Search

Summary: DNA changes that cause premature termination codons (PTCs) represent a large fraction of clinically relevant pathogenic genomic variation. Typically, PTCs induce a transcript's degradation by nonsense-mediated mRNA decay (NMD) and render such changes loss-of-function alleles. However, certain PTC-containing transcripts escape NMD and can exert dominant-negative or gain-of-function (DN/GOF) effects. Therefore, systematic identification of human PTC-causing variants and their susceptibility to NMD contributes to the investigation of the role of DN/GOF alleles in human disease. Here we present aenmd, a software for annotating PTC-containing transcript-variant pairs for predicted escape from NMD. aenmd is user-friendly and self-contained. It offers functionality not currently available in other methods and is based on established and experimentally validated rules for NMD escape; the software is designed to work at scale, and to integrate seamlessly with existing analysis workflows. We applied aenmd to variants in the gnomAD, Clinvar, and GWAS catalog databases and report the prevalence of human PTC-causing variants in these databases, and the subset of these that could exert DN/GOF effects via NMD escape. Availability and implementation: aenmd is implemented in the R programming language. Code is available on GitHub as an R package (github.com/kostkalab/aenmd.git), and as a containerized command-line interface (github.com/kostkalab/aenmd_cli.git).

Motivation Single-cell RNA sequencing (scRNA-seq) continues to expand our knowledge by facilitating the study of transcriptional heterogeneity at the level of single cells. Despite this technology’s utility and success in biomedical research, technical artifacts are present in scRNA-seq data. Doublets/multiplets are a type of artifact that occurs when two or more cells are tagged by the same barcode, and therefore they appear as a single cell. Because this introduces non-existent transcriptional profiles, doublets can bias and mislead downstream analysis. To address this limitation, computational methods to annotate and remove doublets form scRNA-seq datasets are needed. Results We introduce vaeda (Variational Auto-Encoder for Doublet Annotation), a new approach for computational annotation of doublets in scRNA-seq data. Vaeda integrates a variational auto-encoder and Positive-Unlabeled learning to produce doublet scores and binary doublet calls. We apply vaeda, along with seven existing doublet annotation methods, to 16 benchmark datasets and find that vaeda performs competitively in terms of doublet scores and doublet calls. Notably, vaeda outperforms other python-based methods for doublet annotation. Altogether, vaeda is a robust and competitive method for scRNA-seq doublet annotation and may be of particular interest in the context of python-based workflows. Availability and implementation Vaeda is available at https://github.com/kostkalab/vaeda, and the version used for the results we present here is archived at zenodo (https://doi.org/10.5281/zenodo.7199783). Supplementary information Supplementary data are available at Bioinformatics online.

Motivation Single cell RNA sequencing (scRNA-seq) enables transcriptional profiling at the level of individual cells. With the emergence of high-throughput platforms datasets comprising tens of thousands or more cells have become routine, and the technology is having an impact across a wide range of biomedical subject areas. However, scRNA-seq data are high-dimensional and affected by noise, so that scalable and robust computational techniques are needed for meaningful analysis, visualization and interpretation. Specifically, a range of matrix factorization techniques have been employed to aid scRNA-seq data analysis. In this context we note that sources contributing to biological variability between cells can be discrete (or multi-modal, for instance cell-types), or continuous (e.g. pathway activity). However, no current matrix factorization approach is set up to jointly infer such mixed sources of variability. Results To address this shortcoming, we present a new probabilistic single-cell factor analysis model, N on-negative I ndependent F actor A nalysis (NIFA), that combines features of complementary approaches like Independent Component Analysis (ICA), Principal Component Analysis (PCA), and Non-negative Matrix Factorization (NMF). NIFA simultaneously models uni- and multi-modal latent factors and can so isolate discrete cell-type identity and continuous pathway-level variations into separate components. Similar to NMF, NIFA constrains factor loadings to be non-negative in order to increase biological interpretability. We apply our approach to a range of data sets where cell-type identity is known, and we show that NIFA-derived factors outperform results from ICA, PCA and NMF in terms of cell-type identification and biological interpretability. Studying an immunotherapy dataset in detail, we show that NIFA identifies biomedically meaningful sources of variation, derive an improved expression signature for regulatory T-cells, and identify a novel myeloid cell subtype associated with treatment response. Overall, NIFA is a general approach advancing scRNA-seq analysis capabilities and it allows researchers to better take advantage of their data. NIFA is available at https://github.com/wgmao/NIFA . Contact mchikina@pitt.edu

