Biology subjects

Su, Y.

Publications and source records attributed to Su, Y..

At least 181 records · Page 10Linked to original sources

bioRxiv2020

Multiomic Immunophenotyping of COVID-19 Patients Reveals Early Infection Trajectories

Host immune responses play central roles in controlling SARS-CoV2 infection, yet remain incompletely characterized and understood. Here, we present a comprehensive immune response map spanning 454 proteins and 847 metabolites in plasma integrated with single-cell multi-omic assays of PBMCs in which whole transcriptome, 192 surface proteins, and T and B cell receptor sequence were co-analyzed within the context of clinical measures from 50 COVID19 patient samples. Our study reveals novel cellular subpopulations, such as proliferative exhausted CD8+ and CD4+ T cells, and cytotoxic CD4+ T cells, that may be features of severe COVID-19 infection. We condensed over 1 million immune features into a single immune response axis that independently aligns with many clinical features and is also strongly associated with disease severity. Our study represents an important resource towards understanding the heterogeneous immune responses of COVID-19 patients and may provide key information for informing therapeutic development.

immunology↗

bioRxiv2020

Insulin/IGF signaling and TOR promote vitellogenesis via inducing juvenile hormone biosynthesis

Vitellogenesis, including vitellogenin (Vg) production in the fat body and Vg uptake by maturing oocytes, is of great importance for the successful reproduction of adult females. The endocrinal and nutritional regulation of vitellogenesis differs distinctly in insects. Here, the complex crosstalk between juvenile hormone (JH) and the two nutrient sensors, insulin/IGF signaling (IIS) and target of rapamycin (TOR), was investigated to elucidate the molecular mechanisms regulating vitellogenesis in the American cockroach, Periplaneta americana. Our data showed that a block of JH biosynthesis or JH action arrested vitellogenesis, partially by inhibiting the expression of doublesex (Dsx), a key transcription factor gene involved in the sex determination cascade. Depletion of IIS and TOR blocked both JH biosynthesis and vitellogenesis. Importantly, the JH analog methoprene, but not bovine insulin (to restore IIS) and amino acids (to restore TOR activity), restored vitellogenesis in the neck-ligated (both nutrient- and JH-deficient) cockroaches. Combining classic physiology with modern molecular techniques, we have demonstrated that JH signaling alone is able to induce vitellogenesis and thus ovarian maturation. By contrast, IIS and TOR do not induce vitellogenesis independent of JH, the nutrient sensors promote vitellogenesis in an indirect manner via activating JH biosynthesis.

developmental biology↗

bioRxiv2020

Lectin-like Intestinal Defensin Inhibits 2019-nCoV Spike binding to ACE2

The burgeoning epidemic caused by novel coronavirus 2019 (2019-nCoV) is currently a global concern. Angiotensin-converting enzyme-2 (ACE2) is a receptor of 2019-nCoV spike 1 protein (S1) and mediates viral entry into host cells. Despite the abundance of ACE2 in small intestine, few digestive symptoms are observed in patients infected by 2019-nCoV. Herein, we investigated the interactions between ACE2 and human defensins (HDs) specifically secreted by intestinal Paneth cells. The lectin-like HD5, rather than HD6, bound ACE2 with a high affinity of 39.3 nM and weakened the subsequent recruitment of 2019-nCoV S1. The cloak of HD5 on the ligand-binding domain of ACE2 was confirmed by molecular dynamic simulation. A remarkable dose-dependent preventive effect of HD5 on 2019-nCoV S1 binding to intestinal epithelial cells was further evidenced by in vitro experiments. Our findings unmasked the innate defense function of lectin-like intestinal defensin against 2019-nCoV, which may provide new insights into the prevention and treatment of 2019-nCoV infection.

microbiology↗

bioRxiv2020

Discovery of a 382-nt deletion during the early evolution of SARS-CoV-2

To date, the SARS-CoV-2 genome has been considered genetically more stable than SARS-CoV or MERS-CoV. Here we report a 382-nt deletion covering almost the entire open reading frame 8 (ORF8) of SARS-CoV-2 obtained from eight hospitalized patients in Singapore. The deletion also removes the ORF8 transcription-regulatory sequence (TRS), which in turn enhances the downstream transcription of the N gene. We also found that viruses with the deletion have been circulating for at least four weeks. During the SARS-CoV outbreak in 2003, a number of genetic variants were observed in the human population [1], and similar variation has since been observed across SARS-related CoVs in humans and bats. Overwhelmingly these viruses had mutations or deletions in ORF8, that have been associated with reduced replicative fitness of the virus [2]. This is also consistent with the observation that towards the end of the outbreak sequences obtained from human SARS cases possessed an ORF8 deletion that may be associated with host adaptation [1]. We therefore hypothesise that the major deletion revealed in this study may lead to an attenuated phenotype of SARS-CoV-2.

