Deciphering gene expression patterns using large-scale transcriptomic data and its applications.

Bulk and single-cell RNA-seq (scRNA-seq) data are being used as alternatives to traditional technology in biology and medicine research. These data are used, for example, for the detection of differentially expressed (DE) genes. Several statistical methods have been developed for the classification of bulk and single-cell RNA-seq data. These feature genes are vitally important for the classification of bulk and single-cell RNA-seq data. The majority of genes are not DE and they are thus irrelevant for class distinction. To improve the classification performance and save the computation time, removal of irrelevant genes is necessary. Removal will aid the detection of the important feature genes. Widely used schemes in the literature, such as the BSS/WSS (BW) method, assume that data are normally distributed and may not be suitable for bulk and single-cell RNA-seq data. In this article, a category encoding (CAEN) method is proposed to select feature genes for bulk and single-cell RNA-seq data classification. This novel method encodes categories by employing the rank of sequence samples for each gene in each class. Correlation coefficients are considered for gene and class with the rank of sample and a new rank of category. The highest gene correlation coefficients are considered feature genes, which are the most effective for classifying bulk and single-cell RNA-seq dataset. The sure screening method was also established for rank consistency properties of the proposed CAEN method. Simulation studies show that the classifier using the proposed CAEN method performs better than, or at least as well as, the existing methods in most settings. Existing real datasets were analyzed, with the results demonstrating superior performance of the proposed method over current competitors. The application has been coded into an R package named "CAEN" to facilitate wide use.

Structure Guided Design of CCL27 CAR T-Cells Against CCR10 for Multiple Myeloma

Single-cell RNA-sequencing is an invaluable research tool that allows for the investigation of gene expression in heterogeneous cancer cell populations in ways that bulk RNA-seq cannot. However, normal (i.e., non tumor) cells in cancer samples have the potential to confound the downstream analysis of single-cell RNA-seq data. Existing methods for identifying cancer and normal cells include copy number variation inference, marker-gene expression analysis, and expression-based clustering. This work aims to extend the existing approaches for identifying cancer cells in single-cell RNA-seq samples by incorporating variant calling and the identification of putative driver alterations. We found that putative driver alterations can be detected in single-cell RNA-seq data obtained with full-length transcript technologies and noticed that a subset of cells in tumor samples are enriched for putative driver alterations as compared to normal cells. Furthermore, we show that the number of putative driver alterations and inferred copy number variation are not correlated in all samples. Taken together, our findings suggest that augmenting existing cancer-cell filtering methods with variant calling and analysis can increase the number of tumor cells that can be confidently included in downstream analyses of single-cell full-length transcript RNA-seq datasets.

BackgroundCancer-associated fibroblasts (CAFs) are involved in tumor growth, angiogenesis, metastasis, and resistance to therapy. We sought to explore the CAFs characteristics in hepatocellular carcinoma (HCC) and establish a CAF-based risk signature for predicting the prognosis of HCC patients.MethodsThe signal-cell RNA sequencing (scRNA-seq) data was obtained from the GEO database. Bulk RNA-seq data and microarray data of HCC were obtained from the TCGA and GEO databases respectively. Seurat R package was applied to process scRNA-seq data and identify CAF clusters according to the CAF markers. Differential expression analysis was performed to screen differentially expressed genes (DEGs) between normal and tumor samples in TCGA dataset. Then Pearson correlation analysis was used to determine the DEGs associated with CAF clusters, followed by the univariate Cox regression analysis to identify CAF-related prognostic genes. Lasso regression was implemented to construct a risk signature based on CAF-related prognostic genes. Finally, a nomogram model based on the risk signature and clinicopathological characteristics was developed.ResultsBased on scRNA-seq data, we identified 4 CAF clusters in HCC, 3 of which were associated with prognosis in HCC. A total of 423 genes were identified from 2811 DEGs to be significantly correlated with CAF clusters, and were narrowed down to generate a risk signature with 6 genes. These six genes were primarily connected with 39 pathways, such as angiogenesis, apoptosis, and hypoxia. Meanwhile, the risk signature was significantly associated with stromal and immune scores, as well as some immune cells. Multivariate analysis revealed that risk signature was an independent prognostic factor for HCC, and its value in predicting immunotherapeutic outcomes was confirmed. A novel nomogram integrating the stage and CAF-based risk signature was constructed, which exhibited favorable predictability and reliability in the prognosis prediction of HCC.ConclusionCAF-based risk signatures can effectively predict the prognosis of HCC, and comprehensive characterization of the CAF signature of HCC may help to interpret the response of HCC to immunotherapy and provide new strategies for cancer treatment.

