Biology subjects

Shi, X.

Publications and source records attributed to Shi, X..

At least 127 records · Page 7Linked to original sources

bioRxiv2023

Modern wheat breeding selection synergistically improves above- and below-ground traits

The root system, as a fundamental organ for uptaking water and nutrients and interacting with the local environmental conditions, has been postulated to be the foundation for a second Green Revolution. However, the status of the root system during modern wheat breeding remains to be elucidated. Here, by analyzing the phenotypes of 406 wheat accessions on a large scale, we found the root systems of modern cultivars were synchronisely changed along with the above-ground traits. Furthermore, the genomic blocks with phenotypic effects on both above- and below-ground traits were observed to be enriched in the selection sweeps, highlighting that modern wheat breeding contributed to the synchronised changes. More importantly, the haplotypes selected by breeders within the selection sweeps synergistically improved both the above- and below-ground traits, suggesting that modern wheat breeding has improved the root system indirectly, which may contribute to the higher grain yields of modern wheat cultivars. Our results demonstrated that modern wheat breeding synergistically improved the above- and below-ground traits.

genetics↗

bioRxiv2023

A Hill type equation reveals the regulation principle of target protein expression led by p53 pulsing

The central dogma indicates the basic direction of gene expression pathways. For activated gene expression, the quantitative relationship between various links from the binding of transcription factors (TFs) to DNA to protein synthesis remains unclear and debated. There is consensus that at a steady state, protein levels are largely determined by the mRNA level. How can we find this steady state? Taking p53 as an example, based on the previously discovered Hill-type equation that characterizes mRNA expression under p53 pulsing, I proved that the same equation can be used to describe the average steady state of target protein expression. Therefore, at steady state, the average fold changes in mRNA and protein expression under TFs pulsing were the same. This consensus has been successfully demonstrated. For the p53 target gene BAX, the observed fold changes in mRNA and protein expression were 1.40 and 1.28, respectively; the fold changes in mRNA and protein expression calculated using the Hill-type equation were both 1.35. Therefore, using this equation, we can not only fine-tune gene expression, but also predict the proteome from the transcriptome. Furthermore, by introducing two quantitative indicators, we can determine the degree of accumulation and stability of protein expression.

systems biology↗

bioRxiv2023

The SAM Domain of EphA2 Inhibits Ligand-Independent Clustering and Activation

Eph receptors are the largest family of receptor tyrosine kinases (RTKs). They play a role in the pathogenesis of various diseases including cancer, atherosclerosis, fibrosis, infectious diseases, diseases of the central nervous system and age-related cataract. EphA2 has attracted much attention over the years owing to its dysregulation in many diseases. Previous studies have revealed the unique molecular organizations of Eph receptors, and particularly EphA2, into large clusters of receptor-ligand complexes. One unique feature of Eph receptors is a C-terminal sterile alpha motif (SAM) domain, which has been proposed to alter dimerization and kinase activity in EphA2. However, the precise role of the SAM domain in regulating the function and oligomerization state of EphA2 has not been reported. Here we apply a time-resolve fluorescence spectroscopy, PIE-FCCS, to characterize the oligomerization state of EphA2 in live cells and determine the role of the SAM domain. We deleted the SAM domain in the context of full length EphA2 and an intracellular domain (ICD) construct to assess the effect of the SAM domain on oligomerization state, kinase activity, and cellular behavior. Overall, we find that the SAM domain inhibits ligand-independent clustering and kinase activity in both full-length EphA2 and the isolated ICD construct at the cell membrane. These results are consistent with the autoinhibitory features of the C-terminal tail of EGFR and may help resolve the allosteric regulation of other RTKs.

biophysics↗

bioRxiv2023

Depleting glioblastoma cells of very long-chain acyl-CoA synthetase 3 (ACSVL3) produces metabolic alterations in non-lipid pathways

