ARTdeConv: adaptive regularized tri-factor non-negative matrix factorization for cell type deconvolution

Accurate deconvolution of cell types from bulk gene expression is crucial for understanding cellular compositions and uncovering cell-type specific differential expression and physiological states of diseased tissues. Existing deconvolution methods have limitations, such as requiring complete cellular gene expression signatures or neglecting partial biological information. Moreover, these methods often overlook varying cell-type messenger RNA amounts, leading to biased proportion estimates. Additionally, they do not effectively utilize valuable reference information from external studies, such as means and ranges of population cell-type proportions. To address these challenges, we introduce an adaptive regularized tri-factor non-negative matrix factorization approach for deconvolution (ARTdeConv). We rigorously establish the numerical convergence of our algorithm. Through benchmark simulations, we demonstrate the superior performance of ARTdeConv compared to state-of-the-art semi-reference-based and reference-free methods as well as its robustness under challenges to its assumptions. In a real-world application to a dataset from a trivalent influenza vaccine study, our method accurately estimates cellular proportions, as evidenced by the nearly perfect Pearson's correlation between ARTdeConv estimates and flow cytometry measurements. Moreover, our analysis of ARTdeConv estimates in COVID-19 patients reveals patterns consistent with important immunological phenomena observed in other studies. The proposed method, ARTdeConv, is implemented as an R package and can be accessed on GitHub for researchers and practitioners.

Cell-type deconvolution methods for spatial transcriptomics

Spatial transcriptomics is a powerful method for studying the spatial organization of cells, which is a critical feature in the development, function and evolution of multicellular life. However, sequencing-based spatial transcriptomics has not yet achieved cellular-level resolution, so advanced deconvolution methods are needed to infer cell-type contributions at each location in the data. Recent progress has led to diverse tools for cell-type deconvolution that are helping to describe tissue architectures in health and disease. In this Review, we describe the varied types of cell-type deconvolution methods for spatial transcriptomics, contrast their capabilities and summarize them in a web-based, interactive table to enable more efficient method selection.

Transcriptomic analyses of human brains with Alzheimer's disease identified dysregulated epilepsy-causing genes

Background & Objective

Alzheimer's Disease (AD) patients at multiple stages of disease progression have a high prevalence of seizures. However, whether AD and epilepsy share pathophysiological changes remains poorly defined. In this study, we leveraged high-throughput transcriptomic data from sporadic AD cases at different stages of cognitive impairment across multiple independent cohorts and brain regions to examine the role of epilepsy-causing genes.

Methods

Epilepsy-causing genes were manually curated, and their expression levels were analyzed across bulk transcriptomic data from three AD cohorts and three brain regions. RNA-seq data from sporadic AD and control cases from the Knight ADRC, MSBB, and ROSMAP cohorts were processed and analyzed under the same analytical pipeline. An integrative clustering approach employing machine learning and multi-omics data was employed to identify molecularly defined profiles with different cognitive scores.

Results

We found several epilepsy-associated genes/pathways significantly dysregulated in a group of AD patients with more severe cognitive impairment. We observed 15 genes consistently downregulated across the three cohorts, including sodium and potassium channels, suggesting that these genes play fundamental roles in cognitive function or AD progression. Notably, we found 25 of these genes dysregulated in earlier stages of AD and become worse with AD progression.

Conclusion

Our findings showed that epilepsy-causing genes showed changes in the early and late stages of AD progression, suggesting that they might be playing a role in AD progression. We can not establish directionality or cause-effect with our findings. However, changes in the epilepsy-causing genes might underlie the presence of seizures in AD patients, which might be present before or concurrently with the initial stages of AD.

Alleviating batch effects in cell type deconvolution with SCCAF-D

Cell type deconvolution methods can impute cell proportions from bulk transcriptomics data, revealing changes in disease progression or organ development. But benchmarking studies often use simulated bulk data from the same source as the reference, which limits its application scenarios. This study examines batch effects in deconvolution and introduces SCCAF-D, a computational workflow that ensures a Pearson Correlation Coefficient above 0.75 across simulated and real bulk data for various tissue types. Applied to non-alcoholic fatty liver disease, SCCAF-D unveils meaningful insights into changes in cell proportions during disease progression.

Deconvolution from bulk gene expression by leveraging sample-wise and gene-wise similarities and single-cell RNA-Seq data

BACKGROUND

The widely adopted bulk RNA-seq measures the gene expression average of cells, masking cell type heterogeneity, which confounds downstream analyses. Therefore, identifying the cellular composition and cell type-specific gene expression profiles (GEPs) facilitates the study of the underlying mechanisms of various biological processes. Although single-cell RNA-seq focuses on cell type heterogeneity in gene expression, it requires specialized and expensive resources and currently is not practical for a large number of samples or a routine clinical setting. Recently, computational deconvolution methodologies have been developed, while many of them only estimate cell type composition or cell type-specific GEPs by requiring the other as input. The development of more accurate deconvolution methods to infer cell type abundance and cell type-specific GEPs is still essential.

RESULTS

We propose a new deconvolution algorithm, DSSC, which infers cell type-specific gene expression and cell type proportions of heterogeneous samples simultaneously by leveraging gene-gene and sample-sample similarities in bulk expression and single-cell RNA-seq data. Through comparisons with the other existing methods, we demonstrate that DSSC is effective in inferring both cell type proportions and cell type-specific GEPs across simulated pseudo-bulk data (including intra-dataset and inter-dataset simulations) and experimental bulk data (including mixture data and real experimental data). DSSC shows robustness to the change of marker gene number and sample size and also has cost and time efficiencies.

CONCLUSIONS

DSSC provides a practical and promising alternative to the experimental techniques to characterize cellular composition and heterogeneity in the gene expression of heterogeneous samples.

The MIND diet, brain transcriptomic alterations, and dementia

INTRODUCTION

Dietary patterns are associated with dementia risk, but the underlying molecular mechanisms are largely unknown.

METHODS

We used RNA sequencing data from post mortem prefrontal cortex tissue and annual cognitive evaluations from 1204 participants in the Religious Orders Study and Memory and Aging Project. We identified a transcriptomic profile correlated with the MIND diet (Mediterranean-Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay) among 482 individuals who completed ante mortem food frequency questionnaires; and examined its associations with cognitive health in the remaining 722 participants.

RESULTS

We identified a transcriptomic profile, consisting of 50 genes, correlated with the MIND diet score (p = 0.001). Each standard deviation increase in the transcriptomic profile score was associated with a slower annual rate of decline in global cognition (β = 0.011, p = 0.003) and lower odds of dementia (odds ratio = 0.76, p = 0.0002). Expressions of several genes (including TCIM and IGSF5) appeared to mediate the association between MIND diet and dementia.

DISCUSSION

A brain transcriptomic profile for healthy diets revealed novel genes potentially associated with cognitive health.

HIGHLIGHTS

Why healthy dietary patterns are associated with lower dementia risk are unknown. We integrated dietary, brain transcriptomic, and cognitive data in older adults. Mediterranean-Dietary Approaches to Stop Hypertension Intervention for Neurodegenerative Delay (MIND) diet intake is correlated with a specific brain transcriptomic profile. This brain transcriptomic profile score is associated with better cognitive health. More data are needed to elucidate the causality and functionality of identified genes.

Adaptive Regularized Tri-Factor Non-Negative Matrix Factorization for Cell Type Deconvolution

Accurate deconvolution of cell types from bulk gene expression is crucial for understanding cellular compositions and uncovering cell-type specific differential expression and physiological states of diseased tissues. Existing deconvolution methods have limitations, such as requiring complete cellular gene expression signatures or neglecting partial biological information. Moreover, these methods often overlook varying cell-type mRNA amounts, leading to biased proportion estimates. Additionally, they do not effectively utilize valuable reference information from external studies, such as means and ranges of population cell-type proportions. To address these challenges, we introduce an Adaptive Regularized Tri-factor non-negative matrix factorization approach for deconvolution (ARTdeConv). We rigorously establish the numerical convergence of our algorithm. Through benchmark simulations, we demonstrate the superior performance of ARTdeConv compared to state-of-the-art semi-reference-based and reference-free methods. In a real-world application, our method accurately estimates cell proportions, as evidenced by the nearly perfect Pearson's correlation between ARTdeConv estimates and flow cytometry measurements in a dataset from a trivalent influenza vaccine study. Moreover, our analysis of ARTdeConv estimates in COVID-19 patients reveals patterns consistent with important immunological phenomena observed in other studies. The proposed method, ARTdeConv, is implemented as an R package and can be accessed on GitHub for researchers and practitioners.