microbiology↗

bioRxiv2020

Metabolites alleviate staphylococcal bloodstream infection in a NO-dependent manner via arginase inhibition

Staphylococcus aureus is a notorious bacterial pathogen that often causes soft tissue and bloodstream infections and invariably garners resistance mechanisms against new antibiotics. Host innate immunity modulated by metabolites has been proved as a powerful strategy against bacterial infections. However, few studies focus on the application of this strategy against S. aureus infection. Here, we identified four metabolite biomarkers, L-proline, L-isoleucine, L-leucine, and L-valine (PILV), by a metabolomics study. In animal models of S. aureus bloodstream infection, exogenous administration of each metabolite or PILV shows an anti-infective effect, while PILV treatment has higher protection than individual metabolite treatment. Each metabolite targets nitric oxide (NO) to kill S. aureus via arginase inhibition, and PILV exhibits additive inhibition of arginase activity that causes further killing. This suppression also contributes to the metabolite-mediated phagocytic killing of S. aureus in human blood. Our finding demonstrates the metabolite-mediated innate immunity as a therapeutic intervention for S. aureus infection.

microbiology↗

bioRxiv2020

Evolutionary context-integrated deep sequence modeling for protein engineering

Protein engineering seeks to design proteins with improved or novel functions. Compared to rational design and directed evolution approaches, machine learning-guided approaches traverse the fitness landscape more effectively and hold the promise for accelerating engineering and reducing the experimental cost and effort. A critical challenge here is whether we are capable of predicting the function or fitness of unseen protein variants. By learning from the sequence and large-scale screening data of characterized variants, machine learning models predict functional fitness of sequences and prioritize new variants that are very likely to demonstrate enhanced functional properties, thereby guiding and accelerating rational design and directed evolution. While existing generative models and language models have been developed to predict the effects of mutation and assist protein engineering, the accuracy of these models is limited due to their unsupervised nature of the general sequence contexts they captured that is not specific to the protein being engineered. In this work, we propose ECNet, a deep-learning algorithm to exploit evolutionary contexts to predict functional fitness for protein engineering. Our method integrated local evolutionary context from homologous sequences that explicitly model residue-residue epistasis for the protein of interest, as well as the global evolutionary context that encodes rich semantic and structural features from the enormous protein sequence universe. This biologically motivated sequence modeling approach enables accurate mapping from sequence to function and provides generalization from low-order mutants to higher-orders. Through extensive benchmark experiments, we showed that our method outperforms existing methods on [~]50 deep mutagenesis scanning and random mutagenesis datasets, demonstrating its potential of guiding and expediting protein engineering.

bioinformatics↗

bioRxiv2020

ALPPL2 is a highly specific and targetable tumor cell surface antigen

It has been challenging to identify tumor-specific cell surface antigens as the vast majority of tumor-associated antigens are also expressed by some normal tissues. In the course of our study on mesothelioma, we identified a highly specific tumor cell surface antigen that can be targeted for therapy development. Mesothelioma is caused by malignant transformation of the mesothelium, incurable and categorized into three histological subtypes, epithelioid, biphasic and sarcomatoid. To identity novel mesothelioma cell surface antigens with broad subtype coverage and high tissue specificity, we have previously selected phage antibody display libraries on live mesothelioma cells and tissues following counter-selection on normal cells, and identified a panel of human antibodies that bind all subtypes of mesothelioma but not normal mesothelium. One of the antibodies, M25, showed high specificity, and we hereby report the identification of the M25 antigen as ALPPL2. We performed immunohistochemistry on normal human tissues and found that ALPPL2 is expressed only on placental trophoblasts but not any other normal tissues. This exquisite tissue specificity and broad tumor type coverage suggests that ALPPL2 could be an excellent cell surface target for therapeutic development against mesothelioma. To evaluate therapeutic potential of ALPPL2 targeting, we developed an ALPPL2-targeted antibody-drug conjugate and demonstrated potent and specific tumor killing in vitro and in vivo against both epithelioid and sarcomatoid mesothelioma. Thus ALPPL2 belongs to a rare class of cell surface antigens that can be said as being truly tumor specific and is well suited for therapy development against ALPPL2 expressing tumors.