Knockout (KO) of the fatty acid-activation enzyme very long-chain acyl-CoA synthetase 3 (ACSVL3; SLC27A3) in U87MG glioblastoma cells reduced their malignant growth properties both in vitro and in xenografts. These U87-KO glioma cells grew at a slower rate, became adherence-dependent, and were less invasive than parental U87 cells. U87-KO cells produced fewer, slower-growing subcutaneous and intracranial tumors when implanted in NOD-SCID mice. Thus, depleting U87MG cells of ACSVL3 restored these cells to a phenotype more like that of normal astrocytes. To understand the mechanisms underlying these beneficial changes, we investigated several possibilities, including the effects of ACSVL3 depletion on carbohydrate metabolism. Proteomic and metabolomic profiling indicated that ACSVL3 KO produced changes in glucose and energy metabolism. Even though protein levels of glucose transporters GLUT1 and GLUT3 were reduced by KO, cellular uptake of labeled 2-deoxyglucose was unaffected. Glucose oxidation to CO2 was reduced nearly 7-fold by ACSVL3 depletion, and the cellular glucose level was 25% higher in KO cells. Glycolytic enzymes were upregulated by KO, but metabolic intermediates were essentially unchanged. Surprisingly, lactate production and the levels of lactate dehydrogenase isozymes LDHA and LDHB were elevated by ACSVL3 KO. The activity of the pentose phosphate pathway was found to be lower in KO cells. Citric acid cycle enzymes, electron transport chain complexes, and ATP synthase protein levels were all reduced by ACSVL3 depletion. Mitochondria were elongated in KO cells, but had a more punctate morphology in U87 cells. The mitochondrial potential was unaffected by lack of ACSVL3. We conclude that the beneficial effects of ACSVL3 depletion in human glioblastoma cells may result in part from alterations in diverse metabolic processes that are not directly related to role(s) of this enzyme in fatty acid and/or lipid metabolism. (Supported by NIH 5R01NS062043 and KKI institutional funds.)

biochemistry↗

bioRxiv2023

Relationship between reactive astrocytes, by SMBT-1 imaging, with amyloid-beta, tau, glucose metabolism, and microgliosis in mouse models of Alzheimer's disease

PurposeReactive astrocytes play an important role in the development of Alzheimers disease (AD). Here, we aim to investigate the temporospatial relationship between reactive astrocytes, tau and amyloid-{beta}, glucose metabolism, and microgliosis by using multitracer imaging in AD transgenic mouse models. MethodsPositron emission tomography (PET) imaging with [18F]SMBT-1 (monoamine oxidase-B), [18F]florbetapir (A{beta}), [18F]PM-PBB3 (tau), [18F]FDG, and [18F]DPA-714 (translocator protein) was carried out in 5- and 10-month-old APP/PS1, 11-month-old 3xTg mice, and aged-matched wild-type mice. The brain regional referenced standard uptake value (SUVR) was computed with the cerebellum as the reference region. Immunofluorescence staining was performed in mouse brain tissue slices. Results[18F]SMBT-1 and [18F]florbetapir SUVRs were higher in the cortex and hippocampus of 10-month-old APP/PS1 mice than in 5-month-old APP/PS1 mice and wild-type mice. Reduced [18F]FDG SUVR was observed in the thalamus and midbrain of 5-month-old APP/PS1 mice compared to wild-type mice. No significant difference in brain regional [18F]DPA-714 SUVR was observed in 5- and 10-month-old APP/PS1 mice compared to wild- type mice. No significant difference in the SUVRs of any tracers was observed in 11-month-old 3xTg mice compared to age-matched wild-type mice. A positive correlation between the SUVRs of [18F]SMBT-1 and [18F]DPA-714 in the cortex was observed. Immunostaining validated the distribution of MAO-B and TSPO, amyloid and tau inclusions in brain tissue from 10-month-old APP/PS1 mice and limited changes in 11-month- old 3xTg mice. ConclusionThe findings provide in vivo evidence for reactive astrocytes along with amyloid plaque and tau deposition preceding microgliosis in animal models of AD pathologies.