Figure S1. Enhancer sets across all contexts considered differ in both number (A) and length (B): Gm12878; Heart. Figure S2. Enhancer sets have low amounts of overlap with each other in bp-wise comparisons. Figure S3. Enhancer sets overlap more than expected by chance in element-wise comparisons. Figure S4. Enhancer sets have low amounts of overlap with each other in element-wise comparisons. Figure S5. Enhancers identified by different methods differ in enrichment for base pair overlap with functional attributes. Figure S6. Enhancer identification strategies recognize different subsets of validated enhancers. Figure S7. K562 enhancer sets have similar low levels of enrichment for activating regions validated by Sharpr-MPRA. Figure S8. Even among the variants in each functional LD block (r2 > 0.9) with the most enhancer set overlap, there is substantial disagreement between enhancer identification methods. Figure S9. Pairwise similarity for GO Molecular Function (MF) enrichments for enhancer sets based on JEME’s putative mappings to target genes in K562 (A), Gm12878 (B), liver (C), and heart (D). Figure S10. Pairwise similarity for GO Biological Process (BP) for enhancer sets based on JEME’s putative mappings to target genes in K562 (A), Gm12878 (B), liver (C), and heart (D). Figure S11. Pairwise similarity for GO Molecular Function (MF) enrichments from GREAT for liver enhancer sets. Figure S12. There is low pairwise similarity between GO Molecular Function (MF) enrichments calculated with GREAT for enhancer sets in the same context. Figure S13. There is low pairwise similarity between GO Biological Process (BP) enrichments calculated with GREAT for enhancer sets in the same context. Figure S14. Clustering enhancer sets on similarity of enriched transcription factor binding motifs illustrates different clustering of methods. Figure S15. Regions identified as enhancers by multiple methods do not have higher confidence scores than regions identified by a single method. Figure S16. Score distributions for K562 enhancer sets are similar between regions identified as enhancers by a single method and those identified by multiple methods: (A) H3K27acPlusH3K4me1, (B) H3K27acMinusH3K4me3, (C) DNasePlusHistone, (D) EncodeEnhancerlike, and (E) FANTOM. Figure S17. Score distributions for Gm12878 enhancer sets are similar between regions identified as enhancers by a single method and those identified by multiple methods: (A) H3K27acPlusH3K4me1, (B) H3K27acMinusH3K4me3, (C) DNasePlusHistone, (D) EncodeEnhancerlike, and (E) FANTOM. Figure S18. Score distributions for heart enhancer sets are similar between regions identified as enhancers by a single method and those identified by multiple methods: (A) H3K27acPlusH3K4me1, (B) H3K27acMinusH3K4me3, (C) DNasePlusHistone, (D) EncodeEnhancerlike, and (E) FANTOM. Figure S19. Enrichment for functional attributes is not significantly different between regions identified as enhancers by a single method and those identified by multiple methods when focusing on the top 100 predictions from each method. Figure S20. Same as Fig. S19, but considering the top 500 predictions from each method. Table S1. The average distance (in bp) to the closest TSS over all enhancers identified by each method in each cellular context. Table S2. Summary statistics for pairwise percent overlap, both in a base pair and element-wise comparison. Table S3. Number of observed VISTA heart positive and VISTA negative overlaps for each context and enhancer identification method. Table S4. Curated list of relevant GWAS phenotypes for liver (n = 50) and heart (n = 169). Table S5. Enrichments for overlap with context-specific SNPs in liver and heart. Table S6. Number of overlapping GWAS SNPs per enhancer identification method and context. Table S7. Enrichments for overlap with context-specific eQTL in liver and heart. Table S8. Number of overlapping GTEx eQTL per enhancer identification method and context. Table S9. Number of target genes mapped to each enhancer set by JEME. For K562 and Gm12878, p300 and GRO-cap are not included in this mapping. Table S11. Number of enhancers removed by length filtering. (DOCX 36740 kb)