cancer biology↗

bioRxiv2019

LegoBody: facile generation of bispecific and multi-specific antibodies

Bispecific and multi-specific antibodies are capable of recognizing multiple ligands simultaneously or synergistically, creating complex biological interactions not achievable by monoclonal antibodies, thus expanding opportunities for novel therapy development. With the large number of monoclonal antibodies either approved or under clinical development, there are numerous opportunities to combine their specificities to further improve therapeutic potential. Although simple in concept, clinical development of bi- and multi-specific antibodies face several challenges, chief of which is how to efficiently and reliably produce bispecific and multi-specific antibodies with expected specificity and desired biophysical properties. In this study, we developed a modular approach that uses temporary linkers to enforce proper chain pairing and proteases such as thrombin to remove those linkers from the final product. Combined with the knob-into-hole design, we can generate IgG-like, bi- or multi-specific antibodies from any pre-existing monoclonal antibodies. The approach is highly versatile and applicable to any monoclonal antibody pair or panel, expediting evaluation and therapeutic development of bi- and multi-specific antibodies.

bioengineering↗

bioRxiv2019

STAT3 localizes in mitochondria-associated ER membranes instead of in mitochondria

Signal Transducer and Activator of Transcription 3 (STAT3) is a transcription factor (TF) that regulates a variety of biological processes, including a key role in mediating mitochondrial metabolism. It has been shown that STAT3 performs this function by translocating in minute amounts into mitochondria and interacting with mitochondrial proteins and genome. However, whether STAT3 localizes in mitochondria is still up for debate. To decipher the role of mitochondrial STAT3 requires a detailed understanding of its cellular localization. Using Percoll density gradient centrifugation, we surprisingly found that STAT3 is not located in the mitochondrial fraction, but instead, in the mitochondria-associated endoplasmic reticulum membrane (MAM) fraction. This was confirmed by sub-diffraction image analysis of labeled mitochondria in embryonic astrocytes. Also, we find that other TFs that have been previously found to localize in mitochondria are also found instead in the MAM fraction. Our results suggest that STAT3 and other transcriptional factors are, contrary to prior studies, consolidated specifically at MAMs, and further efforts to understand mitochondrial STAT3 function must take into consideration this localization, as the associated functional consequences offer a different interpretation to the questions of STAT3 trafficking and signaling in the mitochondria.

cell biology↗

bioRxiv2019

Pitch of Harmonic Complex Tones: Rate-Place Coding of Resolved Components in Harmonic and Inharmonic Complex Tones in Auditory Midbrain

Harmonic complex tones (HCT) commonly occurring in speech and music evoke a strong pitch at their fundamental frequency (F0), especially when they contain harmonics individually resolved by the cochlea. When all frequency components of an HCT are shifted by the same amount, the pitch of the resulting inharmonic tone (IHCT) also shifts although the envelope repetition rate is unchanged. A rate-place code whereby resolved harmonics are represented by local maxima in firing rates along the tonotopic axis has been characterized in the auditory nerve and primary auditory cortex, but little is known about intermediate processing stages. We recorded single neuron responses to HCT and IHCT with varying F0 and sound level in the inferior colliculus (IC) of unanesthetized rabbits. Many neurons showed peaks in firing rates when a low-numbered harmonic aligned with the neurons characteristic frequency, demonstrating \"rate-place\" coding. The IC rate-place code was most prevalent for F0>800 Hz, was only moderately dependent on sound level over a 40 dB range, and was not sensitive to stimulus harmonicity. A spectral receptive-field model incorporating broadband inhibition better predicted the neural responses than a purely excitatory model, suggesting an enhancement of the rate-place representation by inhibition. Some IC neurons showed facilitation in response to HCT, similar to cortical \"harmonic template neurons\" (Feng and Wang 2017), but to a lesser degree. Our findings shed light on the transformation of rate-place coding of resolved harmonics along the auditory pathway, and suggest a gradual emergence of harmonic templates from low to high processing centers.\n\nSignificance statementHarmonic complex tones are ubiquitous in speech and music and produce strong pitch percepts in human listeners when they contain frequency components that are individually resolved by the cochlea. Here, we characterize a \"rate-place\" code for resolved harmonics in the auditory midbrain that is more robust across sound levels than the peripheral rate-place code and insensitive to the harmonic relationships among frequency components. We use a computational model to show that inhibition may play an important role in shaping the rate-place code. We also show that midbrain auditory neurons can demonstrate similar properties as cortical harmonic template neurons. Our study fills a gap in understanding the transformation in neural representations of resolved harmonics along the auditory pathway.