Assessing transcriptomic heterogeneity of single-cell RNASeq data by bulk-level gene expression data

BACKGROUND

Single-cell RNA sequencing (sc-RNASeq) data illuminate transcriptomic heterogeneity but also possess a high level of noise, abundant missing entries and sometimes inadequate or no cell type annotations at all. Bulk-level gene expression data lack direct information of cell population composition but are more robust and complete and often better annotated. We propose a modeling framework to integrate bulk-level and single-cell RNASeq data to address the deficiencies and leverage the mutual strengths of each type of data and enable a more comprehensive inference of their transcriptomic heterogeneity. Contrary to the standard approaches of factorizing the bulk-level data with one algorithm and (for some methods) treating single-cell RNASeq data as references to decompose bulk-level data, we employed multiple deconvolution algorithms to factorize the bulk-level data, constructed the probabilistic graphical models of cell-level gene expressions from the decomposition outcomes, and compared the log-likelihood scores of these models in single-cell data. We term this framework backward deconvolution as inference operates from coarse-grained bulk-level data to fine-grained single-cell data. As the abundant missing entries in sc-RNASeq data have a significant effect on log-likelihood scores, we also developed a criterion for inclusion or exclusion of zero entries in log-likelihood score computation.

RESULTS

We selected nine deconvolution algorithms and validated backward deconvolution in five datasets. In the in-silico mixtures of mouse sc-RNASeq data, the log-likelihood scores of the deconvolution algorithms were strongly anticorrelated with their errors of mixture coefficients and cell type specific gene expression signatures. In the true bulk-level mouse data, the sample mixture coefficients were unknown but the log-likelihood scores were strongly correlated with accuracy rates of inferred cell types. In the data of autism spectrum disorder (ASD) and normal controls, we found that ASD brains possessed higher fractions of astrocytes and lower fractions of NRGN-expressing neurons than normal controls. In datasets of breast cancer and low-grade gliomas (LGG), we compared the log-likelihood scores of three simple hypotheses about the gene expression patterns of the cell types underlying the tumor subtypes. The model that tumors of each subtype were dominated by one cell type persistently outperformed an alternative model that each cell type had elevated expression in one gene group and tumors were mixtures of those cell types. Superiority of the former model is also supported by comparing the real breast cancer sc-RNASeq clusters with those generated by simulated sc-RNASeq data.

CONCLUSIONS

The results indicate that backward deconvolution serves as a sensible model selection tool for deconvolution algorithms and facilitates discerning hypotheses about cell type compositions underlying heterogeneous specimens such as tumors.

Brain high-throughput multi-omics data reveal molecular heterogeneity in Alzheimer's disease

Unbiased data-driven omic approaches are revealing the molecular heterogeneity of Alzheimer disease. Here, we used machine learning approaches to integrate high-throughput transcriptomic, proteomic, metabolomic, and lipidomic profiles with clinical and neuropathological data from multiple human AD cohorts. We discovered 4 unique multimodal molecular profiles, one of them showing signs of poor cognitive function, a faster pace of disease progression, shorter survival with the disease, severe neurodegeneration and astrogliosis, and reduced levels of metabolomic profiles. We found this molecular profile to be present in multiple affected cortical regions associated with higher Braak tau scores and significant dysregulation of synapse-related genes, endocytosis, phagosome, and mTOR signaling pathways altered in AD early and late stages. AD cross-omics data integration with transcriptomic data from an SNCA mouse model revealed an overlapping signature. Furthermore, we leveraged single-nuclei RNA-seq data to identify distinct cell-types that most likely mediate molecular profiles. Lastly, we identified that the multimodal clusters uncovered cerebrospinal fluid biomarkers poised to monitor AD progression and possibly cognition. Our cross-omics analyses provide novel critical molecular insights into AD.

CATD: a reproducible pipeline for selecting cell-type deconvolution methods across tissues

Motivation

Cell-type deconvolution methods aim to infer cell composition from bulk transcriptomic data. The proliferation of developed methods coupled with inconsistent results obtained in many cases, highlights the pressing need for guidance in the selection of appropriate methods. Additionally, the growing accessibility of single-cell RNA sequencing datasets, often accompanied by bulk expression from related samples enable the benchmark of existing methods.

Results

In this study, we conduct a comprehensive assessment of 31 methods, utilizing single-cell RNA-sequencing data from diverse human and mouse tissues. Employing various simulation scenarios, we reveal the efficacy of regression-based deconvolution methods, highlighting their sensitivity to reference choices. We investigate the impact of bulk-reference differences, incorporating variables such as sample, study and technology. We provide validation using a gold standard dataset from mononuclear cells and suggest a consensus prediction of proportions when ground truth is not available. We validated the consensus method on data from the stomach and studied its spillover effect. Importantly, we propose the use of the critical assessment of transcriptomic deconvolution (CATD) pipeline which encompasses functionalities for generating references and pseudo-bulks and running implemented deconvolution methods. CATD streamlines simultaneous deconvolution of numerous bulk samples, providing a practical solution for speeding up the evaluation of newly developed methods.

Availability and implementation

https://github.com/Papatheodorou-Group/CATD_snakemake.

Identifying lupus Patient Subsets Through Immune Cell Deconvolution of Gene Expression Data in Two Atacicept Phase II Studies

OBJECTIVE

To use cell-based gene signatures to identify patients with systemic lupus erythematous (SLE) in the phase II/III APRIL-SLE and phase IIb ADDRESS II trials most likely to respond to atacicept.

METHODS

A published immune cell deconvolution algorithm based on Affymetrix gene array data was applied to whole blood gene expression from patients entering APRIL-SLE. Five distinct patient clusters were identified. Patient characteristics, biomarkers, and clinical response to atacicept were assessed per cluster. A modified immune cell deconvolution algorithm was developed based on RNA sequencing data and applied to ADDRESS II data to identify similar patient clusters and their responses.

RESULTS

Patients in APRIL-SLE (N = 105) were segregated into the following five clusters (P1-5) characterized by dominant cell subset signatures: high neutrophils, T helper cells and natural killer (NK) cells (P1), high plasma cells and activated NK cells (P2), high B cells and neutrophils (P3), high B cells and low neutrophils (P4), or high activated dendritic cells, activated NK cells, and neutrophils (P5). Placebo- and atacicept-treated patients in clusters P2,4,5 had markedly higher British Isles Lupus Assessment Group (BILAG) A/B flare rates than those in clusters P1,3, with a greater treatment effect of atacicept on lowering flares in clusters P2,4,5. In ADDRESS II, placebo-treated patients from P2,4,5 were less likely to be SLE Responder Index (SRI)-4, SRI-6, and BILAG-Based Combined Lupus Assessment responders than those in P1,3; the response proportions again suggested lower placebo effect and a greater treatment differential for atacicept in P2,4,5.

CONCLUSION

This exploratory analysis indicates larger differences between placebo- and atacicept-treated patients with SLE in a molecularly defined patient subset.

