Biology subjects

Yan, Q.

Publications and source records attributed to Yan, Q..

At least 73 records · Page 4Linked to original sources

bioRxiv2020

AIBP-CAV1-VEGFR3 axis dictates lymphatic cell fateand controls lymphangiogenesis

The lymphatics are essential for the maintenance of tissue fluid homeostasis. Accordingly, lymphatic dysfunction contributes to lymphedema. In development, lymphangiogenesis often requires lymphatic endothelial cell (LEC) lineage specification from the venous ECs and subsequent LEC proliferation and migration, all of which are regulated by the VEGFC/VEGFR3 signaling. Cholesterol is essential for proper cell functions and organ development, yet the molecular mechanism by which cholesterol metabolism controls lymphangiogenesis is unknown. We show that the secreted protein, ApoA1 binding protein (AIBP), dictates lymphatic vessel formation by accelerating cholesterol efflux. Loss of Aibp2, the human paralog in zebrafish, impairs LEC progenitor specification and impedes lymphangiogenesis. Mechanistically, we found that caveolin-1 (CAV-1) suppresses VEGFR3 activation in LECs, and that AIBP-regulated cholesterol efflux disrupts lipid rafts/caveolae and reduces CAV-1 bioavailability, which abolishes the CAV-1 inhibition of VEGFR3 signaling, thereby augmenting VEGFR3 activation and increasing lymphangiogenesis. Enhancement of cholesterol efflux with ApoA1 overexpression or inhibition of cholesterol biosynthesis using atorvastatin restores proper lymphangiogenesis in Aibp2 mutant zebrafish. Loss of Cav-1 increases LEC progenitor specification in zebrafish, and rescues lymphangiogenesis in Aibp2-deficient animals. Recombinant AIBP supplement confers profound LEC fate commitment in the mouse embryonic stem cells (mESC) to LEC differentiation model. Furthermore, enhancement of AIBP-CAV-1-VEGFR3 signaling axis promotes VEGFC-engaged adult lymphangiogenesis in mice. Consistent with these data, AIBP expression is reduced in the epidermis of human lymphedematous skin. These studies identify that AIBP-mediated cholesterol efflux is a critical contributor for lymphangiogenesis. Our studies will provide a new therapeutic avenue for the treatment of lymphatic dysfunctions. One Sentence SummaryOur studies identify that AIBP-CAV-1-VEGFR3 axis enhances VEGFC-elicited lymphangiogenesis, which will guide a new therapeutic strategy for the treatment of lymphatic dysfunctions.

developmental biology↗

bioRxiv2020

CECR2 Drives Breast Cancer Metastasis by Suppressing Macrophage Inflammatory Responses

Epigenetic and transcriptional changes are critical for metastasis, the major cause of cancer-related deaths. Metastatic tumor cells escape immune surveillance more efficiently than tumor cells in the primary sites, but the mechanisms controlling their immune evasion are poorly understood. We found that distal metastases are more immune inert with increased M2 macrophages compared to their matched primary tumors. Acetyl-lysine reader CECR2 is an epigenetic regulator upregulated in metastases and positively associated with M2 macrophages. CECR2 specifically promotes breast cancer metastasis in multiple mouse models, with more profound effect in the immunocompetent setting. Mechanistically, NF-{kappa}B family member RELA recruits CECR2 to activate CSF1 and CXCL1, which are critical for macrophage-mediated immunosuppression at the metastatic sites. Furthermore, pharmacological inhibition of CECR2 bromodomain impedes NF-{kappa}B-mediated immune suppression by macrophages and inhibits breast cancer metastasis. These results reveal novel therapeutic strategies to treat metastatic breast cancer. Statement of SignificanceComparison of matched primary breast tumors and distal metastases show that metastases are more immune inert with increased tumor promoting macrophages. Depletion or pharmacological inhibition of CECR2 inhibits breast cancer metastasis by suppressing macrophage inflammatory responses, nominating CECR2 as a promising therapeutic target for cancer metastasis.

cancer biology↗

bioRxiv2020

Micro-XRF study of the troodontid dinosaur Jianianhualong tengi reveals new biological and taphonomical signals