neuroscience↗

bioRxiv2023

Enhancing Fairness in Disease Prediction with Optimizing Multiple Domain Adversarial Networks

Predictive models in biomedicine need to ensure equitable and reliable outcomes for the populations they are applied to. Unfortunately, biases in medical predictions can lead to unfair treatment and widening disparities, underscoring the need for effective techniques to address these issues. To enhance fairness, we introduce a framework based on a Multiple Domain Adversarial Neural Network (MDANN), which incorporates multiple adversarial components. In an MDANN, an adversarial module is applied to learn a fair pattern by negative gradients back-propagating across multiple sensitive features (i.e., characteristics of individuals that should not be used to discriminate unfairly between individuals when making predictions or decisions.) We leverage loss functions based on the Area Under the Receiver Operating Characteristic Curve (AUC) to address the class imbalance, promoting equitable classification performance for minority groups (e.g., a subset of the population that is underrepresented or disadvantaged.) Moreover, we utilize pre-trained convolutional autoencoders (CAEs) to extract deep representations of data, aiming to enhance prediction accuracy and fairness. Combining these mechanisms, we alleviate biases and disparities to provide reliable and equitable disease prediction. We empirically demonstrate that the MDANN approach leads to better accuracy and fairness in predicting disease progression using brain imaging data for Alzheimers Disease and Autism populations than state-of-the-art techniques.

bioinformatics↗

bioRxiv2023

The tRNA Gm18 Methyltransferase TARBP1 Promotes Hepatocellular Carcinoma Progression via Metabolic Reprogramming of Glutamine

Cancer cells rely on metabolic reprogramming to sustain the prodigious energetic requirements for rapid growth and proliferation. Glutamine metabolism is frequently dysregulated in cancers and is being exploited as a potential therapeutic target. In current study, we identified TARBP1 (TAR (HIV-1) RNA Binding Protein 1) as a novel driver gene critical for glutamine metabolic reprogramming in tumor through the CRISPRi/Cas9 screening. Our in vivo and in vitro assays demonstrated that TARBP1 is the methyltransferase of Guanosine 2-O-methylation targeting position 18 (G18) of tRNAGln (TTG/CTG) and tRNASer (TGA/GCT), and loss of Gm18 modification diminishes the stability of tRNAs. Therefore, TARBP1 is critical for maintaining efficient translation of mRNA, in particular the glutamine transportor-ASCT2 (also known as SCL1A5). Importantly, TARBP1 is frequently amplified and overexpressed in HCC, consequentially promotes the protein synthesis of ASCT2 and glutamine import to fuel the growth of cancer cell, which is associated with poor patient survival. Taken together, this study reveals the critical role of TARBP1 in HCC progression through glutamine metabolic reprogramming and provides a potential target for tumor therapy.

cancer biology↗

bioRxiv2023

High-resolution structure of mouse radial spoke and its in-situ structure in ependymal cilia revealed by cryo-EM and cryo-ET

Radial spokes (RS) transmit mechanochemical signals between the central pair (CP) and axonemal dynein arms to coordinate ciliary motility. However, the high-resolution structure of mammalian RS remains missing. Here, we present the high-resolution cryo-EM structure of mouse RS head-neck complex in both monomer and dimer forms and reveal the intrinsic dynamics of the dimer. We also map the genetic mutations related to primary ciliary dyskinesia and asthenospermia on the head-neck complex. Moreover, we present the unprecedented cryo-ET and sub-tomogram averaging map of mouse ependymal cilia, whose beating creates unidirectional cerebrospinal fluid flow, and build the models for RS1-3, IDAs, and N-DRC. Strikingly, our cryo-ET map reveals the lack of IDA-b/c/e and the absence of tektin filaments within the A-tubule of doublet microtubules in ependymal cilia compared with mammalian respiratory cilia and sperm flagella. This tissue-specific feature may represent an evolutionary choice driven by the functional requirements on ependymal cilia. Our findings reveal the RS head-neck assembly mechanism, shed light on the coordinated rigid and elastic RS-CP interaction modes beneficial for the regulation of asymmetric ciliary beating, and also facilitate understandings on the etiology of ciliary dyskinesia-related ciliopathies.

