Biology subjects

Jiang, Y.

Publications and source records attributed to Jiang, Y..

At least 325 records · Page 18Linked to original sources

bioRxiv2023

Rational engineering of an elevator-type metal transporter ZIP8 reveals a conditional selectivity filter critically involved in determining substrate specificity

Engineering of transporters to alter substrate specificity as desired holds great potential for applications, including metabolic engineering. However, the lack of knowledge on molecular mechanisms of substrate specificity hinders designing effective strategies for transporter engineering. Here, we applied an integrated approach to rationally alter the substrate preference of ZIP8, a Zrt-/Irt-like protein (ZIP) metal transporter with multiple natural substrates, and uncovered the determinants of substrate specificity. By systematically replacing the differentially conserved residues with the counterparts in the zinc transporter ZIP4, we created a zinc-preferring quadruple variant (Q180H/E343H/C310A/N357H), which exhibited largely reduced transport activities towards Cd2+, Fe2+, and Mn2+ whereas increased activity toward Zn2+. Combined mutagenesis, modeling, covariance analysis, and computational studies revealed a conditional selectivity filter which functions only when the transporter adopts the outward-facing conformation. The demonstrated approach for transporter engineering and the gained knowledge about substrate specificity will facilitate engineering and mechanistic studies of other transporters.

biochemistry↗

bioRxiv2023

scATAnno: Automated Cell Type Annotation for single-cell ATAC-seq Data

Recent advances in single-cell epigenomic techniques have created a growing demand for scATAC-seq analysis. One key analysis task is to determine cell type identity based on the epigenetic data. We introduce scATAnno, a python package designed to automatically annotate scATAC-seq data using large-scale scATAC-seq reference atlases. This workflow generates the reference atlases from publicly available datasets enabling accurate cell type annotation by integrating query data with reference atlases, without the use of scRNA-seq data. To enhance annotation accuracy, we have incorporated KNN-based and weighted distance-based uncertainty scores to effectively detect cell populations within the query data that are distinct from all cell types in the reference data. We compare and benchmark scATAnno against 7 other published approaches for cell annotation and show superior performance in multiple data sets and metrics. We showcase the utility of scATAnno across multiple datasets, including peripheral blood mononuclear cell (PBMC), Triple Negative Breast Cancer (TNBC), and basal cell carcinoma (BCC), and demonstrate that scATAnno accurately annotates cell types across conditions. Overall, scATAnno is a useful tool for scATAC-seq reference building and cell type annotation in scATAC-seq data and can aid in the interpretation of new scATAC-seq datasets in complex biological systems.

bioinformatics↗

bioRxiv2023

Haplotype-resolve genome assembly and resequencing provide insights into the origin and domestication of modern rose

Modern rose (Rosa hybrida) is a recently formed interspecific hybrid and has become one of the most important and widely cultivated ornamentals. Here, we report the haplotype-resolved chromosome-scale genome assembly of the tetraploid R. hybrida Samantha and a genome variation map of 233 Rosa accessions involving various wild species, and old and modern cultivars. The two subgenomes of Samantha show no significant global bias in gene loss but substantial expression bias, and homoeologous exchanges are frequently observed between subgenomes. Population genomic and genomic composition analyses reveal the contributions of wild Rosa species to modern roses, and highlight that R. odorata and its derived cultivars contribute more to modern roses than R. chinensis Old Blush. Furthermore, selective sweeps during modern rose breeding associated with major agronomic traits, including continuous flowering, floral organ identity, flower senescence, and disease resistance, are identified. This study provides insights into the genetic basis of modern rose origin and breeding history, and offers unprecedented genomic resources for rose improvement.

genomics↗

bioRxiv2023

The impact of common variants on gene expression in the human brain: from RNA to protein to schizophrenia risk