Contents, Acknowledgements and Introduction i 1.Finding Molecular Characteristics of Disease 1 * Motivation * Supervised Classification of Patients * Discussion and Summary 2\. Communicating Molecular Characteristics of Disease 19 * Motivation * Preprocessing of Oligonucleotide Microarrays * Documentation of Signatures * Application to Data * Compatibility of External Patients to Core Data * Discussion and Summary 3\. Exploring Molecular Characteristics of Disease 45 * Motivation * The dcoex Algorithm * Application to Data * Discussion and Summary Summary and Bibliography 67 Appendices 85, Microarray data data characterizes cells on the transcriptional level. Prominent applications of microarray technology in a clinical setting are the molecular diagnosis of patients and the discovery of disease subtypes by patient stratification (clustering). Lists of differentially expressed genes are often used to guide biological intuition. In general,the data can be utilized to infer novel biological hypotheses by means of pattern mining and to refine or confirm existing knowledge. This thesis contains methodological contributions to both settings. It is composed of three chapters. The first chapter describes statistical learning techniques, which are frequently applied to microarray data with the goal of obtaining rules for molecular diagnosis. The focus lies on characteristics arising from the specific nature of high dimensional microarray data. This chapter concisely integrates concepts, algorithms and practical aspects of microarray data analysis that are usually found in distinct fields of the literature. It provides the theoretical foundation of the other chapters. The second chapter is concerned with the unambiguous documentation of a diagnostic molecular signature or, equivalently, with the unequivocal characterization of disease or subtype of disease. Themotivation to address documentation and communication of molecular signatures is a practical one: Microarray based gene expression signatures have the potential to be powerful tools for patient stratification and diagnosis of disease. But before they can affect clinical practice they need to be communicated to other health care centers with data for independent validation. External validation of a signature can only be meaningful if the new data is transformed to a scale compatible with the original one the signature is tuned to. This scale, in turn, depends on the initial preprocessing applied in the signature deriving study. It needs to be communicated alongside with the signature. Chapter two formalizes this requirement and contains scale adjusting transformations for two popular preprocessing schemes. Using eight clinical microarray data sets I am able to show significantly increased consistency and stability of molecular diagnoses as compared to standard documentation procedures. This underlines the key point of the chapter: Data preprocessing has to be taken into account when documenting molecular characteristics of disease. The third chapter introduces the dcoex algorithm, a method designed to utilize microarray data to reveal groups of genes losing coregulation between two phenotypes. Information about differentially coregulated genes can not only provide a molecular characterization of the phenotypes; it also provides focused information which is useful to generate hypotheses about biological mechanisms underlying the phenotypical differentiation. This chapter introduces the concept, implements an algorithm for detection and demonstrates the biological plausibility of differentially coexpressed genes. In a data set on childhood leukemia we find a biologically plausible group of genes differentially coexpressed between cytogenetically normal children and children bearing a Philadelphia chromosome. After assessing robustness and statistical significance of our findings we conclude that dcoex constitutes a new analysis tool enabling the exploration of differential coexpression patterns., Diese Arbeit dreht sich um die Charakterisierung von Krankheiten mit Hilfe von Genexpressionsdaten. Solche Daten stellen Zellen auf molekularer Ebene dar und können zur Beschreibung von Krankheiten auf zweierlei Art verwendet werden: Zum einen kann man bekannte Krankheiten genauer und verläßlicher diagnostizieren. Zum anderen kann man versuchen, in stetig wiederkehrenden Expressionsmustern entweder neue Krankheitsentitäten zu entdecken, oder aber aufgrund solcher Muster auf biologisch-medizinische Ursachenbekannter Krankheiten zu schließen. Die vorliegende Arbeit enthält methodologisch neue Ansätze für beide Szenarien. Nach einer Einleitung, die unter anderem die Microarray-Technik kurz skizziert, folgt ein ein weiteres einführendes Kapitel. Darin werden Methoden der statistischen Lerntheorie beschrieben, die man benutzen kann um aus Beispieldaten Schemata (oder molekulare Signaturen) für eine Diagnose abzuleiten. Die Darstellung ist auf die Anwendung statistischer Verfahren auf Microarray Daten zugeschnitten und das Kapitel bildet die theoretische Grundlage der folgenden Arbeit. Thema des zweiten Kapitels ist die unzweideutige Dokumentation einmal hergeleiteter molekularer Signaturen. Die Dokumentation einer Expressionssignatur ist ein notwendiger Schritt, falls diese zwischen Wissenschaftlern und Forschungseinrichtungen ausgetauscht werden soll. Ein solcher Austausch aber muss Tests und Validierungen einer Signatur vorangehen, die ihrerseits für den klinischen Einsatz unerläßlich sind. Wir stellen zwei Methoden vor, die gebräuchliche Strategien der Datenvorverarbeitung ergänzen und demonstrieren eine signifikante Erhöhung der Stimmigkeit von Diagnosen an verschiedenen Datensätzen. Im dritten Kapitel wird das Konzept der differentiellen Ko-Expression und der dazugehörige dcoex Algorithmus vorgestellt. Eine Gruppe differentiell koexprimierter Gene hat die Eigenschaft in Proben eines bestimmten Phänotyps kohärent exprimiert zusein, verliert diese Kohärenz allerdings in den Proben eines anderen Phänotyps. Der dcoex Algorithmus ist eine Methode solche Gruppen von differentiell koexprimierten Genen in Datensätzen zu finden, wobei ein kombinatorisches Optimierungsproblem heuristisch gelöst wird. Gruppen differentiell koexprimierter Gene können nicht nur zur molekularen Charakterisierung unterschiedlicher Phänotypen beitragen. Aus den Gengruppen abgeleitete Informationen kann man zur Formulierung fokussierter biologischer Hypothesen verwenden. Wir demonstrieren dies an einem Leukämiedatensatz.