neuroscience↗

bioRxiv2019

Vaping nicotine-containing electronic cigarettes produces addiction-like behaviors and cardiopulmonary abnormalities in rats

The debate about electronic cigarettes has divided healthcare professionals, policymakers, and communities. Central points of disagreement are whether vaping electronic cigarettes are addictive and whether they produce major pulmonary complications. We developed a novel model of nicotine vapor self-administration in rats and found that rats voluntarily exposed themselves to nicotine vapor to the point of reaching blood nicotine levels that are similar to humans, exhibiting both addiction-like behaviors and cardiopulmonary abnormalities. The smoking cessation drug varenicline decreased electronic cigarette self-administration. These findings confirm the addictive properties and harmful effects of nicotine vapor and identify a potential medication for the treatment of electronic cigarette addiction.\n\nOne Sentence SummaryVaping nicotine-containing electronic cigarettes produces cardiopulmonary abnormalities, nicotine dependence and addiction-like behaviors, which are reduced by the smoking cessation drug varenicline.

neuroscience↗

bioRxiv2019

Trajectories from Snapshots: Integrated proteomic and metabolic single-cell assays reveal multiple independent adaptive responses to drug tolerance in a BRAF-mutant melanoma cell line

The determination of individual cell trajectories through a high-dimensional cell-state space is an outstanding challenge, with relevance towards understanding biological changes ranging from cellular differentiation to epigenetic (adaptive) responses of diseased cells to drugging. We report on a combined experimental and theoretic method for determining the trajectories that specific highly plastic BRAFV600E mutant patient-derived melanoma cancer cells take between drug-naive and drug-tolerant states. Recent studies have implicated non-genetic, fast-acting resistance mechanisms are activated in these cells following BRAF inhibition. While single-cell highly multiplex omics tools can yield snapshots of the cell state space landscape sampled at any given time point, individual cell trajectories must be inferred from a kinetic series of snapshots, and that inference can be confounded by stochastic cell state switching. Using a microfludic-based single-cell integrated proteomic and metabolic assay, we assayed for a panel of signaling, phenotypic, and metabolic regulators at four time points during the first five days of drug treatment. Dimensional reduction of the resultant data set, coupled with information theoretic analysis, uncovered a complex cell state landscape and identified two distinct paths connecting drug-naive and drug-tolerant states. Cells are shown to exclusively traverse one of the two pathways depending on the level of the lineage restricted transcription factor MITF in the drug-naive cells. The two trajectories are associated with distinct signaling and metabolic susceptibilities, and are independently druggable. Our results update the paradigm of adaptive resistance development in an isogenic cell population and offer insight into the design of more effective combination therapies.

bioengineering↗

bioRxiv2019

Kinetic Inference Resolves Epigenetic Mechanism of Drug Resistance in Melanoma

Drug-induced dedifferentiation towards a drug-tolerant persister state is a common mechanism cancer cells exploit to escape therapies, posing a significant obstacle to sustained therapeutic efficacy. The dynamic coordination of epigenomic and transcriptomic programs at the early-stage of drug exposure, which initiates and orchestrates these reversible dedifferentiation events, remains largely unexplored. Here we employ high-temporal-resolution multi-omics profiling, information-theoretic approaches, and dynamic system modeling to probe these processes in BRAF-mutant melanoma models and patient specimens. We uncover a hysteretic transition trajectory of melanoma cells in response to oncogene inhibition and subsequent release, driven by the sequential operation of two tightly coupled transcriptional waves, which orchestrate genome-scale chromatin state reconfiguration. Modeling of the transcriptional wave interactions predicts NF-{kappa}B/RelA-driven chromatin remodeling as the underlying mechanism of cell-state dedifferentiation, a finding we validate experimentally. Our results identify critical RelA-target genes that are epigenetically modulated to drive this process, establishing a quantitative epigenome gauge to measure cell-state plasticity in melanomas, which supports the potential use of drugs targeting epigenetic machineries to potentiate oncogene inhibition. Extending our investigation to other cancer models, we identify oxidative stress-mediated NF-{kappa}B/RelA activation as a common mechanism driving cellular transitions towards drug-tolerant persister states, revealing a novel and pivotal role for the NF-{kappa}B signaling axis in linking cellular oxidative stress to cancer progression.