BackgroundThe impact of genetic variants on gene expression has been intensely studied at the transcription level, yielding invaluable insights into the association between genes and the risk of complex disorders, such as schizophrenia (SCZ). However, the downstream impact of these variants and the molecular mechanisms connecting transcription variation to disease risk are not well understood. ResultsWe quantitated ribosome occupancy in prefrontal cortex samples of the BrainGVEX cohort. Together with transcriptomics and proteomics data from the same cohort, we performed cis-Quantitative Trait Locus (QTL) mapping and identified, at 10% false discovery rate, 3,253 expression QTLs (eQTLs), 1,344 ribosome occupancy QTLs (rQTLs), and 657 protein QTLs (pQTLs) out of 7,458 genes from 185 samples. Of the eQTLs identified, only 34% have their effects propagated to the protein level. Further analysis on the effect size of prefrontal cortex eQTLs identified from an independent dataset clearly replicated the post-transcriptional attenuation of eQTL effects. We identified omics-specific QTLs and investigated their potential in driving disease risks. Using both a variant-based approach and a gene-based approach, we identified genes containing expression-specific QTLs (esQTLs), ribosome-occupancy-specific QTLs (rsQTLs), and protein-specific QTLs (psQTLs). Among the variant-based omics-specific QTL, 38 showed strong colocalization with brain associated disorder GWAS signals, 29 of them are esQTLs. From the gene-based approach, we found 11 brain associated disorder risk genes that are driven predominantly by omics-specific QTL, all of them are driven by variants impacting transcriptional regulation. To take a complementary approach to further investigate the functional relevance of genes driven predominantly by attenuated eQTL signals, we identified SCZ risk genes using each omics independently and then investigated the omics-specificity of the driver regulatory process for each risk gene. Using S-PrediXcan we identified 74 SCZ risk genes across the three omics, 30% of which were novel, and 67% of these risk genes were confirmed to be causal in a MR-Egger test. Notably, 52 out of the 74 risk genes were identified using eQTL data and 68% of these SCZ-risk-gene-driving eQTLs show little to no evidence of driving corresponding variations at the protein level. ConclusionThe effect of eQTLs on gene expression in the prefrontal cortex is commonly attenuated post-transcriptionally. Many of the attenuated eQTLs still correlate with GWAS signals of brain associated complex disorders, indicating the possibility that these eQTL variants drive disease risk through mechanisms other than regulating protein expression level. Further investigation is needed to elucidate the mechanistic link between attenuated eQTLs and brain associated complex disorders.

genetics↗

bioRxiv2023

AutoUnmix: an autoencoder-based spectral unmixing method for multi-color fluorescence microscopy imaging

Multiplexed fluorescence microscopy imaging is widely used in biomedical applications. However, simultaneous imaging of multiple fluorophores can result in spectral leaks and overlapping, which greatly degrades image quality and subsequent analysis. Existing popular spectral unmixing methods are mainly based on computational intensive linear models and the performance is heavily dependent on the reference spectra, which may greatly preclude its further applications. In this paper, we propose a deep learning-based blindly spectral unmixing method, termed AutoUnmix, to imitate the physical spectral mixing process. A tranfer learning framework is further devised to allow our AutoUnmix adapting to a variety of imaging systems without retraining the network. Our proposed method has demonstrated real-time unmixing capabilities, surpassing existing methods by up to 100-fold in terms of unmixing speed. We further validate the reconstruction performance on both synthetic datasets and biological samples. The unmixing results of AutoUnmix achieve a highest SSIM of 0.99 in both three- and four-color imaging, with nearly up to 20% higher than other popular unmixing methods. Due to the desirable property of data independency and superior blind unmixing performance, we believe AutoUnmix is a powerful tool to study the interaction process of different organelles labeled by multiple fluorophores.

cell biology↗

bioRxiv2023

Rapid intracellular acidification is a novel plant defense response countered by the brown planthopper