The MIND diet, brain transcriptomic alterations, and dementia

Identifying novel mechanisms underlying dementia is critical to improving prevention and treatment. As an approach to mechanistic discovery, we investigated whether MIND diet (Mediterranean-DASH Diet Intervention for Neurodegenerative Delay), a consistent risk factor for dementia, is correlated with a specific profile of cortical gene expression, and whether such a transcriptomic profile is associated with dementia, in the Religious Orders Study (ROS) and Rush Memory and Aging Project (MAP). RNA sequencing (RNA-Seq) was conducted in postmortem dorsolateral prefrontal cortex tissue from 1,204 deceased participants; neuropsychological assessments were performed annually prior to death. In a subset of 482 participants, diet was assessed ~6 years before death using a validated food-frequency questionnaire; in these participants, using elastic net regression, we identified a transcriptomic profile, consisting of 50 genes, significantly correlated with MIND diet score (P=0.001). In multivariable analysis of the remaining 722 individuals, higher transcriptomic score of MIND diet was associated with slower annual rate of decline in global cognition (β=0.011 per standard deviation increment in transcriptomic profile score, P=0.003) and lower odds of dementia (odds ratio [OR] =0.76, P=0.0002). Cortical expression of several genes appeared to mediate the association between MIND diet and dementia, including TCIM, whose expression in inhibitory neurons and oligodendrocytes was associated with dementia in a subset of 424 individuals with single-nuclei RNA-seq data. In a secondary Mendelian randomization analysis, genetically predicted transcriptomic profile score was associated with dementia (OR=0.93, P=0.04). Our study suggests that associations between diet and cognitive health may involve brain molecular alterations at the transcriptomic level. Investigating brain molecular alterations related to diet may inform the identification of novel pathways underlying dementia.

Expression quantitative trait locus studies in the era of single-cell omics

Genome-wide association studies have revealed that the regulation of gene expression bridges genetic variants and complex phenotypes. Profiling of the bulk transcriptome coupled with linkage analysis (expression quantitative trait locus (eQTL) mapping) has advanced our understanding of the relationship between genetic variants and gene regulation in the context of complex phenotypes. However, bulk transcriptomics has inherited limitations as the regulation of gene expression tends to be cell-type-specific. The advent of single-cell RNA-seq technology now enables the identification of the cell-type-specific regulation of gene expression through a single-cell eQTL (sc-eQTL). In this review, we first provide an overview of sc-eQTL studies, including data processing and the mapping procedure of the sc-eQTL. We then discuss the benefits and limitations of sc-eQTL analyses. Finally, we present an overview of the current and future applications of sc-eQTL discoveries.

CeDAR: incorporating cell type hierarchy improves cell type-specific differential analyses in bulk omics data

Bulk high-throughput omics data contain signals from a mixture of cell types. Recent developments of deconvolution methods facilitate cell type-specific inferences from bulk data. Our real data exploration suggests that differential expression or methylation status is often correlated among cell types. Based on this observation, we develop a novel statistical method named CeDAR to incorporate the cell type hierarchy in cell type-specific differential analyses of bulk data. Extensive simulation and real data analyses demonstrate that this approach significantly improves the accuracy and power in detecting cell type-specific differential signals compared with existing methods, especially in low-abundance cell types.

MS4A4A modifies the risk of Alzheimer disease by regulating lipid metabolism and immune response in a unique microglia state

Genome-wide association studies (GWAS) have identified many modifiers of Alzheimer disease (AD) risk enriched in microglia. Two of these modifiers are common variants in the MS4A locus (rs1582763: protective and rs6591561: risk) and serve as major regulators of CSF sTREM2 levels. To understand their functional impact on AD, we used single nucleus transcriptomics to profile brains from carriers of these variants. We discovered a "chemokine" microglial subpopulation that is altered in MS4A variant carriers and for which MS4A4A is the major regulator. The protective variant increases MS4A4A expression and shifts the chemokine microglia subpopulation to an interferon state, while the risk variant suppresses MS4A4A expression and reduces this subpopulation of microglia. Our findings provide a mechanistic explanation for the AD variants in the MS4A locus. Further, they pave the way for future mechanistic studies of AD variants and potential therapeutic strategies for enhancing microglia resilience in AD pathogenesis.

Characterizing the composition of iPSC derived cells from bulk transcriptomics data with CellMap

Induced pluripotent stem cell (iPSC) derived cell types are increasingly employed as in vitro model systems for drug discovery. For these studies to be meaningful, it is important to understand the reproducibility of the iPSC-derived cultures and their similarity to equivalent endogenous cell types. Single-cell and single-nucleus RNA sequencing (RNA-seq) are useful to gain such understanding, but they are expensive and time consuming, while bulk RNA-seq data can be generated quicker and at lower cost. In silico cell type decomposition is an efficient, inexpensive, and convenient alternative that can leverage bulk RNA-seq to derive more fine-grained information about these cultures. We developed CellMap, a computational tool that derives cell type profiles from publicly available single-cell and single-nucleus datasets to infer cell types in bulk RNA-seq data from iPSC-derived cell lines.

DURIAN: an integrative deconvolution and imputation method for robust signaling analysis of single-cell transcriptomics data

Single-cell RNA sequencing trades read-depth for dimensionality, often leading to loss of critical signaling gene information that is typically present in bulk data sets. We introduce DURIAN (Deconvolution and mUltitask-Regression-based ImputAtioN), an integrative method for recovery of gene expression in single-cell data. Through systematic benchmarking, we demonstrate the accuracy, robustness and empirical convergence of DURIAN using both synthetic and published data sets. We show that use of DURIAN improves single-cell clustering, low-dimensional embedding, and recovery of intercellular signaling networks. Our study resolves several inconsistent results of cell-cell communication analysis using single-cell or bulk data independently. The method has broad application in biomarker discovery and cell signaling analysis using single-cell transcriptomics data sets.

Robust and accurate estimation of cellular fraction from tissue omics data via ensemble deconvolution

MOTIVATION

Tissue-level omics data such as transcriptomics and epigenomics are an average across diverse cell types. To extract cell-type-specific (CTS) signals, dozens of cellular deconvolution methods have been proposed to infer cell-type fractions from tissue-level data. However, these methods produce vastly different results under various real data settings. Simulation-based benchmarking studies showed no universally best deconvolution approaches. There have been attempts of ensemble methods, but they only aggregate multiple single-cell references or reference-free deconvolution methods.

RESULTS

To achieve a robust estimation of cellular fractions, we proposed EnsDeconv (Ensemble Deconvolution), which adopts CTS robust regression to synthesize the results from 11 single deconvolution methods, 10 reference datasets, 5 marker gene selection procedures, 5 data normalizations and 2 transformations. Unlike most benchmarking studies based on simulations, we compiled four large real datasets of 4937 tissue samples in total with measured cellular fractions and bulk gene expression from different tissues. Comprehensive evaluations demonstrated that EnsDeconv yields more stable, robust and accurate fractions than existing methods. We illustrated that EnsDeconv estimated cellular fractions enable various CTS downstream analyses such as differential fractions associated with clinical variables. We further extended EnsDeconv to analyze bulk DNA methylation data.

AVAILABILITY AND IMPLEMENTATION

EnsDeconv is freely available as an R-package from https://github.com/randel/EnsDeconv. The RNA microarray data from the TRAUMA study are available and can be accessed in GEO (GSE36809). The demographic and clinical phenotypes can be shared on reasonable request to the corresponding authors. The RNA-seq data from the EVAPR study cannot be shared publicly due to the privacy of individuals that participated in the clinical research in compliance with the IRB approval at the University of Pittsburgh. The RNA microarray data from the FHS study are available from dbGaP (phs000007.v32.p13). The RNA-seq data from ROS study is downloaded from AD Knowledge Portal.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Macrophage inflammation resolution requires CPEB4-directed offsetting of mRNA degradation

Chronic inflammation is a major cause of disease. Inflammation resolution is in part directed by the differential stability of mRNAs encoding pro-inflammatory and anti-inflammatory factors. In particular, tristetraprolin (TTP)-directed mRNA deadenylation destabilizes AU-rich element (ARE)-containing mRNAs. However, this mechanism alone cannot explain the variety of mRNA expression kinetics that are required to uncouple degradation of pro-inflammatory mRNAs from the sustained expression of anti-inflammatory mRNAs. Here, we show that the RNA-binding protein CPEB4 acts in an opposing manner to TTP in macrophages: it helps to stabilize anti-inflammatory transcripts harboring cytoplasmic polyadenylation elements (CPEs) and AREs in their 3′-UTRs, and it is required for the resolution of the lipopolysaccharide (LPS)-triggered inflammatory response. Coordination of CPEB4 and TTP activities is sequentially regulated through MAPK signaling. Accordingly, CPEB4 depletion in macrophages impairs inflammation resolution in an LPS-induced sepsis model. We propose that the counterbalancing actions of CPEB4 and TTP, as well as the distribution of CPEs and AREs in their target mRNAs, define transcript-specific decay patterns required for inflammation resolution. Thus, these two opposing mechanisms provide a fine-tuning control of inflammatory transcript destabilization while maintaining the expression of the negative feedback loops required for efficient inflammation resolution; disruption of this balance can lead to disease.