BackgroundThe ability of cancer-associated fibroblasts (CAFs) to encourage angiogenesis, tumor cell spread, and increase treatment resistance makes them pro-tumorigenic. We aimed to investigate the CAF signature in Bladder urothelial carcinoma (BLCA) and, for clinical application, to build a CAF-based risk signature to decipher the immune landscape and screen for suitable treatment BLCA samples.MethodsCAF-related genes were discovered by superimposing CAF marker genes discovered from single-cell RNA-seq (scRNA-seq) data taken from the GEO database with CAF module genes discovered by weighted gene co-expression network analysis (WGCNA) using bulk RNA-seq data from TCGA. After identifying prognostic genes related with CAF using univariate Cox regression, Lasso regression was used to build a risk signature. With microarray data from the GEO database, prognostic characteristics were externally verified. For high and low CAF-risk categories, immune cells and immunotherapy responses were analyzed. Finally, a nomogram model based on the risk signature and prospective chemotherapeutic drugs were examined.ResultsCombining scRNA-seq and bulk-seq data analysis yielded a total of 124 CAF-related genes. LRP1, ANXA5, SERPINE2, ECM1, RBP1, GJA1, and FKBP10 were the seven BLCA prognostic genes that remained after univariate Cox regression and LASSO regression analyses. Then, based on these genes, prognostic characteristics were created and validated to predict survival in BLCA patients. Additionally, risk signature had a strong correlation with known CAF scores, stromal scores, and certain immune cells. The CAF-risk signature was identified as an independent prognostic factor for BLCA using multifactorial analysis, and its usefulness in predicting immunotherapy response was confirmed. Based on risk classification, we projected six highly sensitive anticancer medicines for the high-risk group.ConclusionThe prognosis of BLCA may be accurately predicted using CAF-based risk signature. With a thorough understanding of the BLCA CAF-signature, it might be able to explain the BLCA patients’ response to immunotherapy and identify a potential target for BLCA treatment.

Kawasaki Disease (KD) affects young children less than five years old with severe blood vessel inflammation. Despite being treatable, the causes and mechanisms remain elusive. This study conducted a meta-analysis of RNA sequencing (RNA-seq) data from human and animal models to explore KD's transcriptomic profile and evaluate animal models. We retrieved bulk and single-cell RNA-seq data from Gene Expression Omnibus, with blood and coronary artery samples from KD patients, aorta samples from KD mouse models (Lactobacillus casei cell wall extract-injected mice), and their controls. Upon consistent quality control, we applied Fisher's exact test to assess differential gene expression, followed by an enrichment analysis of overlapping genes. These studies identified 400 differentially expressed genes in blood samples of KD patients compared to controls and 413 genes in coronary artery samples. The data from KD blood and KD coronary artery samples shared only 16 differentially expressed genes. Eighty-one genes overlapped between KD human coronary arteries and KD mouse aortas, and 67 of these 81 genes were regulated in parallel in both humans and mice: 30 genes were up-regulated, and 37 were down-regulated. These included previously identified KD-upregulated genes: CD74, SFRP4, ITGA4, and IKZF1. Gene enrichment analysis revealed significant alterations in the cardiomyopathy pathway. Single-cell RNAseq showed a few significant markers, with known KD markers like S100A9, S100A8, CD74, CD14, IFITM2, and IFITM3, being overexpressed in KD cohorts. Gene profiles obtained from KD human coronary artery are more compatible with data from aorta samples of KD mice than blood samples of KD humans, validating KD animal models for identifying therapeutic targets. Although blood samples can be utilized to discover novel biomarkers, more comprehensive single-cell sequencing is required to detail gene expression in different blood cell populations. This study identifies critical genes from human and mouse tissues to help develop new treatment strategies for KD.