The brown planthopper (BPH) is the most destructive insect pest in rice. Through a stylet, BPH secrets a plethora of salivary proteins into rice phloem cells as a crucial step of infestation. However, how various salivary proteins function in rice cells to promote insect infestation is poorly understood. Among them, one of the salivary proteins is predicted to be a carbonic anhydrase (NlCA). The survival rate of the NlCA-RNA interference (RNAi) BPH insects was extremely low on rice, indicating a vital role of this salivary protein in BPH infestation. We generated NICA transgenic rice plants and found that NlCA expressed in rice plants could restore the ability of NlCA-RNAi BPH to survive on rice. Next, we produced rice plants expressing the ratiometric pH sensor pHusion and found that NlCA-RNAi BPH induced rapid intracellular acidification of rice cells during feeding. Further analysis revealed that both NlCA-RNAi BPH feeding and artificial of lowering intracellular pH activated plant defense responses, and that NICA-mediated intracellular pH stabilization is linked to diminished defense responses, including reduced callose deposition at the phloem sieve plates and suppressed defense gene expression. Given the importance of pH homeostasis across the kingdoms of life, discovery of NICA-mediated intracellular pH modulation uncovered a new dimension in the interaction between plants and piecing/sucking insect pests. The crucial role of NICA for BPH infestation of rice suggests that NICA is a promising target for chemical or trans-kingdom RNAi-based inactivation for BPH control strategies in plants. HighlightsO_LIRapid intracellular acidification is a novel plant defense response. C_LIO_LIBrown planthopper delivers into rice tissues a major virulence protein, carbonic anhydrase (NlCA), critical for survival. C_LIO_LINlCA counters intracellular acidification to suppress plant defense. C_LIO_LIResults implicate existence of pH-responsive regulators of defense responses in the rice-brown planthopper interaction. C_LI In briefThe brown planthopper (BPH) is the most devasting insect pest in rice. Jiang et al. discovered that BPH secretes a salivary carbonic anhydrase (NICA) to regulate the intracellular pH of the rice cell to facilitate its feeding and survival on rice plants and that NICA-counters host intracellular pH acidification to diminished defense responses. These findings uncovered that intracellular pH homeostasis is a previously uncharacterized battleground in plant-insect interactions, but also open a door to future discovery of pH-responsive intracellular regulators of defense responses, which could add a new research dimension in the study of plant-biotic interactions.

plant biology↗

bioRxiv2023

The Notch-Pdgfr beta axis suppresses brown adipocyte progenitor differentiation in early postnatal mice

De novo brown adipogenesis holds potential in combating the epidemics of obesity and diabetes. However, the identity of brown adipocyte progenitor cells (APCs) and their regulation have not been extensively studied. Here through in vivo lineage tracing, we observed that PDGFR{beta}+ pericytes give rise to developmental brown adipocytes, but not to those in adult homeostasis. In contrast, TBX18+ pericytes contribute to brown adipogenesis throughout both developmental and adult stages, though in a depot-specific manner. Mechanistically, Notch inhibition in PDGFR{beta}+ pericytes promotes brown adipogenesis through the downregulation of PDGFR{beta}. Furthermore, inhibition of Notch signaling in PDGFR{beta}+ pericytes mitigates HFHS (high-fat, high-sucrose) induced glucose and metabolic impairment in both developmental and adult stages. Collectively, these findings show that the Notch/PDGFR{beta} axis negatively regulates developmental brown adipogenesis, and its repression promotes brown adipose tissue expansion and improves metabolic health. HighlightsO_LIPDGFR{beta}+ pericytes act as an essential developmental brown APC. C_LIO_LITBX18+ pericytes contribute to brown adipogenesis in a depot-specific manner. C_LIO_LIInhibiting Notch-Pdgfr{beta} axis promotes brown APC adipogenesis. C_LIO_LIEnhanced postnatal brown adipogenesis improves metabolic health in adult stage. C_LI

developmental biology↗

bioRxiv2023

The ENCODE4 long-read RNA-seq collection reveals distinct classes of transcript structure diversity