Differential Changes in ACPA Fine Specificity and Gene Expression in a Randomized Trial of Abatacept and Adalimumab in Rheumatoid Arthritis

INTRODUCTION

The biologics abatacept and adalimumab have different mechanisms of action (MoAs). We analyzed data from patients with rheumatoid arthritis treated in AMPLE (NCT00929864) to explore the pharmacodynamic effects of abatacept or adalimumab on anti-citrullinated protein antibodies (ACPAs) and gene expression.

METHODS

AMPLE was a phase IIIb, 2-year, randomized, head-to-head trial of abatacept versus adalimumab. Post hoc analyses of baseline anti-cyclic citrullinated peptide-2 (anti-CCP2, an ACPA surrogate) positive (+) status and ACPA fine-specificity profiles over time, as well as transcriptional profiling (peripheral whole blood), were performed.

RESULTS

Of 646 patients treated (abatacept, n = 318; adalimumab, n = 328), ACPA and gene expression data were available from 508 and 566 patients, respectively. In anti-CCP2+ patients (n = 388), baseline fine specificities for most ACPAs were highly correlated; over 2 years, levels decreased with abatacept but not adalimumab. By year 2, expression of genes associated with T cell co-stimulation and antibody production was lower for abatacept versus adalimumab; expression of genes associated with proinflammatory signaling was lower for adalimumab versus abatacept. Treatment modulated the expression of T- and B-cell gene signatures, with differences in CD8+ T cells, activated T cells, plasma cells, B cells, natural killer cells (all lower with abatacept versus adalimumab), and polymorphonuclear leukocytes (higher with abatacept versus adalimumab).

CONCLUSIONS

In AMPLE, despite similar clinical outcomes, data showed that pharmacodynamic/genetic changes after 2 years of abatacept or adalimumab were consistent with drug MoAs. Further assessment of the relationship between such changes and clinical outcomes, including prediction of response, is warranted.

TRIAL REGISTRATION

ClinicalTrials.gov identifier, NCT00929864.

Identification of Key Modules and Genes Associated with Major Depressive Disorder in Adolescents

Major depressive disorder (MDD) is a leading cause of disability worldwide. Adolescence is a crucial period for the occurrence and development of depression. There are essential distinctions between adolescent and adult depression patients, and the etiology of depressive disorder is unclear. The interactions of multiple genes in a co-expression network are likely to be involved in the physiopathology of MDD. In the present study, RNA-Seq data of mRNA were acquired from the peripheral blood of MDD in adolescents and healthy control (HC) subjects. Co-expression modules were constructed via weighted gene co-expression network analysis (WGCNA) to investigate the relationships between the underlying modules and MDD in adolescents. In the combined MDD and HC groups, the dynamic tree cutting method was utilized to assign genes to modules through hierarchical clustering. Moreover, functional enrichment analysis was conducted on those co-expression genes from interested modules. The results showed that eight modules were constructed by WGCNA. The blue module was significantly associated with MDD after multiple comparison adjustment. Several Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways associated with stress and inflammation were identified in this module, including histone methylation, apoptosis, NF-kappa β signaling pathway, and TNF signaling pathway. Five genes related to inflammation, immunity, and the nervous system were identified as hub genes: CNTNAP3, IL1RAP, MEGF9, UBE2W, and UBE2D1. All of these findings supported that MDD was associated with stress, inflammation, and immune responses, helping us to obtain a better understanding of the internal molecular mechanism and to explore biomarkers for the diagnosis or treatment of depression in adolescents.

Guidelines for bioinformatics of single-cell sequencing data analysis in Alzheimer's disease: review, recommendation, implementation and application

Alzheimer's disease (AD) is the most common form of dementia, characterized by progressive cognitive impairment and neurodegeneration. Extensive clinical and genomic studies have revealed biomarkers, risk factors, pathways, and targets of AD in the past decade. However, the exact molecular basis of AD development and progression remains elusive. The emerging single-cell sequencing technology can potentially provide cell-level insights into the disease. Here we systematically review the state-of-the-art bioinformatics approaches to analyze single-cell sequencing data and their applications to AD in 14 major directions, including 1) quality control and normalization, 2) dimension reduction and feature extraction, 3) cell clustering analysis, 4) cell type inference and annotation, 5) differential expression, 6) trajectory inference, 7) copy number variation analysis, 8) integration of single-cell multi-omics, 9) epigenomic analysis, 10) gene network inference, 11) prioritization of cell subpopulations, 12) integrative analysis of human and mouse sc-RNA-seq data, 13) spatial transcriptomics, and 14) comparison of single cell AD mouse model studies and single cell human AD studies. We also address challenges in using human postmortem and mouse tissues and outline future developments in single cell sequencing data analysis. Importantly, we have implemented our recommended workflow for each major analytic direction and applied them to a large single nucleus RNA-sequencing (snRNA-seq) dataset in AD. Key analytic results are reported while the scripts and the data are shared with the research community through  GitHub. In summary, this comprehensive review provides insights into various approaches to analyze single cell sequencing data and offers specific guidelines for study design and a variety of analytic directions. The review and the accompanied software tools will serve as a valuable resource for studying cellular and molecular mechanisms of AD, other diseases, or biological systems at the single cell level.

Computational challenges in detection of cancer using cell-free DNA methylation

Cell-free DNA(cfDNA) methylation profiling is considered promising and potentially reliable for liquid biopsy to study progress of diseases and develop reliable and consistent diagnostic and prognostic biomarkers. There are several different mechanisms responsible for the release of cfDNA in blood plasma, and henceforth it can provide information regarding dynamic changes in the human body. Due to the fragmented nature, low concentration of cfDNA, and high background noise, there are several challenges in its analysis for regular use in diagnosis of cancer. Such challenges in the analysis of the methylation profile of cfDNA are further aggravated due to heterogeneity, biomarker sensitivity, platform biases, and batch effects. This review delineates the origin of cfDNA methylation, its profiling, and associated computational problems in analysis for diagnosis. Here we also contemplate upon the multi-marker approach to handle the scenario of cancer heterogeneity and explore the utility of markers for 5hmC based cfDNA methylation pattern. Further, we provide a critical overview of deconvolution and machine learning methods for cfDNA methylation analysis. Our review of current methods reveals the potential for further improvement in analysis strategies for detecting early cancer using cfDNA methylation.

Besca, a single-cell transcriptomics analysis toolkit to accelerate translational research

Single-cell RNA sequencing (scRNA-seq) revolutionized our understanding of disease biology. The promise it presents to also transform translational research requires highly standardized and robust software workflows. Here, we present the toolkit Besca, which streamlines scRNA-seq analyses and their use to deconvolute bulk RNA-seq data according to current best practices. Beyond a standard workflow covering quality control, filtering, and clustering, two complementary Besca modules, utilizing hierarchical cell signatures and supervised machine learning, automate cell annotation and provide harmonized nomenclatures. Subsequently, the gene expression profiles can be employed to estimate cell type proportions in bulk transcriptomics data. Using multiple, diverse scRNA-seq datasets, some stemming from highly heterogeneous tumor tissue, we show how Besca aids acceleration, interoperability, reusability and interpretability of scRNA-seq data analyses, meeting crucial demands in translational research and beyond.