Disparities in breast cancer (BC) stage of presentation and survival are well documented in patients of different ethnicities. These differences are undoubtedly due to a combination of extrinsic and intrinsic factors. Our group is focused on the genetic contributions to these disparities. Recent observations by our group and others suggest not only the involvement of specific genes contributing to BC risk in African-American (AA) women, but that multiple distinct ethnic-specific gene expression patterns exist in both matched cancerous and normal breast tissue samples. We report the first results of a parallel study which is addressing an unanswered question: are there ethnic-specific differences in the transcriptome of normal breast tissue? We are completing gene expression profiling studies in normal breast FFPE samples from 25 AA and 25 Caucasian (Cauc) women with no history of breast cancer. Laser capture micro-dissection (LCM) is used to further identify and extract epithelial vs stromal cells from each FFPE slide for use in RNA expression profiling. RNA isolation and hybridization occurs in collaboration with Almac Diagnostics (Dx) using their proprietary Breast Cancer DSA array. Gene expression data analyses involves multiple statistical packages developed for DNA microarray studies, and is focusing on comparison of gene expression profiles across four study groups- AA vs Cauc samples and epithelial vs stroma cells. For baseline comparison, a subset of this study group (10 AA and 10 Cauc samples non-LCM dissected) has also been analyzed. Results from these non-LCM dissected samples displayed expression differences between AA and Cauc in 333 known genes. Applying the stringency criteria of a fold change of two or greater and a statistical p-value of 0.01, reduced the significant gene list to 61 genes. Initial pathway analyses indicate these genes are involved in cell adhesion, G-protein signaling, apoptosis and survival pathways, and cytoskeleton remodeling - all critical pathways for cell survival. These initial results are being validated in the LCM-dissected samples, and our latest results will be presented. Results from analyses of these normal samples will be compared to our previous results obtained in matched tumor-normal triple negative AA and Cauc breast cancer samples (as well as publicly available data sets). Combined analysis of tumor and normal breast expression (transcriptome) data will help to better understand the possible significance of gene expression differences in breast tissue between ethnic groups. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4196. doi:1538-7445.AM2012-4196

Transposable elements (TEs) are an integral part of the host transcriptome. TE-containing noncoding RNAs (ncRNAs) show considerable tissue specificity and play important roles during development, including stem cell maintenance and cell differentiation. Recent advances in single-cell RNA-seq (scRNA-seq) revolutionized cell type–specific gene expression analysis. However, effective scRNA-seq quantification tools tailored for TEs are lacking, limiting our ability to dissect TE expression dynamics at single-cell resolution. To address this issue, we established a TE expression quantification pipeline that is compatible with scRNA-seq data generated across multiple technology platforms. We constructed TE-containing ncRNA references using bulk RNA-seq data and showed that quantifying TE expression at the transcript level effectively reduces noise. As proof of principle, we applied this strategy to mouse embryonic stem cells and successfully captured the expression profile of endogenous retroviruses in single cells. We further expanded our analysis to scRNA-seq data from early stages of mouse embryogenesis. Our results illustrated the dynamic TE expression at preimplantation stages and revealed 146 TE-containing ncRNA transcripts with substantial tissue specificity during gastrulation and early organogenesis.

Single-cell RNA-seq data contain a large proportion of zeros for expressed genes. Such dropout events present a fundamental challenge for various types of data analyses. Here, we describe the SCRABBLE algorithm to address this problem. SCRABBLE leverages bulk data as a constraint and reduces unwanted bias towards expressed genes during imputation. Using both simulation and several types of experimental data, we demonstrate that SCRABBLE outperforms the existing methods in recovering dropout events, capturing true distribution of gene expression across cells, and preserving gene-gene relationship and cell-cell relationship in the data.

