7 (NCOA7) both preserved lysosomal acidification and served as a homeostatic brake to constrain EC immunoactivation. Conversely, NCOA7 deficiency promoted lysosomal dysfunction and proinflammatory oxysterol/bile acid generation that, in turn, contributed to EC pathophenotypes. In vivo, mice deficient for Ncoa7 or exposed to the inflammatory bile acid 7-hydroxy-3-oxo-4-cholestenoic acid (7HOCA) displayed worsened PAH. Emphasizing this mechanism in human PAH, an unbiased, metabolome-wide association study (N=2,756) identified a plasma signature of the same NCOA7-dependent oxysterols/bile acids associated with PAH mortality (P<1.1x10-6). Supporting a genetic predisposition to NCOA7 deficiency, in genome-edited, stem cell-derived ECs, the common variant intronic SNP rs11154337 in NCOA7 regulated NCOA7 expression, lysosomal activity, oxysterol/bile acid production, and EC immunoactivation. Correspondingly, SNP rs11154337 was associated with PAH severity via six-minute walk distance and mortality in discovery (N=93, P=0.0250; HR=0.44, 95% CI [0.21-0.90]) and validation (N=630, P=2x10-4; HR=0.49, 95% CI [0.34-0.71]) cohorts. Finally, utilizing computational modeling of small molecule binding to NCOA7, we predicted and synthesized a novel activator of NCOA7 that prevented EC immunoactivation and reversed indices of rodent PAH. In summary, we have established a genetic and metabolic paradigm and a novel therapeutic agent that links lysosomal biology as well as oxysterol and bile acid processes to EC inflammation and PAH pathobiology. This paradigm carries broad implications for diagnostic and therapeutic development in PAH and in other conditions dependent upon acquired and innate immune regulation of vascular disease. One Sentence SummaryPulmonary arterial hypertension pathophenotypes arise from allele-specific NCOA7 regulation of lysosome function and inflammatory oxysterol generation, as demonstrated by genomic and metabolomic association studies coupled with genetic and pharmacologic mechanistic evidence. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=112 SRC="FIGDIR/small/582142v1_ufig1.gif" ALT="Figure 1"> View larger version (39K): org.highwire.dtl.DTLVardef@896f6forg.highwire.dtl.DTLVardef@f31c2eorg.highwire.dtl.DTLVardef@bf46dcorg.highwire.dtl.DTLVardef@1c55686_HPS_FORMAT_FIGEXP M_FIG Graphical abstract. C_FIG

physiology↗

bioRxiv2024

Multimodal illumination platform for 3D single-molecule super-resolution imaging throughout mammalian cells

Single-molecule super-resolution imaging is instrumental for investigating cellular architecture and organization at the nanoscale. Achieving precise 3D nanometric localization when imaging structures throughout mammalian cells, which can be multiple microns thick, requires careful selection of the illumination scheme in order to optimize the fluorescence signal to background ratio (SBR). Thus, an optical platform that combines different wide-field illumination schemes for target-specific SBR optimization would facilitate more precise, 3D nanoscale studies of a wide range of cellular structures. Here we demonstrate a versatile multimodal illumination platform that integrates the sectioning and background reduction capabilities of light sheet illumination with homogeneous, flat-field epi-and TIRF illumination. Using primarily commercially available parts, we combine the fast and convenient switching between illumination modalities with point spread function engineering to enable 3D singlemolecule super-resolution imaging throughout mammalian cells. For targets directly at the coverslip, the homogenous intensity profile and excellent sectioning of our flat-field TIRF illumination scheme improves single-molecule data quality by providing low fluorescence background and uniform fluorophore blinking kinetics, fluorescence signal, and localization precision across the entire field of view. The increased contrast achieved with LS illumination, when compared with epi-illumination, makes this illumination modality an excellent alternative when imaging targets that extend throughout the cell. We validate our microscopy platform for improved 3D super-resolution imaging by two-color imaging of paxillin - a protein located in the focal adhesion complex - and actin in human osteosarcoma cells.

biophysics↗

bioRxiv2024

Distinct Sequential Effects in Space and Time: Repulsion in Direction Judgments and Attraction in Time Reproduction