The majority of mammalian genes encode multiple transcript isoforms that result from differential promoter use, changes in exonic splicing, and alternative 3 end choice. Detecting and quantifying transcript isoforms across tissues, cell types, and species has been extremely challenging because transcripts are much longer than the short reads normally used for RNA-seq. By contrast, long-read RNA-seq (LR-RNA-seq) gives the complete structure of most transcripts. We sequenced 264 LR-RNA-seq PacBio libraries totaling over 1 billion circular consensus reads (CCS) for 81 unique human and mouse samples. We detect at least one full-length transcript from 87.7% of annotated human protein coding genes and a total of 200,000 full-length transcripts, 40% of which have novel exon junction chains. To capture and compute on the three sources of transcript structure diversity, we introduce a gene and transcript annotation framework that uses triplets representing the transcript start site, exon junction chain, and transcript end site of each transcript. Using triplets in a simplex representation demonstrates how promoter selection, splice pattern, and 3 processing are deployed across human tissues, with nearly half of multitranscript protein coding genes showing a clear bias toward one of the three diversity mechanisms. Evaluated across samples, the predominantly expressed transcript changes for 74% of protein coding genes. In evolution, the human and mouse transcriptomes are globally similar in types of transcript structure diversity, yet among individual orthologous gene pairs, more than half (57.8%) show substantial differences in mechanism of diversification in matching tissues. This initial large-scale survey of human and mouse long-read transcriptomes provides a foundation for further analyses of alternative transcript usage, and is complemented by short-read and microRNA data on the same samples and by epigenome data elsewhere in the ENCODE4 collection.

genomics↗

bioRxiv2023

Radiotherapy induces persistent innate immune reprogramming of microglia into a primed state

More than half of all brain tumour survivors experience debilitating and often progressive cognitive decline after treatment with radiotherapy. Microglia, the resident macrophages in the brain, have been implicated in this decline. In response to various insults microglia can develop innate immune memory (IIM), which can either enhance (priming) or repress (tolerance) the response to subsequent inflammatory challenges. Here, we investigated whether radiation affects the IIM of microglia by irradiating the brains of rats and later exposing them to a secondary inflammatory stimulus. Comparative transcriptomic profiling and protein validation of microglia isolated from irradiated rats showed a stronger immune response to a secondary inflammatory insult demonstrating that radiation can lead to long-lasting molecular reprogramming of microglia. Transcriptomic analysis of post-mortem normal-appearing non-tumour brain tissue of glioblastoma patients indicates that radiation-induced microglial priming is conserved in humans. Targeting microglial priming after radiotherapy or avoiding further inflammatory insults could decrease radiotherapy-induced neurotoxicity.

neuroscience↗

bioRxiv2023

Critical role of down-regulated in adenoma bicarbonate transporter in linaclotide stimulated intestinal bicarbonate secretion

Duodenal bicarbonate secretion is critical to epithelial protection, nutrient digestion/absorption and is impaired in cystic fibrosis (CF). We examined if linaclotide, typically used to treat constipation, may also stimulate duodenal bicarbonate secretion. Bicarbonate secretion was measured in vivo and in vitro using mouse and human duodenum (biopsies and enteroids). Ion transporter localization was identified with confocal microscopy and de novo analysis of human duodenal single cell RNA sequencing (sc-RNAseq) datasets was performed. Linaclotide increased bicarbonate secretion in mouse and human duodenum in the absence of CFTR expression (Cftr knockout mice) or function (CFTRinh-172). NHE3 inhibition contributed to a portion of this response. Linaclotide-stimulated bicarbonate secretion was eliminated by down-regulated in adenoma (DRA, SLC26A3) inhibition during loss of CFTR activity. Sc-RNAseq identified that 70% of villus cells expressed SLC26A3, but not CFTR, mRNA. Loss of CFTR activity and linaclotide increased apical brush border expression of DRA in non-CF and CF differentiated enteroids. These data provide further insights into the action of linaclotide and how DRA may compensate for loss of CFTR in regulating luminal pH. Linaclotide may be a useful therapy for CF individuals with impaired bicarbonate secretion.