Jianianhualong tengi is a key taxon for understanding the evolution of pennaceous feathers as well as troodontid theropods, and it is known by only the holotype, which was recovered from the Lower Cretaceous Yixian Formation of western Liaoning, China. Here, we carried out a large-area micro-X-Ray fluorescence (micro-XRF) analysis on the holotypic specimen of Jianianhualong tengi via a Brucker M6 Jetstream mobile XRF scanner. The elemental distribution measurements of the specimen show an enrichment of typical bones couponing elements such as S, P and Ca allowing to visualize the fossil structure. Additionally, to this, the bones are enriched in several heavier elements such as Sr, Th, Y and Ce over the surrounding rocks. The enrichment is most likely associated to secondary mineralization and the phosphates from the bones. Interestingly the plumage shape correlates with an enrichment in elements such as Cu, Ni and Ti, consistent with a previous study [1] on Archaeopteryx using synchrotron imaging. The analysis presented here provide new biological and taphonomic information of this fossil. An in-situ and nondestructive micro-XRF analysis is currently the most ideal way to map the chemistry of fossils, so far this is manly restricted to small samples. Larger samples usually required a synchrotron facility for analysis. Our study demonstrated that laboratory-based large-area micro-XRF scanner can provides a practical tool for the study of large large-sized specimens allowing collect full chemical data for a better understanding of evolutionary and taphonomic processes.

paleontology↗

bioRxiv2020

A DNAmRep epigenetic fingerprint for determining cellular replication age

ii.Aging is known to elicit dramatic changes to DNA methylation (DNAm), although the causes and consequences of such alterations are not clear and require further exploration. Our ability to experimentally uncover mechanisms of epigenetic aging will be greatly enhanced by our ability to study and manipulate these changes using in vitro models. However, it remains unclear whether the changes elicited by cells in culture can serve as a model of what is observed in aging tissues in vivo. To test this, we serially passaged mouse embryonic fibroblasts (MEFs) and assessed changes in DNAm at each time-point via RRBS. By developing a measure that tracked cellular aging in vitro, we tested whether it tracked physiological aging in various mouse tissues and whether anti-aging interventions modulate this measure. Our measure, termed DNAmCULTURE, was shown to strongly increase with age when examined in multiple tissues (liver, lung, kidney, blood, and adipose). As a control, we confirmed that the measure was not a marker of cellular senescence, suggesting that it reflects a distinct yet progressive cellular aging phenomena that can be induced in vitro. Furthermore, we demonstrated slower epigenetic aging in animals undergoing caloric restriction and a resetting of our measure in lung and kidney fibroblasts when re-programmed to iPSCs. Enrichment and clustering analysis implicated SUZ12, EED and Polycomb group (PcG) factors as potentially important chromatin regulators in translational culture aging phenotypes. Overall, this study supports the concept that physiologically relevant aging changes can be induced in vitro and moving forward, used to uncover mechanistic insights into epigenetic aging.

pathology↗

bioRxiv2020

SARS-CoV-2 manipulates the SR-B1-mediated HDL uptake pathway for its entry

The recently emerged pathogenic severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread rapidly, leading to a global COVID-19 pandemic. Binding of the viral spike protein (SARS-2-S) to cell surface receptor angiotensin-converting enzyme 2 (ACE2) mediates host cell infection. In the present study, we demonstrate that in addition to ACE2, the S1 subunit of SARS-2-S binds to HDL and that SARS-CoV-2 hijacks the SR-B1-mediated HDL uptake pathway to facilitate its entry. SR-B1 facilitates SARS-CoV-2 entry into permissive cells by augmenting virus attachment. MAb (monoclonal antibody)-mediated blocking of SARS-2-S-HDL binding and SR-B1 antagonists strongly inhibit HDL-enhanced SARS-CoV-2 infection. Notably, SR-B1 is co-expressed with ACE2 in human pulmonary and extrapulmonary tissues. These findings revealed a novel mechanism for SARS-CoV-2 entry and could provide a new target to treat SARS-CoV-2 infection.

microbiology↗

bioRxiv2020

Characterization of the gut DNA and RNA viromes in a cohort of Chinese residents and visiting Pakistanis