biophysics↗

bioRxiv2023

Projectome-defined subtypes and modular intra-hypothalamic subnetworks of peptidergic neurons

The hypothalamus plays a vital role in coordinating essential neuroendocrine, autonomic, and somatomotor responses for survival and reproduction. While previous studies have explored population-level projections of hypothalamic neurons, the specific innervation patterns of individual hypothalamic axons remain unclear. To understand the organization of hypothalamic axon projections, we conducted a comprehensive reconstruction of single-cell projectomes from 7,180 mouse hypothalamic neurons expressing specific neuropeptides. Our analysis identified 31 distinct subtypes based on projectome-defined characteristics, with many exhibiting long-range axon collateral projections to multiple brain regions. Notably, these subtypes selectively targeted specific subdomains within downstream areas, either unilaterally or bilaterally. Furthermore, we observed that individual peptidergic neuronal types encompassed multiple projectome-defined subtypes, explaining their diverse functional roles. Additionally, by examining intra-hypothalamic axon projections, we uncovered six modular subnetworks characterized by enriched intramodular connections and distinct preferences for downstream targets. This modular organization of the intra-hypothalamic network likely contributes to the coordinated organization of hypothalamic outputs. In summary, our comprehensive projectome analysis reveals the organizational principles governing hypothalamic axon projections, providing a framework for understanding the neural circuit mechanisms underlying the diverse and coordinated functions of the hypothalamus.

neuroscience↗

bioRxiv2023

Synergistic effects of multiple pesticides on the flight ability of honey bees (Apis cerana F.)

Pesticides cause risks in the environment for Asian honey bees (Apis cerana). Imidacloprid (I), chlorpyrifos (C) and glyphosate (G) are among the most-widely used pesticides in the world. It is not clear on how these pesticides and combination of them affect the flight ability of A. cerana. Here we utilized flight mills to show new results that acute treatment of individual pesticides, such as 0.2 ng/bees imidacloprid (20 L, 10 ng/g), 0.6 ng/bees chlorpyrifos (20 L, 30 ng/g) and 1.2 ng/bees glyphosate (20 L, 60 ng/g), had no effect on the flight ability of bees. However, forager bees showed a significantly decrease in the flying duration and flying distance when oral exposed to two or three these pesticides. This evidence indicated that two or three pesticides can produce synergistic changes in the flight ability and behavior of honeybees. Results showed a light on new understandings of complex effects and potential risks of these three pesticide on bee behavior including homing ability and food-collecting ability. Our results are key information to understand new synergistic potential among pesticide formulations and how they impair bee behavior. Table of Contents Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=104 SRC="FIGDIR/small/541595v1_ufig1.gif" ALT="Figure 1"> View larger version (23K): org.highwire.dtl.DTLVardef@70e24dorg.highwire.dtl.DTLVardef@1195fb8org.highwire.dtl.DTLVardef@db5ad5org.highwire.dtl.DTLVardef@1287b51_HPS_FORMAT_FIGEXP M_FIG C_FIG

animal behavior and cognition↗

bioRxiv2023

A comprehensive catalog of 3D genome organization in diverse human genomes facilitates understanding of the impact of structural variation on chromatin structure