cancer biology↗

bioRxiv2019

Characterizing the Emergence of Liver and Gallbladder from the Embryonic Endoderm through Single-Cell RNA-Seq

The liver and gallbladder are among the most important internal organs derived from the endoderm. Several inductive signals regulate liver development, yet the pure nascent hepatic and gallbladder cells are unable to be isolated due to limited cell markers and cell numbers. The transcriptome networks of the hepatic lineage in the endoderm, and how the gallbladder differentiates from the adjacent endoderm population, are not fully understood. Using a transgenic Foxa2eGFP reporter mouse line, we performed deep single-cell RNA sequencing on 1,966 individual cells, including nascent hepatic and gallbladder cells, isolated from the endoderm and hepatic regions from ten embryonic stages, ranging from day E7.5 to E15.5. We identified the embryonic liver developmental trajectory from primitive streak to hepatoblasts and characterized the transcriptome of the hepatic lineage. During pre-hepatogenesis (5-6 somite stage), we have identified two groups of foregut endoderm cells, one derived from visceral endoderm and the second derived from primitive streak via a mesenchymal-epithelial transition (MET). During the liver specification stages, liver primordium was identified to share both foregut and liver features. We also documented dynamic gene expression during the epithelial-hepatic transition (EHT). Six gene groups were found to switch on or off at different stages during liver specification. Importantly, we found that RXR complex signaling and newly identified transcription factors associated with liver specification. Moreover, we revealed the gallbladder primordium cells at E9.5 and found genes that transcriptionally distinguish them from the liver primordium. The present data provides a high-resolution resource and critical insights for understanding the emergence of the endoderm, liver and gallbladder development.

developmental biology↗

bioRxiv2019

The relative transmissibility of shigellosis among male and female individuals in Hubei Province, China: a modelling study

ObjectiveShigellosis has been a heavy burden in China. However, its relative transmissibility in male and female individuals remains unclear.\n\nMethodA sex-based Susceptible-Exposed-Infectious/Asymptomatic-Recovered (SEIAR) model was applied to explore the dataset of reported shigellosis cases built by Hubei Province from 2005 to 2017. Two indicators, secondary attack rate (SAR) and relative ratio of SAR between males and females, were developed to assess the relative transmissibility in males and females.\n\nResultsThe number of cases and reported incidences in males and females demonstrated a significant decreasing trend (Male trend:{chi} 2 = 11.268, P = 0.001, Female trend:{chi} 2 = 11.144, P = 0.001). SEIAR model had a great fitting effect with the data of shigellosis (P < 0.001). Our simulation revealed that, when parameter {beta}fm = 0, the greatest decrease in cases were obtained for different genders. The median value for SARmm was 2.3225 x 10-8 (Range: 1.7574 x 10-8 - 3.8565 x 10-8), SARmf was 2.5729 x 10-8 (Range: 1.3772 x 10-8 - 3.2773 x 10-8), SARfm was 2.7630 x 10-8 (Range: 1.8387 x 10-8 - 4.2638 x 10-8) and SARff was 2.1061 x 10-8 (Range: 1.0201 x 10-8 - 3.2140 x 10-8). The median value of relative ratio calculated by SAR in mm versus (vs) mf was 0.93 (Range: 0.75 - 1.47), mm vs fm was 0.90 (Range: 0.41 - 1.81), mm vs ff was 1.07 (Range: 0.55 - 2.93), mf vs fm was 0.99 (Range: 0.32 - 1.25), mf vs ff was 1.17 (Range: 0.43 - 3.21) and ff vs fm was 0.75 (Range: 0.35 - 1.06).\n\nConclusionTransmissibility of shigellosis is different among male and female individuals. Shigellosis seems to be more transmissible in males than in females.\n\nAuthor summaryShigellosis, also known as bacillary dysentery, is an infectious disease caused by the genus Shigella spp. Developing countries have high disease burden of shigellosis. However, its relative transmissibility in male and female individuals remains unclear. In this study, we employed a mathematical model to explore the dataset of reported shigellosis cases built by Hubei Province, China from 2005 to 2017. Two indicators, secondary attack rate (SAR) and relative ratio of SAR between males and females, were developed to assess the relative transmissibility in males and females. We found that shigellosis has medium transmissibility among male and female individuals. The disease seems to be more transmissible in males than in females.