Identification of cell-type-specific marker genes from co-expression patterns in tissue samples

MOTIVATION

Marker genes, defined as genes that are expressed primarily in a single-cell type, can be identified from the single-cell transcriptome; however, such data are not always available for the many uses of marker genes, such as deconvolution of bulk tissue. Marker genes for a cell type, however, are highly correlated in bulk data, because their expression levels depend primarily on the proportion of that cell type in the samples. Therefore, when many tissue samples are analyzed, it is possible to identify these marker genes from the correlation pattern.

RESULTS

To capitalize on this pattern, we develop a new algorithm to detect marker genes by combining published information about likely marker genes with bulk transcriptome data in the form of a semi-supervised algorithm. The algorithm then exploits the correlation structure of the bulk data to refine the published marker genes by adding or removing genes from the list.

AVAILABILITY AND IMPLEMENTATION

We implement this method as an R package markerpen, hosted on CRAN (https://CRAN.R-project.org/package=markerpen).

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Diurnal regulation of oxidative phosphorylation restricts hepatocyte proliferation and inflammation

The principles guiding the diurnal organization of biological pathways remain to be fully elucidated. Here, we perturb the hepatic transcriptome through nutrient regulators (high-fat diet and mTOR signaling components) to identify enduring properties of pathway organization. Temporal separation and counter-regulation between pathways of energy metabolism and inflammation/proliferation emerge as persistent transcriptome features across animal models, and network analysis identifies the G0s2 and Rgs16 genes as potential mediators at the metabolism-inflammation interface. Mechanistically, G0s2 and Rgs16 are sequentially induced during the light phase, promoting amino acid oxidation and suppressing overall mitochondrial respiration. In their absence, sphingolipids and diacylglycerides accumulate, accompanied by hepatic inflammation and hepatocyte proliferation. Notably, the expression of G0s2 and Rgs16 is further induced in obese mouse livers, and silencing of their expression accentuates hepatic fibrosis. Therefore, diurnal regulation of energy metabolism alleviates inflammatory and proliferative stresses under physiological and pathological conditions.

Transcriptome Analysis of Post-Mortem Brain Tissue Reveals Up-Regulation of the Complement Cascade in a Subgroup of Schizophrenia Patients

Schizophrenia is a genetically complex neuropsychiatric disorder with largely unresolved mechanisms of pathology. Identification of genes and pathways associated with schizophrenia is important for understanding the development, progression and treatment of schizophrenia. In this study, pathways associated with schizophrenia were explored at the level of gene expression. The study included post-mortem brain tissue samples from 68 schizophrenia patients and 44 age and sex-matched control subjects. Whole transcriptome poly-A selected paired-end RNA sequencing was performed on tissue from the prefrontal cortex and orbitofrontal cortex. RNA expression differences were detected between case and control individuals, focusing both on single genes and pathways. The results were validated with RT-qPCR. Significant differential expression between patient and controls groups was found for 71 genes. Gene ontology analysis of differentially expressed genes revealed an up-regulation of multiple genes in immune response among the patients (corrected p-value = 0.004). Several genes in the category belong to the complement system, including C1R, C1S, C7, FCN3, SERPING1, C4A and CFI. The increased complement expression is primarily driven by a subgroup of patients with increased expression of immune/inflammatory response genes, pointing to important differences in disease etiology within the patient group. Weighted gene co-expression network analysis highlighted networks associated with both synaptic transmission and activation of the immune response. Our results demonstrate the importance of immune-related pathways in schizophrenia and provide evidence for elevated expression of the complement cascade as an important pathway in schizophrenia pathology.

SSMD: a semi-supervised approach for a robust cell type identification and deconvolution of mouse transcriptomics data

Deconvolution of mouse transcriptomic data is challenged by the fact that mouse models carry various genetic and physiological perturbations, making it questionable to assume fixed cell types and cell type marker genes for different data set scenarios. We developed a Semi-Supervised Mouse data Deconvolution (SSMD) method to study the mouse tissue microenvironment. SSMD is featured by (i) a novel nonparametric method to discover data set-specific cell type signature genes; (ii) a community detection approach for fixing cell types and their marker genes; (iii) a constrained matrix decomposition method to solve cell type relative proportions that is robust to diverse experimental platforms. In summary, SSMD addressed several key challenges in the deconvolution of mouse tissue data, including: (i) varied cell types and marker genes caused by highly divergent genotypic and phenotypic conditions of mouse experiment; (ii) diverse experimental platforms of mouse transcriptomics data; (iii) small sample size and limited training data source and (iv) capable to estimate the proportion of 35 cell types in blood, inflammatory, central nervous or hematopoietic systems. In silico and experimental validation of SSMD demonstrated its high sensitivity and accuracy in identifying (sub) cell types and predicting cell proportions comparing with state-of-the-arts methods. A user-friendly R package and a web server of SSMD are released via https://github.com/xiaoyulu95/SSMD.

Infiltrating T-cell abundance combined with EMT-related gene expression as a prognostic factor of colon cancer

EMT-related gene expression reportedly exhibits correlation with the anti-tumor immunity of T cells. In the present study, we explored the factors that might affect the efficacy of immunotherapy in colon cancer with treatment. In this regard, RNA-seq and clinical data of 469 colon cancer samples derived from the Cancer Genome Atlas (TCGA) database were used to calculate infiltrating T-cell abundance (ITA), to illustrate a pathway enrichment analysis, and to construct Cox proportional hazards (CPH) regression models. Subsequently, the RNA-seq and clinical data of 177 colon cancer samples derived from the GSE17536 cohort were used to validate the CPH regression models. We found that ITA showed correlation with EMT-related gene expression, and that it was not an independent prognostic factor for colon cancer. However, upon comparison of two groups with the same ITA, higher EMT expression helped predicted a worse prognosis, whereas a higher ITA could help predict a better prognosis upon comparison of two groups with the same EMT. Additionally, seven genes were found to be statistically related to the prognosis of patients with colon cancer. These results suggest that the balance between ITA and EMT-related gene expression is conducive to the prognosis of patients with colon cancer, and TPM1 is necessary to further explore the common target genes of immune checkpoint blockade.

Intratumor CMS Heterogeneity Impacts Patient Prognosis in Localized Colon Cancer

PURPOSE

The consensus molecular subtypes (CMS) represent a significant advance in the understanding of intertumor heterogeneity in colon cancer. Intratumor heterogeneity (ITH) is the new frontier for refining prognostication and understanding treatment resistance. This study aims at deciphering the transcriptomic ITH of colon cancer and understanding its potential prognostic implications.

EXPERIMENTAL DESIGN

We deconvoluted the transcriptomic profiles of 1,779 tumors from the PETACC8 trial and 155 colon cancer cell lines as weighted sums of the four CMSs, using the Weighted In Silico Pathology (WISP) algorithm. We assigned to each tumor and cell line a combination of up to three CMS subtypes with a threshold above 20%.

RESULTS

Over 55% of tumors corresponded to mixtures of at least two CMSs, demonstrating pervasive ITH in colon cancer. Of note, ITH was associated with shorter disease-free survival (DFS) and overall survival, [HR, 1.34; 95% confidence interval (CI; 1.12-1.59), 1.40, 95% CI (1.14-1.71), respectively]. Moreover, we uncovered specific combinations of CMS associated with dismal prognosis. In multivariate analysis, ITH represents the third parameter explaining DFS variance, after T and N stages. At a cellular level, combined WISP and single-cell transcriptomic analysis revealed that most colon cancer cell lines are a mixture of cells falling into different CMSs, indicating that ITH may correspond to distinct functional statuses of colon cancer cells.

CONCLUSIONS

This study shows that CMS-based transcriptomic ITH is frequent in colon cancer and impacts its prognosis. CMS-based transcriptomic ITH may correspond to distinct functional statuses of colon cancer cells, suggesting plasticity between CMS-related cell populations. Transcriptomic ITH deserves further assessment in the context of personalized medicine.