The human genome is packaged into the three-dimensional (3D) nucleus and organized into functional units known as topologically associating domains (TADs) and chromatin loops. Recent studies show that the 3D genome can be modified by genome structural variants (SVs) through disrupting higher-order chromatin organizations such as TADs, which play an essential role in insulating genes from aberrant regulation by regulatory elements outside TADs. Here, we have developed an integrative Hi-C analysis pipeline to generate a comprehensive catalog of TADs, TAD boundaries, and loops in human genomes to fill the gap of limited resources. We identified 2,293 TADs and 6,810 sub-TADs missing in the previously released TADs of GM12878. We then quantified the impact of SVs overlapping with TAD boundaries and observed that two SVs could significantly alter chromatin architecture leading to abnormal expression and splicing of genes associated with human diseases.

genomics↗

bioRxiv2023

VERY LONG-CHAIN ACYL-CoA SYNTHETASE-3 (ACSVL3) PROMOTES THE MALIGNANT GROWTH BEHAVIOR OF U87 GLIOMA CELLS VIA CHANGES IN CELL CYCLE WITHOUT AFFECTING APOPTOSIS

Decreasing the expression of very long-chain acyl-CoA synthetase 3 (ACSVL3) in U87MG glioblastoma cells by either RNA interference or genomic knockout (KO) significantly decreased their growth rate in culture, as well as their ability to form rapidly growing tumors in mice. U87-KO cells grew at a 9-fold slower rate than U87MG cells. When injected subcutaneously in nude mice, the tumor initiation frequency of U87-KO cells was 70% of that of U87MG cells, and the average growth rate of tumors that did form was decreased by 9-fold. Two hypotheses to explain the decreased growth rate of KO cells were investigated. Lack of ACSVL3 could reduce cell growth either by increasing apoptosis, or via effects on the cell cycle. We examined intrinsic, extrinsic, and caspase-independent apoptosis pathways; none were affected by lack of ACSVL3. However, significant differences in the cell cycle were seen in KO cells, suggesting arrest in S-phase. Levels of cyclin-dependent kinases 1, 2, and 4 were elevated in U87-KO cells, as were regulatory proteins p21 and p53 that promote cell cycle arrest. In contrast, lack of ACSVL3 reduced the level of the inhibitory regulatory protein p27. {gamma}-H2AX, a marker of DNA double strand breaks, was elevated in U87-KO cells, while pH3, a mitotic index marker, was reduced. Previously reported alterations in sphingolipid metabolism in ACSVL3-depleted U87 cells may explain the effect of KO on cell cycle. These studies reinforce the notion that ACSVL3 is a promising therapeutic target in glioblastoma.

biochemistry↗

bioRxiv2023

ZBP1-mediated macrophage necroptosis inhibits ASFV replication

African swine fever (ASF) is an infectious disease characterized by hemorrhagic fever, highly pathogenic, and severe mortality in domestic pigs caused by the African swine fever virus (ASFV). ASFV is a large DNA virus and primarily infects porcine monocyte macrophages. The interaction between ASFV and host macrophages is the major reason for gross pathological lesions caused by ASFV. Necroptosis is an inflammatory programmed cell death and plays an important immune role during virus infection. However, whether and how ASFV induces macrophage necroptosis and what effect of necroptosis signaling on host immunity and ASFV infection remains unknown. This study uncovered that ASFV activates the necroptosis signaling in the spleen, lung, liver, and kidney from ASFV-infected pigs. And macrophage necroptosis also was induced by ASFV infection in vitro. Further evidence showed that the macrophage necroptosis is independent of TNF--RIPK1 or TLR-TRIF pathway but depends on Z-DNA binding protein 1 (ZBP1). ASFV infection upregulates the expression of ZBP1 and RIPK3 to consist of the ZBP1-RIPK3-MLKL necrosome and further activates macrophage necroptosis. Subsequently, multiple Z-DNA sequences were predicted to be present in the ASFV genome. And the Z-DNA signals were further confirmed to be present and colocalized with ZBP1 in the cytoplasm and nucleus of ASFV-infected cells. Moreover, ZBP1-mediated macrophage necroptosis caused the extracellular release of proinflammatory cytokines TNF- and IL-1{beta} induced by ASFV infection. Finally, we demonstrated that ZBP1-mediated necroptosis signaling significantly inhibits ASFV replication in host macrophages. Our findings uncovered a novel mechanism by which ASFV induces macrophage necroptosis by facilitating Z-DNA accumulation and ZBP1 necrosome assembly, providing significant insights into the pathogenesis of ASFV infection. ImportanceASFV infection causes an acute, febrile, hemorrhagic, and severely lethal disease in swine, which seriously threatens the global porcine industry. Understanding the interaction mechanism between ASFV and host macrophages during infection is essential for elucidating the pathogenesis of ASFV. To our knowledge, this study revealed the interaction mechanism between ASFV infection and host macrophage necroptosis. The results showed that ASFV infection induces macrophage necroptosis through ZBP1 activation instead of the TNF--RIPK1 or TLR-TRIF pathway. ASFV infection promotes Z-DNA accumulation, which causes ZBP1-RIPK3-MLKL necrosome assembly and macrophage necroptosis. The ZBP1-mediated necroptosis signaling facilitates the extracellular release of proinflammatory cytokines, inhibiting ASFV replication in host macrophages. This study found a new interaction mechanism between ASFV and host macrophages, which may help understand the pathogenesis caused by ASFV infection.

