Biology subjects

Leung, K.-S.

Publications and source records attributed to Leung, K.-S..

8 recordsLinked to original sources

bioRxiv2026

In-silico cell sorting revealed granulocyte-specific single-cell-type gene expression from peripheral blood bulk expression data and its application as host response biomarkers to discriminate bacterial and viral infections

Peripheral Blood transcriptome analysis evaluated the bulk transcript abundance (TA) covering all leukocyte cell populations. However, there are 2 main problems in using bulk expression as biomarkers: (1) A long list of differential expression genes (DEGs) was found, and (2) DEGs cannot be attributed to a host response of any specific cell-type. TA assays after conventional cell sorting, as the gold-standard method, is too tedious for routine use. Recently, we showed that by using a ratio-based biomarker, RBB (ratio of two stringently selected genes), it is feasible to interrogate the gene expression of a single cell-type (monocyte and B lymphocyte) in peripheral whole blood (WB) directly. Here, we apply this in-silico cell sorting algorithm (DIRECT LS-TA, Direct Leukocyte Single cell-type Transcript Abundance) to granulocytes in WB samples to reveal RBBs specific to granulocytes. This DIRECT LS-TA approach without the need for cell-sorting was applied to public datasets to differentiate the 2 types of infection (bacterial vs viral infection). The following RBBs measured in WB correlate with the expression of target (numerator) genes in purified granulocytes, thus cell-sorting can be avoided by using these RBBs: ARG1/SRGN, ANXA3/SRGN, RSAD2/SRGN. Together with monocyte DIRECT LS-TA biomarkers, IFI27/PSAP, direct quantification of 4 genes provided optimal differentiation of viral from bacterial infection. Meta-analysis and unsupervised machine learning classification confirmed the superior performance of DIRECT LS-TA biomarkers. These RBBs found by prior In-silico cell-sorting identified pairs of genes that are used to formulate as ratio-based biomarkers (RBBs) to represent gene expression of granulocytes inside whole blood cell-mixture samples which was useful to triage febrile patients into two major categories of febrile diseases between viral and bacterial infection with high degree of sensitivity and specificity.

immunology↗

bioRxiv2025

Monocyte single cell-type gene expression measured in peripheral blood by DIRECT LS-TA method: the ratio-based biomarkers of (IFI27/PSAP) showed superior performance than interferon score in triage patients with viral infection

A rapid method to triage febrile patients into different categories of etiologies remains a significant challenge even nowadays, when many molecular tests for pathogens are available. Routine serum protein tests like C-reactive protein and procalcitonin have limited specificity. Host response gene signatures are promising biomarkers but they usually require assaying many genes, e.g. 7 genes are commonly used to calculate the interferon (IFN) score. However, these gene panels fail to capture cell-type-specific host responses. Measuring gene expression of a specified single cell population, like monocytes, offers enhanced biological insight. However, it currently requires laborious cell sorting or costly single-cell sequencing techniques, limiting its clinical applicability. This study aims to develop a simple ratio-based biomarker (RBB) representing monocyte-specific host response to viral infection called DIRECT LS-TA method. A simple ratio of 2 genes (both are shortlist monocyte informative genes) quantified in peripheral blood (PB) samples correlated with gene expression in purified monocytes in the corresponding individual. These RBBs cover 3 interferon-stimulated genes (ISGs): IFI27/PSAP, IFI44L/PSAP and SIGLEC1/PSAP. They are compared to the conventional multi-gene IFN score in the differentiation of viral infection. Public gene expression datasets from NCBI GEO were used to shortlist monocyte-informative genes that can be used as the RBB in PB. The DIRECT LS-TA RBB was calculated as the ratio of the target ISG transcript abundance (TA) to that of another reference gene (PSAP or CTSS) directly quantified from bulk PB data (e.g., Log(IFI27/PSAP) in WB). The correlation (expressed by coefficient of determination, R{superscript 2}) between these DIRECT LS-TA RBBs and the gold-standard target gene TA measured in purified monocytes was assessed. The diagnostic performance of selected RBBs (IFI27/PSAP, IFI44L/PSAP, SIGLEC1/PSAP) was compared against the conventional 8-gene IFN score for differentiating viral infections from controls. Direct LS-TA RBBs measured in PB showed strong correlation with gold-standard gene expression measured in purified monocytes (R2 ranged from 0.53 for the target gene IFI27 to >0.9 for the target gene IFI44L). This high level of correlation supports that this simple RBB (DIRECT LS-TA) method can replace the tedious cell sorting approach to obtain single-cell-type gene expression data. All DIRECT LS-TA results of ISGs were raised during viral infection. The best clinical performance in triaging viral infection patients was achieved by IFI27/PSAP or IFI27/CTSS across all datasets. For example, in the GSE111368 dataset, IFI27/PSAP achieved an AUC of 0.94 (95% CI 0.90-0.97) with 88% sensitivity and 95% specificity, surpassing the IFN scores AUC of 0.90 (95% CI 0.85-0.94) with 79% sensitivity and 93% specificity. ConclusionThe DIRECT LS-TA method, utilizing the format of simple two-gene ratio-based biomarkers like IFI27/PSAP, provides a robust and accurate measure of monocyte-specific interferon pathway activation directly from peripheral blood. The superior performance of the DIRECT LS-TA method makes it a promising, readily implementable tool for clinical triage. Its ability to provide single-cell-type specific information, rapid turnaround using standard qPCR/dPCR technology, and enhanced biological specificity make it a valuable molecular host response assessment.

