Biology subjects

Chen, X.

Publications and source records attributed to Chen, X..

At least 757 records · Page 42Linked to original sources

bioRxiv2023

Efficacy of the first bioactive peptide from the pearl

Pearls have high medicinal value. In the present study, we discovered the first bioactive peptide in pearls. The bioactive peptide, KKCHFWPFPW, was a novel angiotensin I-converting enzyme (ACE)-inhibitory peptide derived from the pearl matrix of Pinctada fucata. It was screened and identified using quadrupole time-of-flight mass spectrometry. The molecular weight of the peptide was 1417.5 Da, and its theoretical isoelectric point was 9.31. The half-maximal inhibitory concentration of the peptide was 4.17M, as determined by high-performance liquid chromatography. The Lineweaver-Burk plot showed that this peptide competitively inhibited ACE activity. As the peptide concentration increased, the ACE inhibition rate also increased. The molecular docking was simulated using Maestro 2022-1 Glide software to understand the potential mechanisms underlying the ACE-inhibitory activity of KKCHFWPFPW. These results indicated that the peptide from the P. martensii pearl matrix might be a potential source of antihypertensive peptides.

biochemistry↗

bioRxiv2023

Epigenetic insights into GABAergic development in Dravet Syndrome iPSC and therapeutic implications

Dravet syndrome (DS) is a devastating early onset refractory epilepsy syndrome caused by variants in the SCN1A gene. A disturbed GABAergic interneuron function is implicated in the progression to DS but the underlying developmental and pathophysiological mechanisms remain elusive, in particularly at the chromatin level. In this study, we utilized induced pluripotent stem cells (iPSCs) derived from DS cases and healthy donors to model disease- associated epigenetic abnormalities of GABAergic development. Employing the ATAC-Seq technique, we assessed chromatin accessibility at multiple time points (Day 0, Day 19, Day 35, and Day 65) of GABAergic differentiation. Additionally, we elucidated the effects of the commonly used anti-seizure drug valproic acid (VPA) on chromatin accessibility in GABAergic cells. The distinct dynamics in chromatin profile of DS iPSC predicted accelerated early GABAergic development, evident at D19, and diverged further from the pattern in control iPSC with continued differentiation, indicating a disrupted GABAergic maturation. Exposure to VPA at D65 reshaped the chromatin landscape at a variable extent in different iPSC-lines and rescued the observed dysfunctional development in some DS iPSC-GABA. This study provides the first comprehensive investigation on the chromatin landscape of GABAergic differentiation in DS-patient iPSC, offering valuable insights into the epigenetic dysregulations associated with interneuronal dysfunction in DS. Moreover, our detailed analysis of the chromatin changes induced by VPA in iPSC-GABA holds the potential to improve development of personalized and targeted anti-epileptic therapies.

cell biology↗

bioRxiv2023

Shared and Distinct Mechanisms of UBA1 Inactivation Across Different Diseases

Most cellular ubiquitin signaling is initiated by UBA1, which activates and transfers ubiquitin to tens of E2 enzymes. Clonally acquired UBA1 missense mutations cause an inflammatory-hematologic overlap disease called VEXAS (vacuoles, E1, X-linked, autoinflammatory, somatic) syndrome. Despite extensive clinical investigation into this lethal disease, little is known about the underlying molecular mechanisms. Here, by dissecting VEXAS-causing UBA1 mutations, we discovered that p.Met41 mutations alter cytoplasmic isoform expression, whereas other mutations reduce catalytic activity of nuclear and cytoplasmic isoforms by diverse mechanisms, including aberrant oxyester formation. Strikingly, non-p.Met41 mutations most prominently affect transthioesterification, revealing ubiquitin transfer to cytoplasmic E2 enzymes as a shared property of pathogenesis amongst different VEXAS syndrome genotypes. A similar E2 charging bottleneck exists in some lung cancer-associated UBA1 mutations, but not in spinal muscular atrophy-causing UBA1 mutations, which instead, render UBA1 thermolabile. Collectively, our results highlight the precision of conformational changes required for faithful ubiquitin transfer, define distinct and shared mechanisms of UBA1 inactivation in diverse diseases, and suggest that specific E1-E2 modules control different aspects of tissue differentiation and maintenance.

biochemistry↗

bioRxiv2023

Collaborative Augmented Reconstruction for Scaled Production of 3D Neuron Morphology in Mouse and Human Brains