Our current perception and decision-making are shaped by recent experiences, a phenomenon known as serial dependence. While serial dependence is well-documented in visual perception and has been recently explored in time perception, their functional similarities across non-temporal and temporal domains remain elusive, particularly in relation to task relevance and working memory load. To address this, we designed a unified experimental paradigm using coherent motion stimuli to test both direction and time reproduction. The direction and time tasks were randomly mixed across trials. Additionally, we introduced pre-cue versus post-cue settings in separate experiments to manipulate working memory load during the encoding phase. We found attractive biases in time reproduction but repulsive biases in direction estimation. Notably, the temporal attraction was more pronounced when the preceding task was also time-related. In contrast, the direction repulsion remained unaffected by the nature of the preceding task. Additionally, both attractive and repulsive biases were enhanced by the post-cue compared to the pre-cue. Our findings suggest that opposing sequential effects in non-temporal and temporal domains may originate from different processing stages linked to sensory adaptation and post-perceptual processes involving working memory.

neuroscience↗

bioRxiv2024

Pulmonary primary oxysterol and bile acid synthesis as a predictor of outcomesin pulmonary arterial hypertension

Pulmonary arterial hypertension (PAH) is a rare and fatal vascular disease with heterogeneous clinical manifestations. To date, molecular determinants underlying the development of PAH and related outcomes remain poorly understood. Herein, we identify pulmonary primary oxysterol and bile acid synthesis (PPOBAS) as a previously unrecognized pathway central to PAH pathophysiology. Mass spectrometry analysis of 2,756 individuals across five independent studies revealed 51 distinct circulating metabolites that predicted PAH-related mortality and were enriched within the PPOBAS pathway. Across independent single-center PAH studies, PPOBAS pathway metabolites were also associated with multiple cardiopulmonary measures of PAH-specific pathophysiology. Furthermore, PPOBAS metabolites were found to be increased in human and rodent PAH lung tissue and specifically produced by pulmonary endothelial cells, consistent with pulmonary origin. Finally, a poly-metabolite risk score comprising 13 PPOBAS molecules was found to not only predict PAH-related mortality but also outperform current clinical risk scores. This work identifies PPOBAS as specifically altered within PAH and establishes needed prognostic biomarkers for guiding therapy in PAH. One-Sentence SummaryThis work identifies pulmonary primary oxysterol and bile acid synthesis as altered in pulmonary arterial hypertension, thus establishing a new prognostic test for this disease.

systems biology↗

bioRxiv2024

PCTA, A PAN-CANCER CELL LINE TRANSCRIPTOME ATLAS

BackgroundA substantial volume of RNA sequencing data were generated from cancer cell lines. However, it requires specific bioinformatics skills to compare gene expression levels across cell lines. This has hindered non-bioinformaticians from fully utilizing these valuable datasets in their research. To bridge this gap, we established a curated Pan-cancer Cell Line Transcriptome Atlas (PCTA) dataset. This resource aims to provide a user-friendly platform, allowing researchers without extensive bioinformatics expertise to access and leverage the wealth of information within the dataset for their studies. Importantly, PCTA stands out by offering sufficient sample numbers per cell line in comparison to other pan-cancer datasets. MethodsCell lines meta data and RNA sequencing data were retrieved from the Cancer Cell Line Encyclopedia (CCLE), SRA and ARCHS4 databases. Utilizing the programming language R, we conducted data retrieval, normalization, and visualization. Only expression data for protein-coding genes and long-non-coding RNAs (LncRNAs) were considered in this study, streamlining the focus to enhance the precision and relevance of the analysis. ResultsThe resulting PCTA dataset encompasses the expression matrix of 24,965 genes, featuring data from 84,385 samples derived from 5,677 studies. This comprehensive compilation spans 535 cell lines, representing a spectrum of 114 cancer types originating from 30 diverse tissue types. On UMAP plots, cell lines originating from the same type of tissue tend to cluster together, illustrating the datasets ability to capture biological relationships. To unravel molecular signatures, marker genes were identified for each cancer type. Additionally, an interactive and user-friendly web application (https://pcatools.shinyapps.io/PCTA_app/) was developed for researchers to explore the PCTA dataset. This platform allows users to examine the expression pattern of their genes of interest across a diverse array of samples. Data are visualized as violin-, box-, and point-plots, enhancing the interpretability of the findings. ConclusionThe PCTA stands as a comprehensive resource, offering insights into gene expression patterns across diverse cancer cell lines and providing a valuable tool to explore molecular signatures and potential therapeutic targets in cancer research.

cancer biology↗

bioRxiv2023

Transcriptomic analysis of melanoma cells reveals an association of α-synuclein with regulation of the inflammatory response.