physiology↗

bioRxiv2023

An Excitatory Projection from the Basal Forebrain to the Ventral Tegmental Area Underlying Anorexia-like Phenotypes

Adaptation to potential threat cues in feeding regulation is key for animal survival. Maladaptation in balancing internal energy needs and external threat cues may result in eating disorders. However, brain mechanisms underlying such maladaptation remain elusive. Here, we identified that the basal forebrain (BF) sends glutamatergic projections to glutamatergic neurons in the ventral tegmental area (VTA). Glutamatergic neurons in both the BF and the VTA displayed correlated responses to various external stressors. Notably, in vivo manipulation of BF terminals in the VTA revealed that the glutamatergic BFVTA circuit reduces appetite, increases locomotion, and elicits avoidance. In consistence, activation of VTA glutamatergic neurons reduced body weight, blunted food motivation, and caused hyperactivity with behavioral signs of anxiety, all hallmarks of typical anorexia phenotypes. Importantly, activation of BF glutamatergic terminals in the VTA reduced dopamine release in the nucleus accumbens (NAc). Collectively, our results point to overactivation of the glutamatergic BFVTA circuit as a potential cause of anorexia-like phenotypes involving reduced dopamine release.

neuroscience↗

bioRxiv2023

Post-translational modification of SPATULA by SECRET AGENT and SPINDLY promotes organ symmetry transition at the gynoecium apex

The establishment of organ symmetry during multicellular development is a fundamental process shared by most living organisms. Here, we investigated how two O-glycosyltransferases of Arabidopsis thaliana, SPINDLY (SPY) and SECRET AGENT (SEC) synergistically promote a rare bilateral-to-radial symmetry transition during patterning of the plant reproductive organ, the gynoecium. SPY and SEC modify N-terminal residues of the bHLH transcription factor SPATULA (SPT) in vivo and in vitro by attaching O-fucose and O-linked-{beta}-N-Acetylglucosamine (O-GlcNAc), to promote style development. This post-translational regulation does not impact SPT homo- and hetero-dimerisation events with INDEHISCENT (IND) and HECATE 1 (HEC1), although it enhances the affinity of SPT for the kinase PINOID (PID) gene locus to promote transcriptional repression. Our findings reveal a previously unrecognized mechanism for O-GlcNAc and O-fucose post-translational decorations in controlling style development and offer the first molecular example of a synergistic role for SEC and SPY in plant post-embryonic organ patterning.

plant biology↗

bioRxiv2023

A genetic mosaic mouse model illuminates the pre-malignant progression of basal-like breast cancer

Basal-like breast cancer is an aggressive breast cancer subtype, often characterized by a deficiency in BRCA1 function and concomitant loss of p53. While conventional mouse models enable the investigation of its malignant stages, one that reveals its initiation and pre-malignant progression is lacking. Here, we leveraged a mouse genetic system known as Mosaic Analysis with Double Markers (MADM) to generate rare GFP-labeled Brca1, p53-deficient cells alongside RFP+ wildtype sibling cells in the mammary gland. The mosaicism resembles the sporadic initiation of human cancer and enables spatially resolved analysis of mutant cells in comparison to paired wildtype sibling cells. Mammary tumors arising in the model show transcriptomic and genomic characteristics similar to human basal-like breast cancer. Analysis of GFP+ mutant cells at interval time points before malignancy revealed a stepwise progression of lesions from focal expansion to hyper-alveolarization and then to micro-invasion. These stereotyped morphologies indicate the pre-malignant stage irrespective of the time point at which it is observed. Paired analysis of GFP-RFP siblings during focal expansion suggested that hyper-alveolarized structures originate from ductal rather than alveolar cells, despite their morphological similarities to alveoli. Evidence for luminal-to-basal transition at the pre-malignant stages was restricted to cells that had escaped hyper-alveoli and progressed to micro-invasive lesions. Our MADM-based mouse model presents a useful tool for studying the pre-malignancy of basal-like breast cancer. Summary statementA mouse model recapitulates the process of human basal-like breast tumorigenesis initiated from sporadic Brca1, p53-deficient cells, empowering spatially-resolved analysis of mutant cells during pre-malignant progression.