A computational method for direct imputation of cell type-specific expression profiles and cellular compositions from bulk-tissue RNA-Seq in brain disorders

The importance of cell type-specific gene expression in disease-relevant tissues is increasingly recognized in genetic studies of complex diseases. However, most gene expression studies are conducted on bulk tissues, without examining cell type-specific expression profiles. Several computational methods are available for cell type deconvolution (i.e. inference of cellular composition) from bulk RNA-Seq data, but few of them impute cell type-specific expression profiles. We hypothesize that with external prior information such as single cell RNA-seq and population-wide expression profiles, it can be computationally tractable to estimate both cellular composition and cell type-specific expression from bulk RNA-Seq data. Here we introduce CellR, which addresses cross-individual gene expression variations to adjust the weights of cell-specific gene markers. It then transforms the deconvolution problem into a linear programming model while taking into account inter/intra cellular correlations and uses a multi-variate stochastic search algorithm to estimate the cell type-specific expression profiles. Analyses on several complex diseases such as schizophrenia, Alzheimer’s disease, Huntington’s disease and type 2 diabetes validated the efficiency of CellR, while revealing how specific cell types contribute to different diseases. In summary, CellR compares favorably against competing approaches, enabling cell type-specific re-analysis of gene expression data on bulk tissues in complex diseases.

stochprofML: stochastic profiling using maximum likelihood estimation in R

BACKGROUND

Tissues are often heterogeneous in their single-cell molecular expression, and this can govern the regulation of cell fate. For the understanding of development and disease, it is important to quantify heterogeneity in a given tissue.

RESULTS

We present the R package stochprofML which uses the maximum likelihood principle to parameterize heterogeneity from the cumulative expression of small random pools of cells. We evaluate the algorithm's performance in simulation studies and present further application opportunities.

CONCLUSION

Stochastic profiling outweighs the necessary demixing of mixed samples with a saving in experimental cost and effort and less measurement error. It offers possibilities for parameterizing heterogeneity, estimating underlying pool compositions and detecting differences between cell populations between samples.

Computational deconvolution to estimate cell type-specific gene expression from bulk data

Computational deconvolution is a time and cost-efficient approach to obtain cell type-specific information from bulk gene expression of heterogeneous tissues like blood. Deconvolution can aim to either estimate cell type proportions or abundances in samples, or estimate how strongly each present cell type expresses different genes, or both tasks simultaneously. Among the two separate goals, the estimation of cell type proportions/abundances is widely studied, but less attention has been paid on defining the cell type-specific expression profiles. Here, we address this gap by introducing a novel method Rodeo and empirically evaluating it and the other available tools from multiple perspectives utilizing diverse datasets.

Unveiling the immune infiltrate modulation in cancer and response to immunotherapy by MIXTURE-an enhanced deconvolution method

The accurate quantification of tumor-infiltrating immune cells turns crucial to uncover their role in tumor immune escape, to determine patient prognosis and to predict response to immune checkpoint blockade. Current state-of-the-art methods that quantify immune cells from tumor biopsies using gene expression data apply computational deconvolution methods that present multicollinearity and estimation errors resulting in the overestimation or underestimation of the diversity of infiltrating immune cells and their quantity. To overcome such limitations, we developed MIXTURE, a new ν-support vector regression-based noise constrained recursive feature selection algorithm based on validated immune cell molecular signatures. MIXTURE provides increased robustness to cell type identification and proportion estimation, outperforms the current methods, and is available to the wider scientific community. We applied MIXTURE to transcriptomic data from tumor biopsies and found relevant novel associations between the components of the immune infiltrate and molecular subtypes, tumor driver biomarkers, tumor mutational burden, microsatellite instability, intratumor heterogeneity, cytolytic score, programmed cell death ligand 1 expression, patients' survival and response to anti-cytotoxic T-lymphocyte-associated antigen 4 and anti-programmed cell death protein 1 immunotherapy.

Benchmarking of cell type deconvolution pipelines for transcriptomics data

Many computational methods have been developed to infer cell type proportions from bulk transcriptomics data. However, an evaluation of the impact of data transformation, pre-processing, marker selection, cell type composition and choice of methodology on the deconvolution results is still lacking. Using five single-cell RNA-sequencing (scRNA-seq) datasets, we generate pseudo-bulk mixtures to evaluate the combined impact of these factors. Both bulk deconvolution methodologies and those that use scRNA-seq data as reference perform best when applied to data in linear scale and the choice of normalization has a dramatic impact on some, but not all methods. Overall, methods that use scRNA-seq data have comparable performance to the best performing bulk methods whereas semi-supervised approaches show higher error values. Moreover, failure to include cell types in the reference that are present in a mixture leads to substantially worse results, regardless of the previous choices. Altogether, we evaluate the combined impact of factors affecting the deconvolution task across different datasets and propose general guidelines to maximize its performance.

Inferring cell type proportions from transcriptomics data is affected by data transformation, normalization, choice of method and the markers used. Here, the authors use single-cell RNAseq datasets to evaluate the impact of these factors and propose guidelines to maximise deconvolution performance.

ITGA6+ Human Testicular Cell Populations Acquire a Mesenchymal Rather than Germ Cell Transcriptional Signature during Long-Term Culture

Autologous spermatogonial stem cell transplantation is an experimental technique aimed at restoring fertility in infertile men. Although effective in animal models, in vitro propagation of human spermatogonia prior to transplantation has proven to be difficult. A major limiting factor is endogenous somatic testicular cell overgrowth during long-term culture. This makes the culture both inefficient and necessitates highly specific cell sorting strategies in order to enrich cultured germ cell fractions prior to transplantation. Here, we employed RNA-Seq to determine cell type composition in sorted integrin alpha-6 (ITGA6+) primary human testicular cells (n = 4 donors) cultured for up to two months, using differential gene expression and cell deconvolution analyses. Our data and analyses reveal that long-term cultured ITGA6+ testicular cells are composed mainly of cells expressing markers of peritubular myoid cells, (progenitor) Leydig cells, fibroblasts and mesenchymal stromal cells and only a limited percentage of spermatogonial cells as compared to their uncultured counterparts. These findings provide valuable insights into the cell type composition of cultured human ITGA6+ testicular cells during in vitro propagation and may serve as a basis for optimizing future cell sorting strategies as well as optimizing the current human testicular cell culture system for clinical use.

Gene co-expression networks in peripheral blood capture dimensional measures of emotional and behavioral problems from the Child Behavior Checklist (CBCL)

The U.S. National Institute of Mental Health (NIMH) introduced the research domain criteria (RDoC) initiative to promote the integration of information across multiple units of analysis (i.e., brain circuits, physiology, behavior, self-reports) to better understand the basic dimensions of behavior and cognitive functioning underlying normal and abnormal mental conditions. Along those lines, this study examined the association between peripheral blood gene expression levels and emotional and behavioral problems in school-age children. Children were chosen from two age- and sex-matched groups: those with or without parental reports of any prior or current psychiatric diagnosis. RNA-sequencing was performed on whole blood from 96 probands aged 6-12 years who were medication-free at the time of assessment. Module eigengenes were derived using weighted gene co-expression network analysis (WGCNA). Associations were tested between module eigengene expression levels and eight syndrome scales from parent ratings on the Child Behavior Checklist (CBCL). Nine out of the 36 modules were significantly associated with at least one syndrome scale measured by the CBCL (i.e., aggression, social problems, attention problems, and/or thought problems) after accounting for covariates and correcting for multiple testing. Our study demonstrates that variation in peripheral blood gene expression relates to emotional and behavioral profiles in children. If replicated and validated, our results may help in identifying problem or at-risk behavior in pediatric populations, and in elucidating the biological pathways that modulate complex human behavior.

Dissecting differential signals in high-throughput data from complex tissues

MOTIVATION

Samples from clinical practices are often mixtures of different cell types. The high-throughput data obtained from these samples are thus mixed signals. The cell mixture brings complications to data analysis, and will lead to biased results if not properly accounted for.