Digital reconstruction of the intricate 3D morphology of individual neurons from microscopic images is widely recognized as a crucial challenge in both individual research laboratories and large-scale scientific projects focusing on cell types and brain anatomy. This task often fails both conventional manual reconstruction and state-of-the-art automatic reconstruction algorithms, even many of which are based on artificial intelligence (AI). It is also critical but challenging to organize multiple neuroanatomists to produce and cross-validate biologically relevant and agreeable reconstructions in scaled data production. Here we propose an approach based on collaborative human intelligence augmented by AI. Specifically, we have developed a Collaborative Augmented Reconstruction (CAR) platform for neuron reconstruction at scale. This platform allows for immersive interaction and efficient collaborative-editing of neuron anatomy using a variety of client devices, such as desktop workstations, virtual reality headsets, and mobile phones, enabling users to contribute anytime and anywhere and take advantage of several AI-based automation tools. We have tested CARs applicability for challenging mouse and human neurons towards a scaled and faithful data production. Our data indicate that the CAR platform is suitable for generating tens of thousands of neuronal reconstructions used in our companion studies.

neuroscience↗

bioRxiv2023

The serine/threonine kinase CcAkt regulates the fertility of Coridius chinensis

Akt (also known as protein kinase B) belongs to the multifunctional serine/threonine kinase family and is an important component of the insulin signaling pathway that plays a key role in many biological processes such as cell growth, proliferation, and survival. However, few studies have reported the effect of Akt on reproduction in Hemiptera. In this study, we cloned and characterized the Akt gene from Coridius chinensis (CcAkt). The open reading frame of CcAkt has a length of 1,563 bp and encodes 520 amino acids. It has a conserved pleckstrin homology domain, a serine/threonine protein kinase catalytic domain, and a serine/threonine protein kinase domain. Phylogenetic analysis showed that CcAkt and HhAkt of Halyomorpha halys had the highest similarity. Analysis of temporal and spatial expression patterns revealed that CcAkt is expressed throughout development and in various tissues of C. chinensis adults. CcAkt expression was high in female adult and the fourth instar nymph stage of C. chinensis. In C. chinensis adult, CcAkt expression was highest in the testis and ovary. Injection of bovine insulin induced CcAkt expression, whereas that of 20-hydroxyecdysone significantly reduced CcAkt expression. Both hormones, however, induced the expression of vitellogenin (Vg) and vitellogenin receptor (VgR). In unmated females, CcAkt knockout resulted in decreased expression of CcVg and CcVgR, stunted the development of the ovarioles, decreased the number of eggs and hatching rate. These findings suggest that CcAkt may be involved in regulating the reproduction of C. chinensis.

molecular biology↗

bioRxiv2023

Spiral NeuroString: High-Density Soft Bioelectronic Fibers for Multimodal Sensing and Stimulation

Bioelectronic fibers hold promise for both research and clinical applications due to their compactness, ease of implantation, and ability to incorporate various functionalities such as sensing and stimulation. However, existing devices suffer from bulkiness, rigidity, limited functionality, and low density of active components. These limitations stem from the difficulty to incorporate many components on one-dimensional (1D) fiber devices due to the incompatibility of conventional microfabrication methods (e.g., photolithography) with curved, thin and long fiber structures. Herein, we introduce a fabrication approach, "spiral transformation'', to convert two-dimensional (2D) films containing microfabricated devices into 1D soft fibers. This approach allows for the creation of high density multimodal soft bioelectronic fibers, termed Spiral NeuroString (S-NeuroString), while enabling precise control over the longitudinal, angular, and radial positioning and distribution of the functional components. We show the utility of S-NeuroString for motility mapping, serotonin sensing, and tissue stimulation within the dynamic and soft gastrointestinal (GI) system, as well as for single-unit recordings in the brain. The described bioelectronic fibers hold great promises for next-generation multifunctional implantable electronics.

bioengineering↗

bioRxiv2023

RAG1 and RAG2 non-core regions are implicated in the leukemogenesis and off-target V(D)J recombination in BCR-ABL1-driven B cell lineage lymphoblastic leukemia.