BackgroundTrillions of viruses inhabit the gastrointestinal tract. Some of them have been well-studied on their roles in infection and human health, but the majority remain unsurveyed. It has been established that the composition of the gut virome is highly variable based on the changes of diet, physical state, and environmental factors. However, the effect of host genetic factors, e.g. ethnic origin, on the gut virome is rarely investigated. Methods and ResultsHere, we characterized and compared the gut virome in a cohort of local Chinese residents and visiting Pakistani individuals, each group containing 24 healthy adults and 6 children. Using metagenomic shotgun sequencing and assembly of fecal samples, a huge number of viral operational taxonomic units (vOTUs) were identified for profiling the DNA and RNA viromes. National background contributed a primary variation to individuals gut virome. Compared with the Chinese adults, the Pakistan adults showed higher macrodiversity and different compositional and functional structures in their DNA virome and lower diversity and altered composition in their RNA virome. The virome variations of Pakistan children were inherited from the that of the adults but also tended to share similar characteristics with the Chinese cohort. We also analyzed and compared the bacterial microbiome between two cohorts and further revealed numerous connections between virus and bacterial host. Statistically, the gut DNA and RNA viromes were covariant to some extent (p<0.001), and they both influenced the holistic bacterial composition and vice versa. ConclusionsThis study provides an overview of gut viral community in Chinese and visiting Pakistanis and proposes a considerable role of ethnic origin in shaping the virome.

microbiology↗

bioRxiv2020

Genome-wide CRISPR screen reveals host genes that regulate SARS-CoV-2 infection

Identification of host genes essential for SARS-CoV-2 infection may reveal novel therapeutic targets and inform our understanding of COVID-19 pathogenesis. Here we performed a genome-wide CRISPR screen with SARS-CoV-2 and identified known SARS-CoV-2 host factors including the receptor ACE2 and protease Cathepsin L. We additionally discovered novel pro-viral genes and pathways including the SWI/SNF chromatin remodeling complex and key components of the TGF-{beta} signaling pathway. Small molecule inhibitors of these pathways prevented SARS-CoV-2-induced cell death. We also revealed that the alarmin HMGB1 is critical for SARS-CoV-2 replication. In contrast, loss of the histone H3.3 chaperone complex sensitized cells to virus-induced death. Together this study reveals potential therapeutic targets for SARS-CoV-2 and highlights host genes that may regulate COVID-19 pathogenesis.

genetics↗

bioRxiv2020

Functional requirement of the Arabidopsis importin-α nuclear transport receptor family in autoimmunity mediated by the NLR protein SNC1

IMPORTIN-3/MOS6 (MODIFIER OF SNC1, 6) is one of nine importin- isoforms in Arabidopsis that recruit nuclear localization signal (NLS)-containing cargo proteins to the nuclear import machinery. IMP-3/MOS6 is required genetically for full autoimmunity of the nucleotide-binding leucine-rich repeat (NLR) immune receptor mutant snc1 (suppressor of npr1-1, constitutive 1) and MOS6 also contributes to basal disease resistance. Here, we investigated the contribution of the other importin- genes to both types of immune responses, and we analyzed potential interactions of all importin- isoforms with SNC1. By using reverse-genetic analyses in Arabidopsis and protein-protein interaction assays in N. benthamiana we provide evidence that among the nine -importins in Arabidopsis, IMP-3/MOS6 is the main nuclear transport receptor of SNC1, and that IMP-3/MOS6 is required selectively for autoimmunity of snc1 and basal resistance to mildly virulent Pseudomonas syringae in Arabidopsis. SIGNIFICANCE STATEMENTSpecific requirement for the Arabidopsis -importin MOS6 in snc1-mediated autoimmunity is explained by selective formation of MOS6-SNC1 nuclear import complexes.

plant biology↗

bioRxiv2020

Expression quantitative trait methylation analysis reveals methylomic associations with gene expression in childhood asthma

Nasal airway epithelial methylation profiles have been associated with asthma, but the effects of such profiles on expression of distant cis-genes are largely unknown. We identified 16,867 significant methylation-gene expression pairs in nasal epithelium from Puerto Rican children and adolescents (with and without asthma) in an expression quantitative trait methylation (eQTM) analysis of cis-genes located within 1 Mb of the methylation probes tested. Most eQTM methylation probes were distant from their target genes, and more likely located in enhancer regions of their target genes in lung tissue than control probes. The top 500 eQTM genes were enriched in pathways for immune processes and epithelial integrity, and also more likely to be differentially expressed in atopic asthma. Moreover, we identified 5,934 paths through which methylation probes could affect atopic asthma through gene expression. Our findings suggest that distant epigenetic regulation of gene expression in airway epithelium plays a role in atopic asthma.

genomics↗

bioRxiv2020

Analysis of metagenome-assembled genomes from the mouse gut microbiota reveals distinctive strain-level characteristics