RESULTS

We develop a method to model the high-throughput data from mixed, heterogeneous samples, and to detect differential signals. Our method allows flexible statistical inference for detecting a variety of cell-type specific changes. Extensive simulation studies and analyses of two real datasets demonstrate the favorable performance of our proposed method compared with existing ones serving similar purpose.

AVAILABILITY AND IMPLEMENTATION

The proposed method is implemented as an R package and is freely available on GitHub (https://github.com/ziyili20/TOAST).

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Compressing gene expression data using multiple latent space dimensionalities learns complementary biological representations

BACKGROUND

Unsupervised compression algorithms applied to gene expression data extract latent or hidden signals representing technical and biological sources of variation. However, these algorithms require a user to select a biologically appropriate latent space dimensionality. In practice, most researchers fit a single algorithm and latent dimensionality. We sought to determine the extent by which selecting only one fit limits the biological features captured in the latent representations and, consequently, limits what can be discovered with subsequent analyses.

RESULTS

We compress gene expression data from three large datasets consisting of adult normal tissue, adult cancer tissue, and pediatric cancer tissue. We train many different models across a large range of latent space dimensionalities and observe various performance differences. We identify more curated pathway gene sets significantly associated with individual dimensions in denoising autoencoder and variational autoencoder models trained using an intermediate number of latent dimensionalities. Combining compressed features across algorithms and dimensionalities captures the most pathway-associated representations. When trained with different latent dimensionalities, models learn strongly associated and generalizable biological representations including sex, neuroblastoma MYCN amplification, and cell types. Stronger signals, such as tumor type, are best captured in models trained at lower dimensionalities, while more subtle signals such as pathway activity are best identified in models trained with more latent dimensionalities.

CONCLUSIONS

There is no single best latent dimensionality or compression algorithm for analyzing gene expression data. Instead, using features derived from different compression models across multiple latent space dimensionalities enhances biological representations.

DeconPeaker, a Deconvolution Model to Identify Cell Types Based on Chromatin Accessibility in ATAC-Seq Data of Mixture Samples

While our understanding of cellular and molecular processes has grown exponentially, issues related to the cell microenvironment and cellular heterogeneity have sparked a new debate concerning the cell identity. Cell composition (chromatin and nuclear architecture) poses a strong risk for dynamic changes in the diseased condition. Since chromatin accessibility patterns play a major role in human diseases, it is therefore anticipated that a deconvolution tool based on open chromatin data will provide better performance in identifying cell composition. Herein, we have designed the deconvolution tool "DeconPeaker," which can precisely define the uniqueness among subpopulations of cells using open chromatin datasets. Using this tool, we simultaneously evaluated chromatin accessibility and gene expression datasets to estimate cell types and their respective proportions in a mixture of samples. In comparison to other known deconvolution methods, we observed the lowest average root-mean-square error (RMSE = 0.042) and the highest average correlation coefficient (r = 0.919) between the prediction and "true" proportion. As a proof-of-concept, we also tested chromatin accessibility data from acute myeloid leukemia (AML) and successfully obtained unique cell types associated with AML progression. Furthermore, we showed that chromatin accessibility represents more essential characteristics in the identification of cell types than gene expression. Taken together, DeconPeaker as a powerful tool has the potential to combine different datasets (primarily, chromatin accessibility and gene expression) and define different cell types in mixtures. The Python package of DeconPeaker is now available at https://github.com/lihuamei/DeconPeaker.

Standard machine learning approaches outperform deep representation learning on phenotype prediction from transcriptomics data

BACKGROUND

The ability to confidently predict health outcomes from gene expression would catalyze a revolution in molecular diagnostics. Yet, the goal of developing actionable, robust, and reproducible predictive signatures of phenotypes such as clinical outcome has not been attained in almost any disease area. Here, we report a comprehensive analysis spanning prediction tasks from ulcerative colitis, atopic dermatitis, diabetes, to many cancer subtypes for a total of 24 binary and multiclass prediction problems and 26 survival analysis tasks. We systematically investigate the influence of gene subsets, normalization methods and prediction algorithms. Crucially, we also explore the novel use of deep representation learning methods on large transcriptomics compendia, such as GTEx and TCGA, to boost the performance of state-of-the-art methods. The resources and findings in this work should serve as both an up-to-date reference on attainable performance, and as a benchmarking resource for further research.

RESULTS

Approaches that combine large numbers of genes outperformed single gene methods consistently and with a significant margin, but neither unsupervised nor semi-supervised representation learning techniques yielded consistent improvements in out-of-sample performance across datasets. Our findings suggest that using l2-regularized regression methods applied to centered log-ratio transformed transcript abundances provide the best predictive analyses overall.

CONCLUSIONS

Transcriptomics-based phenotype prediction benefits from proper normalization techniques and state-of-the-art regularized regression approaches. In our view, breakthrough performance is likely contingent on factors which are independent of normalization and general modeling techniques; these factors might include reduction of systematic errors in sequencing data, incorporation of other data types such as single-cell sequencing and proteomics, and improved use of prior knowledge.

A reference profile-free deconvolution method to infer cancer cell-intrinsic subtypes and tumor-type-specific stromal profiles

BACKGROUND

Patient stratification based on molecular subtypes is an important strategy for cancer precision medicine. Deriving clinically informative cancer molecular subtypes from transcriptomic data generated on whole tumor tissue samples is a non-trivial task, especially given the various non-cancer cellular elements intertwined with cancer cells in the tumor microenvironment.

METHODS

We developed a computational deconvolution method, DeClust, that stratifies patients into subtypes based on cancer cell-intrinsic signals identified by distinguishing cancer-type-specific signals from non-cancer signals in bulk tumor transcriptomic data. DeClust differs from most existing methods by directly incorporating molecular subtyping of solid tumors into the deconvolution process and outputting molecular subtype-specific tumor reference profiles for the cohort rather than individual tumor profiles. In addition, DeClust does not require reference expression profiles or signature matrices as inputs and estimates cancer-type-specific microenvironment signals from bulk tumor transcriptomic data.

RESULTS

DeClust was evaluated on both simulated data and 13 solid tumor datasets from The Cancer Genome Atlas (TCGA). DeClust performed among the best, relative to existing methods, for estimation of cellular composition. Compared to molecular subtypes reported by TCGA or other similar approaches, the subtypes generated by DeClust had higher correlations with cancer-intrinsic genomic alterations (e.g., somatic mutations and copy number variations) and lower correlations with tumor purity. While DeClust-identified subtypes were not more significantly associated with survival in general, DeClust identified a poor prognosis subtype of clear cell renal cancer, papillary renal cancer, and lung adenocarcinoma, all of which were characterized by CDKN2A deletions. As a reference profile-free deconvolution method, the tumor-type-specific stromal profiles and cancer cell-intrinsic subtypes generated by DeClust were supported by single-cell RNA sequencing data.

CONCLUSIONS

DeClust is a useful tool for cancer cell-intrinsic molecular subtyping of solid tumors. DeClust subtypes, together with the tumor-type-specific stromal profiles generated by this pan-cancer study, may lead to mechanistic and clinical insights across multiple tumor types.

Surfactant Expression Defines an Inflamed Subtype of Lung Adenocarcinoma Brain Metastases that Correlates with Prolonged Survival

PURPOSE

To provide a better understanding of the interplay between the immune system and brain metastases to advance therapeutic options for this life-threatening disease.

EXPERIMENTAL DESIGN

Tumor-infiltrating lymphocytes (TIL) were quantified by semiautomated whole-slide analysis in brain metastases from 81 lung adenocarcinomas. Multi-color staining enabled phenotyping of TILs (CD3, CD8, and FOXP3) on a single-cell resolution. Molecular determinants of the extent of TILs in brain metastases were analyzed by transcriptomics in a subset of 63 patients. Findings in lung adenocarcinoma brain metastases were related to published multi-omic primary lung adenocarcinoma The Cancer Genome Atlas data (n = 230) and single-cell RNA-sequencing (scRNA-seq) data (n = 52,698).