immunology↗

bioRxiv2025

Compositional data modeling of high-dimensional single cell RNA-seq (CoDA-hd): its advantages over commonly used normalization approaches

Compositional data analysis (CoDA) is an emerging statistical framework and has been extended to microbiome, bulk RNA-seq, and cell type proportions in single-cell RNA-seq (scRNA-seq) (with 50-200 components). Here, we explore the high-dimensional application of CoDA (CoDA-hd) and its various log-ratio (LR) transformations to raw count matrices of scRNA-seq which has over 20,000 components (e.g., protein coding genes). scRNA-seq matrices are typically sparse and high-dimensional. Common approaches of normalization such as log-normalization may lead to suspicious findings as previously shown for trajectory inference. Although RNA-seq is compositional data by nature, the geometry of CoDA in high-dimensional simplex is not compatible with most downstream analyses of scRNA-seq which are based on Euclidean space. In this study, we explored if CoDA is adaptable with scRNA-seq in various downstream applications. Specifically, we attempted to study: (1) CoDA adaptability to scRNA-seq; (2) handling of zero data: prior-log-normalized, imputation or with specific count addition; (3) transformation to Euclidean space and compatibility with downstream analyses. Our results suggested that (1) the innovative count addition schemes (e.g., SGM) enable the application of CoDA to high dimensional sparse data (i.e., scRNA-seq); (2) log-normalized data could be transformed to CoDA LR transformation as an approximation; (3) CoDA LR transformations such as count-added centered-log-ratio (CLR) had some advantages in dimension reduction visualization, clustering, and trajectory inference in the tested real & simulated datasets. CLR provided more decent and separated clusters in dimension reductions, improved the Slingshot trajectory inference, and eliminated the suspicious trajectory that is probably caused by the dropouts. We therefore concluded that CoDA may be a preferred scale-free model to handle scRNA-seq data for these downstream applications. Additionally, an R package CoDAhd was developed for conducting CoDA LR transformations for high dimensional scRNA-seq data. The code for implementing CoDA-hd and some example datasets were placed at https://github.com/GO3295/CoDAhd.

bioinformatics↗

bioRxiv2025

Direct single cell-type gene expression analysis in Whole blood: Novel ratio-based gene expression biomarkers using 2 novel monocyte reference genes (PSAP and CTSS) for monitoring bacterial infection