The Parkinsons disease protein, alpha-synuclein (-syn/SNCA), is highly expressed in neurons and melanomas. The goal of this study was to reveal the mechanism(s) of -syns involvement in melanoma pathogenesis. To decipher the genes and pathways affected by -syn, we conducted an RNA sequencing analysis of human SK-MEL-28 cells and several SK-MEL-28 SNCA-KO clones. We identified 1098 significantly up-regulated genes and 660 significantly down-regulated genes. Several of the upregulated genes are related to the immune system, i.e., the inflammatory response and the matrisome. We validated five upregulated genes (IL-1y, SAA1, IGFBP5, CXCL8, and CXCL10) by RT-qPCR and detected IGFBP5 and IL-1y in spent media of control and SNCA-KO cells. The levels of each of these secreted proteins were significantly higher in the spent media of the SNCA-KO clones than control cells. We suggest that the loss of -syn expression unleashes chemokine/cytokine secretion, which could help melanoma cells evade the immune system.

cancer biology↗

bioRxiv2023

Utilizing Non-Invasive Prenatal Test Sequencing Data Resource for Human Genetic Investigation

Non-invasive prenatal testing (NIPT) employs ultra-low-pass sequencing of maternal plasma cell-free DNA to detect fetal trisomy. With exceptional sensitivity, specificity, and safety, NIPT has gained global adoption, exceeding ten million tests, establishing it as one of the largest human genetic resources. This resource holds immense potential for exploring population genetic variations and their correlations with phenotypes. Here, we present comprehensive methods tailored for analyzing large, low-depth NIPT genetic datasets, involving customized algorithms and software for genetic variation detection, genotype imputation, and genome-wide association analysis. Through evaluations, we demonstrate that, when integrated with appropriate probabilistic models and population-specific haplotype reference panels, accurate allele frequency estimation and high genotype imputation accuracy (0.8 to 0.9) are achievable for genetic variants with alternative allele frequencies between 0.01 and 0.05, at sequencing depths of 0.1x to 0.25x. Additionally, we attained an R-square exceeding 0.9 for estimating genetic effect sizes across various sequencing platforms. These findings establish a robust theoretical and practical foundation for leveraging NIPT data in advancing medical genetic studies, not only in realms of maternal and child health, but also for long-term health outcomes. HighlightsO_LIIntroduction of probabilistic model integration for analyzing large-scale, low-pass non-invasive prenatal test (NIPT) sequencing data C_LIO_LIEvaluation of protocols for variant detection, genotype imputation, and genome-wide association analyses with NIPT data C_LI

bioinformatics↗

bioRxiv2023

Golgi-associated retrograde protein (GARP) complex recruits retromer to trans-Golgi network for FgKex2 and FgSnc1 recycling, necessary for development and pathogenicity of Fusarium graminearum

In eukaryotic cells, the retromer complex plays a crucial role in orchestrating the sorting and retrograde transport of cargo proteins from endosomes to the trans-Golgi network (TGN). Despite its significance, the molecular details of this intracellular trafficking process remain unclear. Here, we identified a Golgi-associated retrograde protein (GARP) complex as a mediator of vesicular transport, facilitating the recruitment of the retromer complex to TGN to exert its functions. The GARP complex is mainly localized to the TGN, where it interacts with the retromer complex. This interaction is evolutionarily conserved across different species. Furthermore, we identified FgKex2 and Fgsnc1 as cargo proteins in the GARP/retromer-mediated recycling pathway. Loss of GARP or retromer results in a complete mis-sorting of FgKex2 and FgSnc1 to the vacuolar degradation pathway, which affects the growth, development, toxisome biogenesis, and pathogenicity of F. graminearum. In summary, we assert that GARP facilitates the recruitment of retromer from endosomes to the TGN, orchestrating the recycling of FgKex2 and FgSnc1. This is integral to sustaining continuous growth, development and contributes significantly to the pathogenicity of F. graminearum.

microbiology↗

bioRxiv2023

Hemodynamics regulate spatiotemporal artery muscularization in the developing circle of Willis