The evolutionary conservation of non-core RAG regions suggests significant roles that might involve quantitative or qualitative alterations in RAG activity. Off-target V(D)J recombination contributes to lymphomagenesis and is exacerbated by RAG2 C-terminus absence in Tp53-/- mice thymic lymphomas. However, the genomic stability effects of non-core regions from both cRAG1 and cRAG2 in BCR-ABL1+ Blymphoblastic leukemia (BCR-ABL1+ B-ALL), the characteristics, and mechanisms of non-core regions in suppressing off-target V(D)J recombination remains unclear. Here, we established three mouse models of BCR-ABL1+ B-ALL in mice expressing full-length RAG (fRAG), core RAG1 (cRAG1), and core RAG2 (cRAG2). The cRAG (cRAG1 and cRAG2) leukemia cells exhibited greater malignant tumor characteristics compared to fRAG cells. Additionally, cRAG cells showed higher frequency of off-target V(D)J recombination and oncogenic mutations than fRAG. We also revealed decreased RAG cleavage accuracy in cRAG cells and a smaller recombinant size in cRAG1 cells, which could potentially exacerbate off-target V(D)J recombination in cRAG cells. In conclusion, these findings indicate that the non-core RAG regions, particularly the non-core region of RAG1, play a significant role in preserving V(D)J recombination precision and genomic stability in BCR-ABL1+ B-ALL.

immunology↗

bioRxiv2023

Central pore-opening structure and gating of the mechanosensitive PIEZO1 channel

PIEZO1 is a mechanically activated cation channel that undergoes force-induced activation and inactivation to control diverse physiological processes1-4. However, the distinct functional states and gating transitions remain structurally undefined. In response to changes in membrane forces, PIEZO1 transitions from a curved to a flattened shape, which might lead to opening and subsequent inactivation of the ion-conducting pore5,6. Here we employed a PIEZO1 mutant to capture various conformational states including an intermediate state with an opened central pore, allowing us to reveal the gating dynamics of this multiple-gate system. Upon transition from the curved to the partially flattened intermediate state, the trimeric cap undergoes a downward and rotational motion to open the lateral cap-gate, while the loop that connects the cap to the pore-lining inner helix acts as a spring-linker, switching from an extended to a compressed state. Along with the partial flattening of the peripheral blades, these conformational changes collectively dilate the cap-gate, spring-gate and the hydrophobic transmembrane-gate, leading to ion permeation through the transmembrane pore. Upon inactivation, while the blades remain flattened, the cap and spring-linker return to an up and extended state. Molecular dynamics simulations verify the ion-permeating pathway and the distinct ion-conducting states of the assigned close, intermediate, open, and inactivated structures. Mutagenesis and electrophysiological characterizations reveal that domains and residues involved in conformational dynamics of the cap and spring-linker are critical for mechanical activation of PIEZO1. Together, these studies have structurally defined the distinct functional states and the curvature-based activation and inactivation mechanisms of PIEZO1.

neuroscience↗

bioRxiv2023

A resource for probing influences of perinatal factors on neurodevelopment in 3-10-years-old Chinese children

Adverse perinatal factors can disrupt the normal development of the brain, with potential long-term impacts on childrens overall development. Currently, the neuropathological mechanisms by which these factors lead to various neurodevelopmental disorders (NDDs) remain largely unknown. An open resource that integrate perinatal factors with brain and mental health development is essential for investigating NDD-related aetiology. In this Data Descriptor, we introduce a multicentre database containing information on perinatal factors that influence childrens brain-mind development, namely, periCBD, that combines neuroimaging and behavioural phenotypes with perinatal factors associated with a high incidence of NDDs at county/region/central district hospitals. PeriCBD was designed to establish a platform for the investigation of individual differences in brain-mind development among children aged 3-10 years are associated with perinatal factors. Ultimately, our goal was to develop an early prediction and screening model for NDDs that leverages normative data to facilitate NDD aetiology research. Herein, we provide a systematic overview of the data acquisition/cleaning/quality control/sharing, processes of periCBD and present preliminary brain-mind associations.

neuroscience↗

bioRxiv2023

Integrated single-cell multiome analysis reveals muscle fiber-type gene regulatory circuitry modulated by endurance exercise

Endurance exercise is an important health modifier. We studied cell-type specific adaptations of human skeletal muscle to acute endurance exercise using single-nucleus (sn) multiome sequencing in human vastus lateralis samples collected before and 3.5 hours after 40 min exercise at 70% VO2max in four subjects, as well as in matched time of day samples from two supine resting circadian controls. High quality same-cell RNA-seq and ATAC-seq data were obtained from 37,154 nuclei comprising 14 cell types. Among muscle fiber types, both shared and fiber-type specific regulatory programs were identified. Single-cell circuit analysis identified distinct adaptations in fast, slow and intermediate fibers as well as LUM-expressing FAP cells, involving a total of 328 transcription factors (TFs) acting at altered accessibility sites regulating 2,025 genes. These data and circuit mapping provide single-cell insight into the processes underlying tissue and metabolic remodeling responses to exercise.