While recent studies have highlighted oxidative stress (OS) as a pivotal factor influencing tumor dynamics, its specific interactions within the tumor microenvironment (TME) of colorectal cancer (CRC) remain elusive. This study seeks to unveil the impact of OS on the CRC TME and to develop an advanced OS-related risk signature (OSRRS) model. We analyzed OS-related pathway activities using both single-cell and bulk RNA-seq data. An unsupervised clustering algorithm was utilized to identify OS-related subtypes. Based on genes associated with the OS pathways, we constructed an OSRRS model employing the LASSO Cox analysis. For validation, we employed quantitative real-time polymerase chain reaction (qRT-PCR) coupled with immunohistochemical (IHC) analyses on tissue microarrays (TMA) to confirm the expression of the identified gene. Examination of the single-cell RNA-seq GSE132465 dataset revealed a universal elevation in OS-associated pathway activities within malignant cells. By integrating this with the bulk RNA-seq TCGA-CRC dataset, we identified two unique OS-specific clusters. This led to the establishment of a 12-gene OSRRS using the LASSO Cox method. The robustness of our model was further verified using the GSE39582 and GSE17538 cohorts. Notably, increased expression of the UCN gene was observed in CRC specimens, as confirmed by qRT-PCR and IHC assays on TMA. In this research, we delineated two distinct subtypes of CRC associated with OS. The developed OSRRS holds promise as a candidate prognostic and stratification tool for CRC management. Collectively, these results shed light on the intricate role of OS in CRC pathology.

Single-cell RNA-seq overcomes the shortcomings of conventional transcriptome sequencing technology and could provide a powerful tool for distinguishing the transcriptome characteristics of various cell types in biological tissues, and comprehensively revealing the heterogeneity of gene expression between cells. Many Intelligent Computing methods have been presented to infer gene regulatory network (GRN) with single-cell RNA-seq data. In this paper, we investigate the performances of seven classifiers including support vector machine (SVM), random forest (RF), Naive Bayesian (NB), GBDT, logical regression (LR), decision tree (DT) and K-Nearest Neighbor (KNN) for solving the binary classification problems of GRN inference with single-cell RNA-seq data (Single_cell_GRN). In SVM, three different kernel functions (linear, polynomial and radial basis function) are utilized, respectively. Three real single-cell RNA-seq datasets from mouse and human are utilized. The experiment results prove that in most cases supervised learning methods (SVM, RF, NB, GBDT, LR, DT and KNN) perform better than unsupervised learning method (GENIE3) in terms of AUC. SVM, RF and KNN have the better performances than other four classifiers. In SVM, linear and polynomial kernels are more fit to model single-cell RNA-seq data.

Androgens mediate their effects through the androgen receptor (AR) and have antiproliferative effects on uterine endometrial cells. In this report, we investigated methylation status and the expression of the AR gene in normal endometrium and uterine endometrial cancer (UEC) tissues using methylation-specific polymerase chain reaction (MSP) and immunohistochemical staining. Seventy of 89 cancer samples were AR negative, although 39 of 46 normal samples were AR positive by immunohistochemistry. By MSP, 64 of 89 cancer samples showed only methylated AR alleles, although all normal tissues showed both unmethylated and methylated AR alleles. To determine whether similar changes occurred in methylation status in the UEC carcinogenesis, we studied AR methylation using pairs of cancerous and normal samples from 28 patients. Twenty-three of 28 cancer samples showed only methylated AR alleles, although all normal samples showed both unmethylated and methylated alleles. All of the 23 cancer samples that lost unmethylated alleles were negative for AR by immunohistochemical analysis. Reverse transcription-polymerase chain reaction was performed by using UEC cell lines with and without treatment by the demethylating reagent 5-aza-2'-deoxycytidine. No AR expression was found in any of the UEC cell lines, except for MFE-296 without 5-aza-2'-deoxycytidine. Treatment with 5-aza-2'-deoxycytidine restored AR expression in all of the UEC cell lines that showed no AR expression before treatment. This study is the first to report that the possible mechanism of AR inactivation in endometrial cancer is through hypermethylation of the AR gene CpG islands.