epidemiology↗

bioRxiv2019

Accelerating iterative deconvolution and multiview fusion by orders of magnitude

We describe theoretical and practical advances in algorithm and software design, resulting in ten to several thousand-fold faster deconvolution and multiview fusion than previous methods. First, we adapt methods from medical imaging, showing that an unmatched back projector accelerates Richardson-Lucy deconvolution by at least 10-fold, in most cases requiring only a single iteration. Second, we show that improvements in 3D image-based registration with GPU processing result in speedups of 10-100-fold over CPU processing. Third, we show that deep learning can provide further accelerations, particularly for deconvolution with a spatially varying point spread function. We illustrate the power of our methods from the subcellular to millimeter spatial scale, on diverse samples including single cells, nematode and zebrafish embryos, and cleared mouse tissue. Finally, we show that our methods facilitate the use of new microscopes that improve spatial resolution, including dual-view cleared tissue light-sheet microscopy and reflective lattice light-sheet microscopy.

bioengineering↗

bioRxiv2019

QTL and genetic analysis controlling fiber quality traits using paternal backcross population in Upland cotton

In present study, F14 recombinant inbred line (RIL) population was backcrossed to paternal parent for a paternal backcross (BC/P) population, deriving from one Upland cotton hybrid. Three repetitive BC/P field trials and one BC/M field trial were performed including both two BC populations and the original RIL population. Totally, for fiber quality traits, 24 novel QTLs were detected and 13 QTLs validated previous results. And 19 quantitative trait loci (QTL) in BC/P populations explained 5.01% - 22.09% of phenotype variation (PV). Among the 19 QTLs, three QTLs were detected simultaneously in BC/M population. The present study provided novel alleles of male parent for fiber quality traits with positive genetic effects. Particularly, qFS-Chr3-1 controlling fiber strength explained 22.09% of PV in BC/P population, which increased 0.48 cN/tex for fiber strength. A total of seven, two, eight, two and six QTLs explained over 10.00% of PV for fiber length, fiber uniformity, fiber strength, fiber elongation and fiber micronaire, respectively. In the RIL population, six common QTLs detected in more than one environment such as qFL-Chr1-2, qFS-Chr5-1, qFS-Chr9-1, qFS-Chr21-1, qFM-Chr9-1 and qFM-Chr9-2. Two common QTLs of qFE-Chr2-2 (TMB2386-SWU12343) and qFM-Chr9-1 (NAU2873-CGR6771) explained 22.42% and 21.91% of PV. In addition, a total of 142 and 46 epistatic QTLs and QTL x environments (E-QTLs and QQEs) were identified in RIL-P and BC/P populations, respectively.

plant biology↗

bioRxiv2019

Mitigating Data Scarcity in Protein Binding Prediction Using Meta-Learning

A plethora of biological functions are performed through various types of protein-peptide binding. Prime examples include the protein kinase phosphorylation on peptide substrates and the binding of major histocompatibility complex to neoantigens in the immune system. Understanding the specificity of protein-peptide interactions is critical for unraveling the architectures of functional pathways and the mechanisms of cellular processes in human cells. Despite mass-spectrometric techniques were developed for the identification of protein-peptide interactions, our understanding of the preferences of proteins on their binding peptides is still rudimentary. As a complementary direction, a line of computational prediction methods has been recently proposed to predict protein-peptide bindings which efficiently provide rich functional annotations on a large scale. To achieve a high prediction accuracy, these computational methods require a sufficient amount of data to build the prediction model. However, the number of experimentally verified protein-peptide bindings is often limited in real cases. For example, a majority of protein kinases have very few experimentally verified phosphorylation sites (e.g., less than 30 sites) in existing databases. These methods are thus limited to building accurate prediction models for only well-characterized proteins with a large volume of known binding peptides and cannot be extended to predict new binding peptides for less-studied proteins. In this paper, we introduce a generic framework to address this issue of data scarcity in protein binding prediction. We demonstrate the applicability of our framework in predicting kinase-specific phosphorylation sites. Our method uses an effective training strategy to build a prediction model with robust transferability. The model is able to predict the phosphorylation sites of a less-studied kinase, even if there is only a small number of phosphorylation sites known for this kinase. To achieve this, we train the model via a meta-learning phase followed by a few-shot learning phase. We demonstrate our framework has better transferability than state-of-the-art methods and is effective in utilizing limited data to accurately predict phosphorylation sites for less-characterized kinases. The implementation of our framework is available at https://github.com/luoyunan/MetaKinase.

bioinformatics↗