genomics↗

bioRxiv2023

Tissue-specific knockout in Drosophila neuromuscular system reveals ESCRT's role in formation of synapse-derived extracellular vesicles

Tissue-specific gene knockout by CRISPR/Cas9 is a powerful approach for characterizing gene functions in animal development. However, this approach has been successfully applied in only a small number of Drosophila tissues. The Drosophila motor nervous system is an excellent model system for studying the biology of neuromuscular junction (NMJ). To expand tissue-specific CRISPR to the Drosophila motor system, here we present a CRISPR-mediated tissue-restricted mutagenesis (CRISPR-TRiM) toolkit for knocking out genes in motoneurons, muscles, and glial cells. We validated the efficacy of this toolkit by knocking out known genes in each tissue, demonstrated its orthogonal use with the Gal4/UAS binary expression system, and showed simultaneous knockout of multiple redundant genes. Using these tools, we discovered an essential role for SNARE pathways in NMJ maintenance. Furthermore, we demonstrate that the canonical ESCRT pathway suppresses NMJ bouton growth by downregulating the retrograde Gbb signaling. Lastly, we found that axon termini of motoneurons rely on ESCRT-mediated intra-axonal membrane trafficking to lease extracellular vesicles at the NMJ. SIGNIFICANCE STATEMENTIn this study, we developed a tissue-specific Cas9 toolkit that enables gene knockout specifically in motor neurons, glial cells, and muscle cells, the three cell types of the Drosophila peripheral motor system. Complementary to existing RNAi methods, this versatile tissue-specific knockout system offers unique advantages for dissecting gene functions at the neuromuscular junction (NMJ). Using these tools, we discovered that SNARE-mediated secretory pathways are required to maintain the integrity of the NMJ and that ESCRT components play critical yet differential roles in the biogenesis of extracellular vesicles, bouton growth, and membrane turnover at the NMJ. This CRISPR toolkit can be applied to study many biological questions in the neuromuscular system.

neuroscience↗

bioRxiv2023

Cardiac disease diagnosis based on GAN in case of missing data

In daily life, two common algorithms are used for collecting medical disease data: data integration of medical institutions and questionnaires. However, these statistical methods require collecting data from the entire research area, which consumes a significant amount of manpower and material resources. Additionally, data integration is difficult and poses privacy protection challenges, resulting in a large number of missing data in the dataset. The presence of incomplete data significantly reduces the quality of the published data, hindering the timely analysis of data and the generation of reliable knowledge by epidemiologists, public health authorities, and researchers. Consequently, this affects the downstream tasks that rely on this data. To address the issue of discrete missing data in cardiac disease, this paper proposes the AGAN (Attribute Generative Adversarial Nets) architecture for missing data filling, based on generative adversarial networks. This algorithm takes advantage of the strong learning ability of generative adversarial networks. Given the ambiguous meaning of filling data in other network structures, the attribute matrix is designed to directly convert it into the corresponding data type, making the actual meaning of the filling data more evident. Furthermore, the distribution deviation between the generated data and the real data is integrated into the loss function of the generative adversarial networks, improving their training stability and ensuring consistency between the generated data and the real data distribution. This approach establishes the missing data filling mechanism based on the generative adversarial networks, which ensures the rationality of the data distribution while filling the missing data samples. The experimental results demonstrate that compared to other filling algorithms, the data matrix filled by the proposed algorithm in this paper has more evident practical significance, fewer errors, and higher accuracy in downstream classification prediction.

physiology↗

bioRxiv2023

Bio-inspired mineralization collagen induce fibrocartilage regeneration after tendon-bone injury by activating Gli1+Dkk3+ progenitor cells