The laboratorial mouse harbors a unique gut microbiota with potential value for human microbiota-associated studies. Mouse gut microbiota has been explored at the genus and species levels, but features rarely been showed at the strain level. The identification of 833,051 and 658,438 nonredundant genes of faeces and gut content samples from the laboratorial C57/BL mice showed over half of these genes were newly found compared to the previous mouse gut microbial gene catalogue. Metagenome-assembled genomes (MAGs) was used to reconstruct 46 nonredundant MAGs belonging to uncultured specieses. These MAGs included members across all phyla in mouse gut (i.e. Firmicutes, Bacteroidetes, Proteobacteria, Deferribacteres, Verrucomicrobia, and Tenericutes) and allowed a strain-level delineating of the mouse gut microbiota. Comparison of MAGs with human gut colonies revealed distinctive genomic and functional characteristics of mouses Bacteroidetes and Firmicutes strains. Genomic characteristics of rare phyla in mouse gut microbiota were demonstrated by MAG approach, including strains of Mucispirillum schaedleri, Parasutterella excrementihominis, Helicobacter typhlonius, and Akkermansia muciniphila. ImportanceThe identification of nonredundant genes suggested the existence of unknown microbes in the mouse gut samples. The metagenome-assembled genomes (MAGs) instantiated the specificity of mouse gut species and revealed an intestinal microbial correlation between mouse and human. The cultivation of faeces and gut contents sample validated the existence of MAGs and estimate their accuracy. Full-length 16S ribosomal RNA gene sequencing enabled taxonomic characterization. This study highlighted a unique ecosystem in the gut of laboratorial mice that obviously differed with the human gut flora at the strain level. The outcomes may be beneficial to researches based on laboratorial mouse models.

microbiology↗

bioRxiv2019

Sample demultiplexing, multiplet detection, experiment planning and novel cell type verification in single cell sequencing

Identifying and removing multiplets from downstream analysis is essential to improve the scalability and reliability of single cell RNA sequencing (scRNA-seq). High multiplet rates create artificial cell types in the dataset. Sample barcoding, including the cell hashing technology and the MULTI-seq technology, enables analytical identification of a fraction of multiplets in a scRNA-seq dataset.\n\nWe propose a Gaussian-mixture-model-based multiplet identification method, GMM-Demux. GMM-Demux accurately identifies and removes the sample-barcoding-detectable multiplets and estimates the percentage of sample-barcoding-undetectable multiplets in the remaining dataset. GMM-Demux describes the droplet formation process with an augmented binomial probabilistic model, and uses the model to authenticate cell types discovered from a scRNA-seq dataset.\n\nWe conducted two cell-hashing experiments, collected a public cell-hashing dataset, and generated a simulated cellhashing dataset. We compared the classification result of GMM-Demux against a state-of-the-art heuristic-based classifier. We show that GMM-Demux is more accurate, more stable, reduces the error rate by up to 69x, and is capable of reliably recognizing 9 multiplet-induced fake cell types and 8 real cell types in a PBMC scRNA-seq dataset.

bioinformatics↗

bioRxiv2019

Structural changes in the gut microbiome of short-term and long-term medical workers compared to normal controls