BackgroundTo determine single cell-type specific gene expression in peripheral blood (PB) requires either prior labour-intense cell sorting or expensive single-cell RNA sequencing. We developed and validated a novel ratio-based biomarker (RBB) called Direct Leukocyte Subpopulation-Transcript Abundance (DIRECT LS-TA) assay that allows quantification of monocyte-specific gene expression directly from peripheral blood samples without cell sorting. MethodsThe DIRECT LS-TA method leverages known cell-type proportions and differential gene expression profiles among leukocyte subpopulations (e.g. monocytes, lymphocytes and granulocytes) to identify monocyte-informative genes. We shortlisted genes that had 2.5-fold higher expression in isolated monocytes compared to PB samples, indicating >50% transcript of genes in PB are contributed by monocytes alone. Public gene expression datasets were used to generate a list of monocyte informative genes with which DIRECT LS-TA assay is applicable in PB samples. PSAP and CTSS were identified as the monocyte informative reference genes, based on low biological variation (CV <12%) and high monocyte specificity. They were used as the denominator and together with another monocyte informative target gene as the numerator, the value of this new DIRECT LS-TA can be determined which is a kind of ratio-based biomarker (RBB). The clinical utility was the differentiation of patients with bacterial infection from control subjects in a discovery dataset and 4 other replication datasets. Methods to convert DIRECT LS-TA results to multiple of control median (MoM) provided approximations to delta-delta CT values of relative quantification commonly used in qPCR or digital PCR assays. ResultsOver 50 monocyte-informative genes were identified, including key immune response genes like IFI44L, IL1B, VNN1 and NFKBIZ. DIRECT LS-TA results showed excellent correlation with the gold standard results, gene expression in isolated monocyte expression (R2=0.55-0.97). Then, DIRECT LS-TA of these 50 target genes were evaluated to identify the best RBB to detect the host response to bacterial infection. Monocyte DIRECT LS-TA VNN1 RBB showed consistent upregulation across five independent datasets (median fold change 2.7-fold, p<10-8) with strong diagnostic performance (AUC=0.84-0.99). The expected corresponding delta-delta CT value is more than 1.4 cycle which can be reliably measured by qPCR. Additional monocyte-informative genes including NLRC4, CYP1B1, PFKFB3, LILRA5, NFKBIA, and NFKBIZ also demonstrated significant diagnostic ability (AUC>0.8). ConclusionsThe DIRECT LS-TA method (as a simple RBB) provides reliable quantification of monocyte-specific gene expression directly from whole blood samples without cell separation. The robust performance in bacterial infection diagnosis demonstrates its potential clinical utility for rapid infection differentiation and informed antibiotic stewardship. DIRECT LS-TA will emerge as a new kind of in vitro diagnostics (IVD) which can convey single cell-type gene expression information from PB samples. The new kind of IVD and uniqueness of the information, together with the ease of implementation will make it very useful in clinics

immunology↗

bioRxiv2024

scCaT: an explainable capsulating architecture for sepsis diagnosis transferring from single-cell RNA sequencing

Sepsis is a life-threatening condition characterized by an exaggerated immune response to pathogens, leading to organ damage and high mortality rates in the intensive care unit. Although deep learning has achieved impressive performance on prediction and classification tasks in medicine, it requires large amounts of data and lacks explainability, which hinder its application to sepsis diagnosis. We introduce a deep learning framework, called scCaT, which blends the capsulating architecture with Transformer to develop a sepsis diagnostic model using single-cell RNA sequencing data and transfers it to bulk RNA data. The capsulating architecture effectively groups genes into capsules based on biological functions, which provides explainability in encoding gene expressions. The Transformer serves as a decoder to classify sepsis patients and controls. Our model achieves high accuracy with an AUROC of 0.93 on the single-cell test set and an average AUROC of 0.98 on seven bulk RNA cohorts. Additionally, the capsules can recognize different cell types and distinguish sepsis from control samples based on their biological pathways. This study presents a novel approach for learning gene modules and transferring the model to other data types, offering potential benefits in diagnosing rare diseases with limited subjects. Author summaryDeep learning models used in disease diagnosis usually suffer from insufficient data for training and the lack of explainability, especially in rare diseases. These shortages hinder their application to sepsis diagnosis. Here we propose a diagnostic framework name scCaT(https://github.com/Kimxbzheng/CaT), which transfers knowledge learned from single-cell RNA-seq, for diseases with insufficient bulk data. The framework uses capsulating architecture to group genes into capsules and provide explainability to the deep learning model for sepsis diagnosis. ScCaT achieves robust and outstanding performance for sepsis diagnosis in both scRNA-seq and bulk RNA datasets. This architecture offers potential approaches in diagnosing rare diseases with limited subjects with explainability.

bioinformatics↗

bioRxiv2022

Noncoding RNA's competing endogenous gene pair as motif in serous ovarian cancer

The non-coding RNA (ncRNA) regulation apprears to be associated to the diagnosis and targeted therapy of complex diseases. Motifs of non-coding RNAs and genes in the competing endogenous RNA (ceRNA) network would probably contribute to the accurate prediction of serous ovarian carcinoma (SOC). Hence, we profiled a whole transcriptome of eight human SOCs and eight controls and constructed a ceRNA network including mRNAs, long ncRNAs, and circular RNAs (circRNAs). We identified the mRNA-ncRNA-mRNA motifs in the ceRNA network named the non-coding RNAs competing endogenous gene pairs (ceGPs), through the denoised individualized pair analysis of gene expression (deiPAGE) proposed in this study. 18 cricRNAs ceGPs (cceGPs) were identified from multiple cohorts and were fused as an indicator (SOC index) for SOC discrimination, which carried a high predictive capacity in independent cohorts. It was found that the index was negatively correlated with the CD8+/CD4+ ratio in tumour-infiltration, reflecting the migration and growth of tumour cells in ovarian cancer progression.