A fibrocartilaginous connection between the tendon and bone, plays a critical role in transferring force from muscle to bone to enable joint movement. However, due to the high mechanical stress it experiences, the enthesis is vulnerable to injury and incapable of regenerating. The spatial relationship and functional basis of the principal components of the fibrocartilage - mineral and collagen - have not been clearly elucidated, which is a significant remaining gap in reconstructing complex architectures for promoting interface tissue regeneration. Here, using three-dimensional electron tomography imaging and high-resolution two-dimensional electron microscopy, we discover that mineral particles form a continuous cross-fibrillar phase within the fibrocartilage region. By developing a "floating mineralization" system, we fabricate a three-layer hydrogel that mimics the hierarchical nano- to micro-scale structure of tendon-bone interface (TBI). The middle layer is noteworthy for its resemblance to the nanostructure of fibrocartilage and its superior ability to induce mineralized fibrochondrogenesis in vitro. Based on motor function analysis, imaging diagnosis, histological staining, immunofluorescence staining, and biomechanics performance, we demonstrate that in situ transplantation of the gradient hydrogel achieved tendon-fibrocartilage-bone synchronous regeneration and result in 68% maximum mechanical recovery at 8-week postoperation. Single-cell RNA sequencing analysis reveals that a unique atlas of in situ stem/progenitor cells is generated during the TBI healing in vivo. Notably, the bio-inspired hydrogel microenvironment drived endogenous Gli1+Dkk3+ progenitor cells, playing a key role in TBI regeneration. Therefore, we have successfully decoded and reconstructed the nanostructure of fibrocartilage, which has great potential in TBI regeneration.

bioengineering↗

bioRxiv2023

Genome-wide distribution of 5-hydroxymethyluracil and chromatin accessibility in the Breviolum minutum genome

In dinoflagellates, a unique and extremely divergent genomic and nuclear organization has evolved. The highly unusual features of dinoflagellate nuclei and genomes include permanently condensed liquid crystalline chromosomes, primarily packaged by proteins other than histones, genes organized in very long unidirectional gene arrays, a general absence of transcriptional regulation, high abundance of the otherwise very rare DNA modification 5-hydroxymethyluracil (5-hmU), and many others. While most of these fascinating properties were originally identified in the 1970s and 1980s, they have not yet been investigated using modern genomic tools. In this work, we address some of the outstanding questions regarding dinoflagellate genome organization by mapping the genome-wide distribution of 5-hmU (using both immunoprecipitation-based and basepair-resolution chemical mapping approaches) and of chromatin accessibility in the genome of the Symbiodiniaceae dinoflagellate Breviolum minutum. We find that the 5-hmU modification is preferentially enriched over certain classes of repetitive elements, often coincides with the boundaries between gene arrays, and is generally correlated with decreased chromatin accessibility, the latter otherwise being largely uniform along the genome. We discuss the potential roles of 5-hmU in the functional organization of dinoflagellate genomes and its relationship to the transcriptional landscape of gene arrays.

genomics↗

bioRxiv2023

Kirigami electronics for long-term electrophysiological recording of human neural organoids and assembloids

Organoids and assembloids have emerged as a promising platform to model aspects of nervous system development. Longterm, minimally-invasive recordings in these multi-cellular systems are essential for developing disease models. Current technologies, such as patch-clamp, penetrating microelectrodes, planar electrode arrays and substrate-attached flexible electrodes, do not, however, allow chronic recording of organoids in suspension, which is necessary to preserve their architecture. Inspired by the art of kirigami, we developed flexible electronics that transition from a 2D pattern to a 3D basketlike configuration to accommodate the long-term culture of organoids in suspension. This platform, named kirigami electronics (KiriE), integrates with and enables chronic recording of cortical organoids while preserving morphology, cytoarchitecture, and cell composition. KiriE can be integrated with optogenetic and pharmacological stimulation and model disease. Moreover, KiriE can capture activity in cortico-striatal assembloids. Moving forward, KiriE could reveal disease phenotypes and activity patterns underlying the assembly of the nervous system.