Vascular smooth muscle cells (VSMCs) envelop vertebrate brain arteries, playing a crucial role in regulating cerebral blood flow and neurovascular coupling. The dedifferentiation of VSMCs is implicated in cerebrovascular diseases and neurodegeneration. Despite its importance, the process of VSMC differentiation on brain arteries during development remains inadequately characterized. Understanding this process could aid in reprogramming and regenerating differentiated VSMCs in cerebrovascular diseases. In this study, we investigated VSMC differentiation on the zebrafish circle of Willis (CoW), comprising major arteries that supply blood to the vertebrate brain. We observed that the arterial expression of CoW endothelial cells (ECs) occurs after their migration from the cranial venous plexus to form CoW arteries. Subsequently, acta2+ VSMCs differentiate from pdgfrb+ mural cell progenitors upon recruitment to CoW arteries. The progression of VSMC differentiation exhibits a spatiotemporal pattern, advancing from anterior to posterior CoW arteries. Analysis of blood flow suggests that earlier VSMC differentiation in anterior CoW arteries correlates with higher red blood cell velocity wall shear stress. Furthermore, pulsatile blood flow is required for differentiation of human brain pdgfrb+ mural cells into VSMCs as well as VSMC differentiation on zebrafish CoW arteries. Consistently, the flow-responsive transcription factor klf2a is activated in ECs of CoW arteries prior to VSMC differentiation, and klf2a knockdown delays VSMC differentiation on anterior CoW arteries. In summary, our findings highlight the role of blood flow activation of endothelial klf2a as a mechanism regulating the initial VSMC differentiation on vertebrate brain arteries.

developmental biology↗

bioRxiv2023

Development of a Next Generation SNP Genotyping Array for Wheat

High throughput genotyping arrays have provided a cost effective, reliable and interoperable system for genotyping hexaploid wheat and its related germplasm pool. Existing, highly cited arrays including our 35K Axiom Wheat Breeders genotyping array and the Illumina 90K iSelect array were designed based on a limited amount of varietal sequence diversity and with imperfect knowledge of SNP positions. Recent progress in sequencing wheat varieties and landraces has given us access to a vast pool of SNP diversity, whilst technological improvements in array design has allowed us to fit significantly more probes onto a 384-well format Axiom array than was previously possible. Here we describe a novel High Density Axiom genotyping array, the Triticum aestivum Next Generation array (TaNG), largely derived from whole genome skim sequencing of 204 elite wheat lines and 111 wheat landraces taken from the Watkins "Core Collection". We use a novel "minimal marker" optimisation approach to select up to six SNPs in each 1.5 MB region of the wheat genome with the highest combined varietal discrimination potential. A design iteration step allowed us to test and replace skim-sequence derived SNPs which failed to convert to reliable Axiom markers, resulting in a final design, designated TaNG1.1 with 43,372 SNPs derived from a haplotype-optimised combination of novel SNPs, DArTAG-derived and legacy wheat Axiom markers. We show that this design has an even distribution of SNPs across chromosomes and sub-genomes compared to previous arrays and can be used to generate genetic maps with a significantly higher number of distinct bins than our previous Axiom array. We also demonstrate the improved performance of TaNG1.1 for Genome Wide Association Studies (GWAS) and its utility for Copy Number Variation (CNV) analysis. The array is commercially available, and the marker annotations, initial genotyping results and software used to generate the optimised marker sets are freely available.

genomics↗

bioRxiv2023

Global remapping emerges as the mechanism for renewal of context-dependent behavior in a reinforcement learning model

The hippocampal formation exhibits complex and context-dependent activity patterns and dynamics, e.g., place cell activity during spatial navigation in rodents or remapping of place fields when the animal switches between contexts. Furthermore, rodents show context-dependent renewal of extinguished behavior. However, the link between context-dependent neural codes and context-dependent renewal is not fully understood. We use a reinforcement learning agent based on deep neural networks to study the learning dynamics that occur during spatial learning and context switching in a simulated ABA extinction and renewal paradigm in a 3D virtual environment. Despite its simplicity, the network exhibits a number of features typically found in the CA1 and CA3 regions of the hippocampus. A significant proportion of neurons in deeper layers of the network are tuned to a specific spatial position of the agent in the environment - similar to place cells in the hippocampus. These spatial representations exhibit global remapping when the agent is exposed to a new context. The spatial maps are restored when the agent returns to the previous context, accompanied by renewal of the conditioned behavior. Remapping is facilitated by memory replay of experiences during training. These complex spatial representations and dynamics occur spontaneously in the hidden layer of a deep network during learning. Our results show that integrated codes that jointly represent spatial and task-relevant contextual variables are the mechanism underlying renewal in a simulated DQN agent.

neuroscience↗