cancer biology↗

bioRxiv2023

How hibernation in frogs drives brain and reproductive evolution in opposing directions

Environmental seasonality can promote the evolution of larger brains through cognitive and behavioral flexibility but can also hamper it when temporary food shortage is buffered by stored energy. Multiple hypotheses linking brain evolution with resource acquisition and allocation have been proposed for warm-blooded organisms, but it remains unclear how these extend to cold-blooded taxa whose metabolism is tightly linked to ambient temperature. Here, we integrated these hypotheses across frogs and toads in the context of varying brumation ( hibernation) durations and their environmental correlates. We showed that protracted brumation covaried negatively with brain size but positively with reproductive investment, likely in response to brumation-dependent changes in the socio-ecological context and associated selection on different tissues. Our results provide novel insights into resource allocation strategies and possible constraints in trait diversification, which may have important implications for the adaptability of species under sustained environmental change.

evolutionary biology↗

bioRxiv2023

A Comprehensive Benchmark Study on Biomedical Text Generation and Mining with ChatGPT

In recent years, the development of natural language process (NLP) technologies and deep learning hardware has led to significant improvement in large language models(LLMs). The ChatGPT, the state-of-the-art LLM built on GPT-3.5, shows excellent capabilities in general language understanding and reasoning. Researchers also tested the GPTs on a variety of NLP related tasks and benchmarks and got excellent results. To evaluate the performance of ChatGPT on biomedical related tasks, this paper presents a comprehensive benchmark study on the use of ChatGPT for biomedical corpus, including article abstracts, clinical trials description, biomedical questions and so on. Through a series of experiments, we demonstrated the effectiveness and versatility of Chat-GPT in biomedical text understanding, reasoning and generation.

bioinformatics↗

bioRxiv2023

S2' cleavage site plays a decisive role in expansion of Vero cell tropism by infectious bronchitis virus HV80, with Q855H promoting cell-to-cell fusion

Infectious bronchitis virus (IBV) has restricted cell tropism. Apart from the Beaudette strain, other IBVs cannot infect mammalian cell lines. The limited cell tropism of other IBVs has hindered the development of IBV vaccines and research on mechanisms of IBV infection. In a previous study, a new Vero-cell-adapted strain HV80 was obtained via serial chicken embryo and cell passaging of strain H120 and 17 mutations leading to amino acid substitutions occurred in replication gene 1a, S gene and E gene. This study, we constructed recombinants that expressed chimeric S glycoprotein, S1 or S2 subunit of strain H120, and demonstrated that mutations in S2 subunit were related to the Vero cell adaption of strain HV80. With a genome backbone of strain HV80 or H120, and expression of chimeric S2' cleavage site of H120 or HV80, two recombinants demonstrated that the RRRR690/S motif at the S2' cleavage site played a key role in Vero cell adaption of strain HV80. Another six amino acid substitutions in the S2 subunit of the recombinants showed that F692V enhanced the capability of invasion of HV80 strain, and Q855H induced the formation of syncytia. A transient transfection assay demonstrated different mechanisms for virus-to-cell fusion and cell-to-cell fusion induced by S glycoprotein. The PRRR690/S motif at the S2' cleavage site could be activated by proteases in the process of cell-to-cell fusion, while H855Q substitution did not affect the cell invasion of HV80, but hindered the cell-to-cell fusion by blocking activation of the S2' cleavage site. IMPORTANCEInfectious bronchitis is an acute respiratory disease that has caused large economic losses to the poultry industry. As a member of the gamma-coronaviruses, the restricted cell tropism of infectious bronchitis virus (IBV) limits the development of cellular vaccines and research on infection mechanisms. As a strain that can replicate effectively in mammalian cell lines, studies of HV80s adaptive mechanisms point a way for engineering other IBVs for adaptation in mammalian cell lines. In our study, different recombinants were constructed by reverse genetic techniques, and demonstrated the different mechanism between virus-to-cell and cell-to-cell fusion induced by HV80 S glycoprotein. The acquisition of a highly efficient S2' cleavage site enabled the virus to invade Vero cells. The Q855H substitution played a key role in cell-to-cell fusion, and provided a more efficient model of infection in Vero cells. Our study provides new theoretical insights into mechanisms of IBV adaptation in mammalian cell lines.