microbiology↗

bioRxiv2023

Using macromolecular electron densities to improve the enrichment of active compounds in virtual screening

The quest for more effective virtual screening algorithms is hindered by the scarcity of training data, calling for innovative approaches. This study presents the first use of experimental electron density (ED) data for improving active compound enrichment in virtual screening, supported by EDs ability to reflect the time-averaged behavior of ligands and solvents in the binding pocket. Experimental ED-based grid matching score (ExptGMS) was developed to score compounds by measuring the degree of matching between their binding conformations and a series of multi-resolution experimental ED grids. The efficiency of ExptGMS was validated using both in-silico tests with the Directory of Useful Decoys-Enhanced dataset and wet-lab tests on Covid-19 3CLpro-inhibitors. ExptGMS improved the active compound enrichment in top-ranked molecules by approximately 20%. Furthermore, ExptGMS helped identify four new and active inhibitors of 3CLpro, with the top showing an IC50 value of 1.9 {micro}M. To facilitate the use of ExptGMS, we developed an online database containing experimental ED grids for over 17,000 proteins.

bioinformatics↗

bioRxiv2023

Hepatic conversion of acetyl-CoA to acetate plays crucial roles in energy stresses

Accumulating evidences indicate that acetate is increased in energy stresses such as diabetes mellitus and prolonged starvation. However, it is largely unknown how and where acetate is produced and what is its biological significance. We observed overproduction of acetate in an amount comparable to ketone bodies in patients and mice with diabetes or starvation. Mechanistically, ACOT 12&8 are dramatically upregulated in liver to convert FFA-derived acetyl-CoA to acetate and CoA. This conversion not only provides large amount of acetate which fuels brain preferentially rather than muscle, but also recycles CoA which is required for sustained fatty acid oxidation and ketogenesis. Taken together, we suggest that acetate is an emerging novel "ketone body" and may be used as a parameter to evaluate the progression of energy stress in the future.

biochemistry↗

bioRxiv2023

E2EGraph: An End-to-end Graph Learning Model for Interpretable Prediction of Pathlogical Stages in Prostate Cancer