ObjectiveHospital environment has been implicated in enrichment and exchange of pathogens and antibiotic resistances, but its potential in shaping the symbiotic microbial community of the hospital staff is unclear. This study was designed to evaluate the alteration of gut microbiome in medical workers compared to non-medical controls.\n\nDesignProspective cross-sectional cohort study.\n\nSettingIntensive care unit (ICU) and other departments from a center in northeast China.\n\nSubjects175 healthy medical workers (1-3 months short-term workers, n = 80; >1 year long-term workers, n = 95) and 80 healthy normal controls.\n\nInterventionsNone.\n\nMeasurements and Main ResultsFecal samples of all subjects were analyzed using the 16S rRNA gene sequencing. Medical workers exhibited remarkable deviation in gut microbial within-sample diversity and enterotypes stratification, and shift in overall microbial structure. Short-term workers were significantly more abundant in taxa including Lactobacillus, Butyrivibrio, Clostridiaceae_Clostridium, Ruminococcus, Dialister, Bifidobacterium, Odoribacter and Desulfovibrio, and with lower abundances of Bacteroides and Blautia compared with the controls. While long-term workers were enriched in taxa including Dialister, Veillonella, Clostridiaceae_Clostridium, Bilophila, Desulfovibrio, Pseudomonas and Akkermansia, with lower abundances of Bacteroides and Coprococcus compared with the controls. In addition, medical workers working years (short-term vs. long-term), hospital department (resident doctor vs. nursing staff) and work position (ICU vs. not-ICU) revealed considerable effects on their gut microbiome. Moreover, by analyzing the environmental samples (n = 9) around the inpatient wards and the hospital, we showed that the gut microbiota of medical workers was closer to environmental microbiota than that of the normal controls, probably in correlation to lasting exposure to the pathogenic taxa (e.g. Pseudomonas) in health workers.\n\nConclusionsOur findings demonstrated structural changes in the gut microbial community of the medical workers. Further studies are proposed for investigating the potentially physiological influence of the altered gut microbiome in medical participants.\n\nIMPORTANCEIn this study, we for the first time focused on the influence of hospital environmental factors on gut microbiota of medical workers. The significance of our study is not limited to revealing the remodeling effect of the hospital environment on the gut microbiota of medical workers. Based on these, we also propose targeted and operational recommendations that can promote the health of hospital staff.

microbiology↗

bioRxiv2019

Genome-wide development of miRNA-based SSR markers in Cleistogenes songorica with their transferability analysis to gramineae and non- gramineae species

SSR markers are commonly used for many genetic applications, such as map construction, fingerprinting and genetic diversity analysis due to their high reproducibility, levels of polymorphism and abundance. As endogenous, small RNAs, miRNAs have essential roles in plant development and gene expression under diverse stress conditions, including various biotic and abiotic stress conditions. In the present study, we predicted 110 pre-miRNAs sequences from 287 precursor miRNAs and used them as queries for SSR marker development. Among 110 primer pairs, 85 were successfully amplified and examined for transferability to other gramineae and non-gramineae species. The results showed that all 82 primer pairs yielded unambiguous and strong amplification, and across the 23 studied Cleistogenes accessions, a total of 385 alleles were polymorphic. The number of alleles produced per primer varied from 3 to 11, with an average of 4.69 per locus. The expected heterozygosity (He) ranged from 0.44 to 0.88, with an average of 0.74 per locus, and the PIC (Polymorphism Information Content) values ranged from 0.34 to 0.87, with an average of 0.69 per locus. In this study, 1422 miRNA target genes were predicted and analyzed using the GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) databases. The results showed that this miRNA-based microsatellite marker system can be very useful for genetic diversity and marker-assisted breeding studies.

genomics↗

bioRxiv2019

The Connectivity Index: An Effective Metric for Grading Epileptogenicity

ObjectiveTo investigate the performance of a metric of functional connectivity to classify and grade the excitability of brain regions based on evoked potentials to single pulse electrical stimulation (SPES).\n\nMethodsPatients who received 1-Hz frequency stimulation between 2003 and 2014 at Yale at prospectively selected contacts were included. The stimulated contacts were classified as seizure onset zone (SOZ), highly irritative zone (IZp) or control. Response contacts were classified as seizure onset zone (SOZ), active interictal (IZp), quiet or other. The normalized number of responses was defined as the number of contacts with any evoked responses divided by the total number of recorded contacts, and the normalized distance is the ratio of the average distance between the site of stimulation and sites of evoked responses to the average distances between the site of stimulation and all other recording contacts. A new metric we labeled the connectivity index (CI) is defined as the product of the two values.\n\nResults57 stimulation-sessions in 22-patients were analyzed. The connectivity index (CI) of the SOZ was higher than control (median CI of 0.74 vs. 0.16, p = 0.0002). The evoked responses after stimulation of SOZ were seen at further distance compared to control (median normalized distance 0.96 vs. 0.62, p = 0.0005). It was 1.8 times more likely to record a response at SOZ than in non-epileptic contacts after stimulation of a control site. Habitual seizures were triggered in 27% of patients and 35 % of SOZ contacts (median stimulation intensity 4 mA) but in none of the control or IZp contacts. Non-SOZ contacts in multifocal or poor surgical outcome cases had a higher CI than non-SOZ contacts in those with localizable onsets (medians CI of 0.5 vs. 0.12, p = 0.04). There was a correlation between the stimulation current intensity and the normalized number of evoked responses (r = + 0.49, p 0.01) but not with distance (r = + 0.1, p 0.64)\n\nConclusionsWe found enhanced connectivity when stimulating the SOZ compared to stimulating control contacts; responses were more distant as well. Habitual auras and seizures provoked by SPES were highly predictive of brain sites involved in seizure generation.