RESULTS

TIL numbers within tumor islands was an independent prognostic marker in patients with lung adenocarcinoma brain metastases. Comparative transcriptomics revealed that expression of three surfactant metabolism-related genes (SFTPA1, SFTPB, and NAPSA) was closely associated with TIL numbers. Their expression was not only prognostic in brain metastasis but also in primary lung adenocarcinoma. Correlation with scRNA-seq data revealed that brain metastases with high expression of surfactant genes might originate from tumor cells resembling alveolar type 2 cells. Methylome-based estimation of immune cell fractions in primary lung adenocarcinoma confirmed a positive association between lymphocyte infiltration and surfactant expression. Tumors with a high surfactant expression displayed a transcriptomic profile of an inflammatory microenvironment.

CONCLUSIONS

The expression of surfactant metabolism-related genes (SFTPA1, SFTPB, and NAPSA) defines an inflamed subtype of lung adenocarcinoma brain metastases characterized by high abundance of TILs in close vicinity to tumor cells, a prolonged survival, and a tumor microenvironment which might be more accessible to immunotherapeutic approaches.

The TMEM106B FTLD-protective variant, rs1990621, is also associated with increased neuronal proportion

Apart from amyloid β deposition and tau neurofibrillary tangles, Alzheimer's disease (AD) is a neurodegenerative disorder characterized by neuronal loss and astrocytosis in the cerebral cortex. The goal of this study is to investigate genetic factors associated with the neuronal proportion in health and disease. To identify cell-autonomous genetic variants associated with neuronal proportion in cortical tissues, we inferred cellular population structure from bulk RNA-Seq derived from 1536 individuals. We identified the variant rs1990621 located in the TMEM106B gene region as significantly associated with neuronal proportion (p value = 6.40 × 10-07) and replicated this finding in an independent dataset (p value = 7.41 × 10-04) surpassing the genome-wide threshold in the meta-analysis (p value = 9.42 × 10-09). This variant is in high LD with the TMEM106B non-synonymous variant p.T185S (rs3173615; r2 = 0.98) which was previously identified as a protective variant for frontotemporal lobar degeneration (FTLD). We stratified the samples by disease status, and discovered that this variant modulates neuronal proportion not only in AD cases, but also several neurodegenerative diseases and in elderly cognitively healthy controls. Furthermore, we did not find a significant association in younger controls or schizophrenia patients, suggesting that this variant might increase neuronal survival or confer resilience to the neurodegenerative process. The single variant and gene-based analyses also identified an overall genetic association between neuronal proportion, AD and FTLD risk. These results suggest that common pathways are implicated in these neurodegenerative diseases, that implicate neuronal survival. In summary, we identified a protective variant in the TMEM106B gene that may have a neuronal protection effect against general aging, independent of disease status, which could help elucidate the relationship between aging and neuronal survival in the presence or absence of neurodegenerative disorders. Our findings suggest that TMEM106B could be a potential target for neuronal protection therapies to ameliorate cognitive and functional deficits.

Multiscale, multimodal analysis of tumor heterogeneity in IDH1 mutant vs wild-type diffuse gliomas

Glioma is recognized to be a highly heterogeneous CNS malignancy, whose diverse cellular composition and cellular interactions have not been well characterized. To gain new clinical- and biological-insights into the genetically-bifurcated IDH1 mutant (mt) vs wildtype (wt) forms of glioma, we integrated data from protein, genomic and MR imaging from 20 treatment-naïve glioma cases and 16 recurrent GBM cases. Multiplexed immunofluorescence (MxIF) was used to generate single cell data for 43 protein markers representing all cancer hallmarks, Genomic sequencing (exome and RNA (normal and tumor) and magnetic resonance imaging (MRI) quantitative features (protocols were T1-post, FLAIR and ADC) from whole tumor, peritumoral edema and enhancing core vs equivalent normal region were also collected from patients. Based on MxIF analysis, 85,767 cells (glioma cases) and 56,304 cells (GBM cases) were used to generate cell-level data for 24 biomarkers. K-means clustering was used to generate 7 distinct groups of cells with divergent biomarker profiles and deconvolution was used to assign RNA data into three classes. Spatial and molecular heterogeneity metrics were generated for the cell data. All features were compared between IDH mt and IDHwt patients and were finally combined to provide a holistic/integrated comparison. Protein expression by hallmark was generally lower in the IDHmt vs wt patients. Molecular and spatial heterogeneity scores for angiogenesis and cell invasion also differed between IDHmt and wt gliomas irrespective of prior treatment and tumor grade; these differences also persisted in the MR imaging features of peritumoral edema and contrast enhancement volumes. A coherent picture of enhanced angiogenesis in IDHwt tumors was derived from multiple platforms (genomic, proteomic and imaging) and scales from individual proteins to cell clusters and heterogeneity, as well as bulk tumor RNA and imaging features. Longer overall survival for IDH1mt glioma patients may reflect mutation-driven alterations in cellular, molecular, and spatial heterogeneity which manifest in discernable radiological manifestations.

Mega-Analysis of Gene Expression in Mouse Models of Alzheimer's Disease

While multiple studies have been conducted of gene expression in mouse models of Alzheimer's disease (AD), their findings have not reached a clear consensus and have not accounted for the potentially confounding effects of changes in cellular composition. To help address this gap, we conducted a re-analysis based meta-analysis (mega-analysis) of ten independent studies of hippocampal gene expression in mouse models of AD. We used estimates of cellular composition as covariates in statistical models aimed to identify genes differentially expressed (DE) at either early or late stages of progression. Our analysis revealed changes in gene expression at early phases shared across studies, including dysregulation of genes involved in cholesterol biosynthesis and the complement system. Expression changes at later stages were dominated by cellular compositional effects. Thus, despite the considerable heterogeneity of the mouse models, we identified common patterns that may contribute to our understanding of AD etiology. Our work also highlights the importance of controlling for cellular composition effects in genomics studies of neurodegeneration.

ADAPTS: Automated deconvolution augmentation of profiles for tissue specific cells

Immune cell infiltration of tumors and the tumor microenvironment can be an important component for determining patient outcomes. For example, immune and stromal cell presence inferred by deconvolving patient gene expression data may help identify high risk patients or suggest a course of treatment. One particularly powerful family of deconvolution techniques uses signature matrices of genes that uniquely identify each cell type as determined from single cell type purified gene expression data. Many methods from this family have been recently published, often including new signature matrices appropriate for a single purpose, such as investigating a specific type of tumor. The package ADAPTS helps users make the most of this expanding knowledge base by introducing a framework for cell type deconvolution. ADAPTS implements modular tools for customizing signature matrices for new tissue types by adding custom cell types or building new matrices de novo, including from single cell RNAseq data. It includes a common interface to several popular deconvolution algorithms that use a signature matrix to estimate the proportion of cell types present in heterogenous samples. ADAPTS also implements a novel method for clustering cell types into groups that are difficult to distinguish by deconvolution and then re-splitting those clusters using hierarchical deconvolution. We demonstrate that the techniques implemented in ADAPTS improve the ability to reconstruct the cell types present in a single cell RNAseq data set in a blind predictive analysis. ADAPTS is currently available for use in R on CRAN and GitHub.

Psychiatric Genetics, Epigenetics, and Cellular Models in Coming Years

Psychiatric genetic studies have uncovered hundreds of loci associated with various psychiatric disorders. We take the opportunity to review achievements in the past and provide our view of what is coming in the fields of molecular genetics, epigenetics, and cellular models. We expect that SNP-array and sequencing-based studies of genetic associations will continue to expand, covering more disorders, drug responses, phenotypes, and diverse populations. Epigenetic studies of psychiatric disorders will be another promising field with the growing recognition that environmental factors impact the risk for psychiatric disorders by modulating epigenetic factors. Functional studies of genetic findings will be needed in cellular models to provide important connections between genetic and epigenetic variants and biological phenotypes.