neuroscience↗

bioRxiv2023

Diversity and distribution of viruses inhabiting the oil reservoirs

Oil reservoirs, being one of the most significant subsurface repositories of energy and carbon, have long hosted diverse microorganisms affecting energy production and carbon emissions. Viruses play crucial roles in the ecology of microbiomes, however, their distribution and ecological significance in oil reservoirs remain undetermined. Here, we assembled an extensive catalogue encompassing viral and prokaryotic genomes sourced from oil reservoirs. The catalogue comprises 7,229 prokaryotic genomes and 6,218 viral genomes from 182 oil reservoir metagenomes, respectively. Based on sequence clustering we identified 3,886 Operational Taxonomic Units (vOTUs) approximately at species level, 94.80% of which were not found in any other environments. The results showed that viruses were widely distributed in oil reservoirs, and oil reservoirs contain a significant number of unique and unexplored viruses. We also constructed a catalogue of 322,060 viral gene clusters including 105 virus-encoded putative auxiliary metabolic genes (AMGs) that participate in host metabolism and adaptation to the environment. Furthermore, our investigation has yielded a total of 7,197 putative virus-host pairs based on CRISPR and tRNA profiles. Viruses within oil reservoirs profoundly infected bacterial Actinobacteriota, Desulfobacterota, as well as archaeal Halobacteriota and Methanobacteriota. Combined microcosm enrichment experiments and bioinformatics analysis, we validated the ecological roles of viruses in regulating the community structure of sulfate reduction microorganisms, primarily through the predation of virulent viruses. Collectively, these findings have unveiled a rich diversity of novel viruses and their ecological functions within oil reservoirs. This study provides a comprehensive understanding of the role of viral communities in the biogeochemical cycle of the deep biosphere. It also facilitates microbial applications aimed at addressing challenges related to fossil-fuel production and carbon emissions in the petroleum industry.

microbiology↗

bioRxiv2023

Picomolar-Level Sensing of Cannabidiol by Metal Nanoparticles Functionalized with Chemically Induced Dimerization Binders

Simple and fast detection of small molecules is critical to health and environmental monitoring. Methods for chemical detection often use mass spectrometers or enzymes; the former relies on expensive equipment and the latter is limited to those that can act as enzyme substrates. Affinity reagents like antibodies can target a variety of small-molecule analytes, but the detection requires successful design of chemically conjugated targets or analogs for competitive binding assays. Here, we developed a generalizable method for highly sensitive and specific in-solution detection of small molecules, using cannabidiol (CBD) as an example. Our sensing platform uses gold nanoparticles (AuNPs) functionalized with a pair of chemically induced dimerization (CID) nanobody binders (nano-binders), where CID triggers AuNPs aggregation and sedimentation in the presence of CBD. Despite moderate binding affinities of the two nano-binders to CBD (KDs of [~]6 and [~]56 {micro}M), a scheme consisting of CBD-AuNP pre-analytical incubation, centrifugation, and electronic detection (ICED) was devised to demonstrate a high sensitivity (limit of detection of [~]100 picomolar) in urine and saliva, a relatively short assay time ([~]2 hours), a large dynamic range (5 logs), and a sufficiently high specificity to differentiate CBD from its analog, tetrahydrocannabinol. The high sensing performance was achieved with the multivalency of AuNP sensing, the ICED scheme that increases analyte concentrations in a small assay volume, and a portable electronic detector. This sensing system is readily coupled to other binders for wide molecular diagnostic applications.

bioengineering↗

bioRxiv2023

A Neural Speech Decoding Framework Leveraging Deep Learning and Speech Synthesis

Decoding human speech from neural signals is essential for brain-computer interface (BCI) technologies restoring speech function in populations with neurological deficits. However, it remains a highly challenging task, compounded by the scarce availability of neural signals with corresponding speech, data complexity, and high dimensionality, and the limited publicly available source code. Here, we present a novel deep learning-based neural speech decoding framework that includes an ECoG Decoder that translates electrocorticographic (ECoG) signals from the cortex into interpretable speech parameters and a novel differentiable Speech Synthesizer that maps speech parameters to spectrograms. We develop a companion audio-to-audio auto-encoder consisting of a Speech Encoder and the same Speech Synthesizer to generate reference speech parameters to facilitate the ECoG Decoder training. This framework generates natural-sounding speech and is highly reproducible across a cohort of 48 participants. Among three neural network architectures for the ECoG Decoder, the 3D ResNet model has the best decoding performance (PCC=0.804) in predicting the original speech spectrogram, closely followed by the SWIN model (PCC=0.796). Our experimental results show that our models can decode speech with high correlation even when limited to only causal operations, which is necessary for adoption by real-time neural prostheses. We successfully decode speech in participants with either left or right hemisphere coverage, which could lead to speech prostheses in patients with speech deficits resulting from left hemisphere damage. Further, we use an occlusion analysis to identify cortical regions contributing to speech decoding across our models. Finally, we provide open-source code for our two-stage training pipeline along with associated preprocessing and visualization tools to enable reproducible research and drive research across the speech science and prostheses communities.

neuroscience↗