microbiology↗

bioRxiv2023

ME-Bayes SL: Enhanced Bayesian Polygenic Risk Prediction Leveraging Information across Multiple Ancestry Groups

Polygenic risk scores (PRS) are now showing promising predictive performance on a wide variety of complex traits and diseases, but there exists a substantial performance gap across different populations. We propose MUSSEL, a method for ancestry-specific polygenic prediction that borrows information in the summary statistics from genome-wide association studies (GWAS) across multiple ancestry groups. MUSSEL conducts Bayesian hierarchical modeling under a MUltivariate Spike-and-Slab model for effect-size distribution and incorporates an Ensemble Learning step using super learner to combine information across different tuning parameter settings and ancestry groups. In our simulation studies and data analyses of 16 traits across four distinct studies, totaling 5.7 million participants with a substantial ancestral diversity, MUSSEL shows promising performance compared to alternatives. The method, for example, has an average gain in prediction R2 across 11 continuous traits of 40.2% and 49.3% compared to PRS-CSx and CT-SLEB, respectively, in the African Ancestry population. The best-performing method, however, varies by GWAS sample size, target ancestry, underlying trait architecture, and the choice of reference samples for LD estimation, and thus ultimately, a combination of methods may be needed to generate the most robust PRS across diverse populations.

genetics↗

bioRxiv2023

TCR2vec: a deep representation learning framework of T-cell receptor sequence and function

The T-cell receptor (TCR) repertoires are critical components of the adaptive immune system, and machine learning methods were proposed to analyze the TCR repertoire data. However, most methods work solely on the hypervariable CDR3 regions of TCRs, overlooking the information encoded in other domains. Representing full TCRs as informative vectors can be indispensable for developing reliable and effective machine learning models. We introduce TCR2vec, a deep representation learning framework with 12 layers of transformer blocks, to pave the way for downstream modelings of full TCRs. Together with masked language modeling (MLM), we propose a novel pretraining task named similarity preservation modeling (SPM) to capture the sequence similarities of TCRs. Through a multi-task pretraining procedure on MLM and SPM, TCR2vec learns a contextual understanding of TCRs within a similarity-preserved vector space. We first verify the effectiveness of TCR2vec in predicting TCRs binding specificity and TCR clustering through comparison with three other embedding approaches. TCR2vec can be finetuned on small task-specific labeled data for enhanced performance, which outperforms state-of-the-art models by 2-25% in predicting TCRs binding specificity. Next, we compare the performance of two versions of TCR2vec pretrained on full TCRs (TCR2vec) or CDR3s (CDR3vec) and demonstrate that TCR2vec consistently outperforms CDR3vec by 12-14%. Further analysis of attention maps reveals that residues outside CDR3 also make notable contributions to the recognition of antigens. TCR2vec is available at https://github.com/jiangdada1221/TCR2vec.

bioinformatics↗