Prostate cancer is one of the deadliest cancers worldwide. An accurate prediction of pathological stages using the expressions and interactions of genes is effective for clinical assessment and treatment. However, identification of interactions using biological procedure is time consuming and prohibitively expensive. A graph is a powerful representation for the complex interactome of genes, their transcripts, and proteins. Recently, Graph Neural Networks (GNNs) have gained great attention in machine learning due to their capability to capture the graphical interactions among data entities. To leverage GNNs for predicting pathological stage stages, we developed an end-to-end graph representation and learning model, namely E2EGraph, which can automatically generate a graph representation using gene expression data and a multi-head graph attention network to learn the strength of interactions among genes and make the prediction. To ensure the reliability of model prediction, we identify critical components of graph representation and GNN model to interpret prediction results from multiple perspectives at gene and patient levels. We evaluated E2EGraph to predict pathological stages of prostate cancer using The Cancer Genome Atlas (TCGA) data. Our experimental results demonstrate that E2EGraph reaches the state-of-art prediction performance while being effective in identifying marker genes indicated by interpretability. Our results point to a direction where adaptive graph construction and attention based GNNs can be leveraged for various prediction tasks and interpretation of model prediction in a variety of data domains including disease prediction.

bioinformatics↗

bioRxiv2023

CYP4F2 is a human-specific determinant of circulating N-acyl amino acid levels

N-acyl amino acids are a large family of circulating lipid metabolites that modulate energy expenditure and fat mass in rodents. However, little is known about the regulation and potential cardiometabolic functions of N-acyl amino acids in humans. Here, we analyze the cardiometabolic phenotype associations and genetic regulation of four plasma N-fatty acyl amino acids (N-oleoyl-leucine, N-oleoyl-phenylalanine, N-oleoyl-serine, and N-oleoyl-glycine) in 2,351 individuals from the Jackson Heart Study. N-oleoyl-leucine and N-oleoyl-phenylalanine were positively associated with traits related to energy balance, including body mass index, waist circumference, and subcutaneous adipose tissue. In addition, we identify the CYP4F2 locus as a human-specific genetic determinant of plasma N-oleoyl-leucine and N-oleoyl-phenylalanine levels. In vitro, CYP4F2-mediated hydroxylation of N-oleoyl-leucine and N-oleoyl-phenylalanine results in metabolic diversification and production of many previously unknown lipid metabolites with varying characteristics of the fatty acid tail group, including several that structurally resemble fatty acid hydroxy fatty acids (FAHFAs). By contrast, FAAH-regulated N-oleoyl-glycine and N-oleoyl-serine were inversely associated with traits related to glucose and lipid homeostasis. These data uncover a human-specific enzymatic node for the metabolism of a subset of N-fatty acyl amino acids and establish a framework for understanding the cardiometabolic roles of individual N-fatty acyl amino acids in humans.

biochemistry↗

bioRxiv2023

Split-Transformer Impute (STI): Genotype Imputation Using a Transformer-Based Model

MotivationDespite recent advances in sequencing technologies, genome-scale datasets continue to have missing bases and genomic segments. Such incomplete datasets can undermine downstream analyses, such as disease risk prediction and association studies. Consequently, the imputation of missing information is a common pre-processing step for which many methodologies have been developed. However, the imputation of genotypes of certain genomic regions and variants, including large structural variants, remains a challenging problem. ResultsHere, we present a transformer-based deep learning framework, called a split-transformer impute (STI) model, for accurate genome-scale genotype imputation. Empowered by the attention-based transformer model, STI can be trained for any collection of genomes automatically using self-supervision. STI handles multi-allelic genotypes naturally, unlike other models that need special treatments. STI models automatically learned genome-wide patterns of linkage disequilibrium (LD), evidenced by much higher imputation accuracy in high LD regions. Also, STI models trained through sporadic masking for self-supervision performed well in imputing systematically missing information. Our imputation results on the human 1000 Genomes Project show that STI can achieve high imputation accuracy, comparable to the state-of-the-art genotype imputation methods, with the additional capability to impute multi-allelic structural variants and other types of genetic variants. Moreover, STI showed excellent performance without needing any special presuppositions about the patterns in the underlying data when applied to a collection of yeast genomes, pointing to easy adaptability and application of STI to impute missing genotypes in any species.

genomics↗