neuroscience↗

bioRxiv2018

Mapping native R-loops genome-wide using a targeted nuclease approach

R-loops are three-stranded DNA:RNA hybrids that are pervasive in the eukaryotic and prokaryotic genomes and have been implicated in a variety of nuclear processes, including transcription, replication, DNA repair, and chromosome segregation. While R-loops may have physiological roles, the formation of stable, aberrant R-loops has been observed in disease, particularly neurological disorders and cancer. Despite the importance of these structures, methods to assess their distribution in the genome invariably rely on affinity purification, which requires large amounts of input material, is plagued by high level of noise, and is poorly suited to capture dynamic and unstable R-loops. Here, we present a new method that leverages the affinity of RNase H for DNA:RNA hybrids to target micrococcal nuclease to genomic sites that contain R-loops, which are subsequently cleaved, released, and sequenced. Our R-loop mapping method, MapR, is as specific as existing techniques, less prone to recover non-specific repetitive sequences, and more sensitive, allowing for genome-wide coverage with low input material and read numbers, in a fraction of the time.

molecular biology↗

bioRxiv2018

Precise temporal regulation of alternative splicing during neural development

Alternative splicing (AS) is a crucial step of gene expression that must be tightly controlled, but the precise timing of dynamic splicing switches during neural development and the underlying regulatory mechanisms are poorly understood. Here we systematically analyzed the temporal regulation of AS in a large number of transcriptome profiles of developing mouse cortices, in vivo purified neuronal subtypes, and neurons differentiated in vitro. Our analysis revealed early- and late-switch exons in genes with distinct functions, and these switches accurately define neuronal maturation stages. Integrative modeling suggests that these switches are under direct and combinatorial regulation by distinct sets of neuronal RNA-binding proteins including Nova, Rbfox, Mbnl and Ptbp. Surprisingly, various neuronal subtypes in the sensory systems lack Nova and/or Rbfox expression. These neurons retain the \"immature\" splicing program in early-switch exons, affecting numerous synaptic genes. These results provide new insights into the organization and regulation of the neurodevelopmental transcriptome.

molecular biology↗

bioRxiv2017

A cytoplasmic Argonaute protein promotes the inheritance of RNAi

RNAi-elicited gene silencing is heritable and can persist for multiple generations after its initial induction in C. elegans. However, the mechanism by which parental-acquired trait-specific information from RNAi is inherited by the progenies is not fully understood. Here, we identified a cytoplasmic Argonaute protein, WAGO-4, necessary for the inheritance of RNAi. WAGO-4 exhibits asymmetrical translocation to the germline during early embryogenesis, accumulates at the perinuclear foci in the germline, and is required for the inheritance of exogenous RNAi targeting both germline- and soma-expressed genes. WAGO-4 binds to 22G-RNAs and their mRNA targets. Interestingly, WAGO-4-associated endogenous 22G-RNAs target the same cohort of germline genes as CSR-1 and contain untemplated addition of uracil at the 3 ends. The poly(U) polymerase CDE-1 is required for the untemplated uridylation of 22G-RNAs and inheritance of RNAi. Therefore, we conclude that, in addition to the nuclear RNAi pathway, the cytoplasmic RNAi machinery also promotes RNAi inheritance.

genetics↗

bioRxiv2017

Revealing structure components of the retina by deep learning networks

Deep convolutional neural networks (CNNs) have demonstrated impressive performance on visual object classification tasks. In addition, it is a useful model for predication of neuronal responses recorded in visual system. However, there is still no clear understanding of what CNNs learn in terms of visual neuronal circuits. Visualizing CNNs features to obtain possible connections to neuronscience under-pinnings is not easy due to highly complex circuits from the retina to higher visual cortex. Here we address this issue by focusing on single retinal ganglion cells with a simple model and electrophysiological recordings from salamanders. By training CNNs with white noise images to predicate neural responses, we found that convolutional filters learned in the end are resembling to biological components of the retinal circuit. Features represented by these filters tile the space of conventional receptive field of retinal ganglion cells. These results suggest that CNN could be used to reveal structure components of neuronal circuits.

neuroscience↗