Functional gene expression signatures (FGES) are widely used as biomarkers in cancer diagnostics as descriptive and predictive models for treatment selection. Since FGES scores are based solely on gene expression profiles, they can be applied to data from bulk and single-cell RNA-seq or gene expression microarrays. However, FGES development and implementation has several challenges including technical limitations of calculating FGES scores based on ranked gene expressions such as background noise and gene cross-correlations and difficulty in assessing the biological relevance of FGES. Here, we developed an FGES validation pipeline that addressed these limitations by examining publicly available and internally-developed FGES. The pipeline was developed by applying a modified single-sample gene set enrichment analysis (ssGSEA) to a manually curated database of more than 50,000 RNA-Seq samples, including sorted cells, cell lines, and tumor and normal tissue biopsies. We defined technical and biological criteria to validate signature quality: 1) single gene expression noise distribution minimum is greater than zero; 2) cross-correlation of genes comprising a signature is positive; 3) specificity, whereby a FGES differentiates a group of interest and has a minimal overlap with other unrelated signatures and molecular pathways; 4) generalizability, requiring FGES score to differentiate a group of interest across datasets of different origin, cell type, or diagnosis. We tested these criteria on FGES that describe cellular processes such as epithelial-mesenchymal transition (EMT) and senescence as well as FGES that can delineate cell types such as macrophages. We were able to better distinguish 12 different cell types using the corresponding cell type-specific FGES with significantly higher ssGSEA scores (p-adj ≤ 0.00004 in 99.9% cases) than 911 analogous MSigDB gene signatures. Our EMT FGES, tuned specifically for melanoma, exhibited higher precision than the MSigDB Hallmark EMT gene set (p < 0.001) for predicting histological pigment scores. The senescence FGES also accurately distinguished senescent from non-senescent cells in datasets of differing cell origin and data acquisition methods, demonstrating signature generalizability. Finally, using macrophage-describing FGES, we showed that specificity of a publicly available gene set can be significantly improved (p-adj = 0.00001) by tuning it according to the aforementioned criteria. Taken together, our proposed workflow enables better assessment of the technical quality and biological adequacy of FGES for their further use in transcriptomic analyses and computational models of complex biological processes and tumorigenesis. Citation Format: Elena Ocheredko, Nadezhda Lukashevich, Maria Savchenko, Sofia Kust, Syune Ambaryan, Dmitry Tabakov, Maria Sorokina, Linda Balabanian, Kushal Suryamohan, Anastasiya Zotova, Nathan Fowler, Aleksander Bagaev. Development and refinement of functional gene expression signatures as a computational tool for comprehensive characterization of transcriptomic data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6538.

The excessive amount of zeros in single-cell RNA-seq (scRNA-seq) data includes 'real' zeros due to the on-off nature of gene transcription in single cells and 'dropout' zeros due to technical reasons. Existing differential expression (DE) analysis methods cannot distinguish these two types of zeros. We developed an R package DEsingle which employed Zero-Inflated Negative Binomial model to estimate the proportion of real and dropout zeros and to define and detect three types of DE genes in scRNA-seq data with higher accuracy. The R package DEsingle is freely available at Bioconductor (https://bioconductor.org/packages/DEsingle). Supplementary data are available at Bioinformatics online.

This study aims to analyze the RNA expression and alternative polyadenylation (APA) events and identify APA tuned genes with prognostic significance in lung adenocarcinoma (LUAD). Genome-wide RNA expression profile and APA events were acquired in LUAD cancer and normal samples in GSE197346. Comparative analysis screened common deregulated genes and transcripts. All 11 and 19 transcripts were up and down expressed and polyadenylated in cancer samples, respectively. Clinical analysis found eight genes with prognostic significance, such as coiled-coil domain containing 137 (CCDC137). Role of CCDC137 in LUAD was first reported in this study. The cellular and animal experiments indicated that downregulated CCDC137 suppressed the malignant tumor phenotype and tumor growth in LUAD. Then, to identify APA regulators for elevated CCDC137, we analyzed the expression of 26 APA regulators in GSE197346 and The Cancer Genome Atlas (TCGA), and found 4 differential regulators: CPSF1, CELF2, NUDT21, and ELAVL1. At last, the correlation of eight genes with four differential APA regulators was analyzed, and CPSF1 showed a strong positive correlation with CCDC137. Based on the above results, we propose an oncogenic axis of CPSF1-CCDC137 in LUAD. This study first constructed a polyadenylation tuned RNA expression map in LUAD, and the proposed oncogenic axis of CPSF1-CCDC137 would shed light on the pathogenesis of LUAD.