bioinformatics↗

bioRxiv2020

Direct measurement of B lymphocyte gene expression biomarkers in peripheral blood enables early prediction of seroconversion after vaccination

Vaccination is a common and efficient means to reduce the mortality and morbidity of emerging infectious diseases. Among responders, injected antigen induces acquired immunity pathways and leads to the final production of antigen-specific antibodies. The whole process may take weeks to months, depending on the antigen. Typically, seroconversion to influenza vaccine is expected after one month with a responder rate of ~50%. An early biomarker to predict response is desirable. Peripheral blood gene expression (or transcript abundance, TA) datasets in the public domain were analyzed for early biomarkers among responders. As peripheral blood samples (such as peripheral blood mononuclear cells, PBMC) are cell mixture samples containing various blood cell-types (leukocyte subpopulations, LS). We first develop a model that enables the determination of TA in B lymphocytes of certain genes directly in PBMC samples without the need of prior cell isolation. These genes are called B cell informative genes. Then a ratio of two B cell informative genes (a target gene and a stably expressed reference gene) measured in PBMC was used as a new biomarker to gauge the target gene expression in B lymphocytes. This method having an obvious advantage over conventional methods by eliminating the tedious procedure of cell sorting and enables directly determining TA of a leukocyte subpopulation in cell mixture samples is called Direct LS-TA method. By using a B lymphocyte-specific gene such as TNFRSF17 or TXNDC5 as target genes with either TNFRSF13C or FCRLA as reference genes, the B cell biomarkers were determined directly in PBMC which was highly correlated with TA of target genes in purified B lymphocytes. These Direct LS-TA biomarkers in PBMC increased significantly early after vaccination in both the discovery dataset and a meta-analysis of 7 datasets. Responders had almost a 2-fold higher Direct LS-TA biomarker level of TNFRSF17 (SMD=0.84, 95% CI=0.47-1.21 after log2). And Direct LS-TA biomarkers of TNFRSF17 and TXNDC5 measured at day 7 predict responder with sensitivity values of higher than 0.7. The Area-under curves (AUC) in receiver operation curve (ROC) analysis were over 0.8. Here, we report a straightforward approach to directly analyses B lymphocyte gene expression in PBMC, which could be used in a routine clinical setting as it avoids the labor-intensive procedures of B lymphocyte isolation. And the method allows the practice of precision medicine in the prediction of vaccination response. Furthermore, response to vaccination could be predicted as early as on day 7. As vaccination response is based on the similar acquired immunology pathway in the upcoming worldwide vaccination campaign against COVID-19, these biomarkers could also be useful to predict seroconversion for individuals.

immunology↗

bioRxiv2018

Identification and characterization of moonlighting long non-coding RNAs based on RNA and protein interactome

Moonlighting proteins are a class of proteins having multiple distinct functions, which play essential roles in a variety of cellular and enzymatic functioning systems. Although there have long been calls for computational algorithms for the identification of moonlighting proteins, research on approaches to identify moonlighting long non-coding RNAs (lncRNAs) has never been undertaken. Here, we introduce a methodology, MoonFinder, for the identification of moonlighting lncRNAs. MoonFinder is a statistical algorithm identifying moonlighting lncRNAs without a priori knowledge through the integration of protein interactome, RNA-protein interactions, and functional annotation of proteins. We identify 155 moonlighting lncRNA candidates and uncover that they are a distinct class of lncRNAs characterized by specific sequence and cellular localization features. The non-coding genes that transcript moonlighting lncRNAs tend to have shorter but more exons and the moonlighting lncRNAs have a localization tendency of residing in the cytoplasmic compartment in comparison with the nuclear compartment. Moreover, moonlighting lncRNAs and moonlighting proteins are rather mutually exclusive in terms of both their direct interactions and interacting partners. Our results also shed light on how the moonlighting candidates and their interacting proteins implicated in the formation and development of cancers and other diseases.

systems biology↗