Search

Advances in next generation sequencing (NGS) technologies resulted in a broad array of large-scale gene expression studies and an unprecedented volume of whole messenger RNA (mRNA) sequencing data, or the transcriptome (also known as RNA sequencing, or RNA-seq). These include the Genotype Tissue Expression project (GTEx) and The Cancer Genome Atlas (TCGA), among others. Here we cover some of the commonly used datasets, provide an overview on how to begin the analysis pipeline, and how to explore and interpret the data provided by these publicly available resources.

Noninvasive prenatal diagnosis (NIPD) has become a common, safe, and effective procedure for detection of inherited diseases early in pregnancy. It is based on the analysis of fetal cell-free DNA (cffDNA) derived from the placenta, circulating in the maternal plasma. De novo mutations, although rare, cause a considerable number of dominant genetic disorders. Due to the sparse representation of fetal-derived sequences in the blood, the challenge of detecting low frequency fetal de novo mutations becomes preponderant. Hence, this detection type requires deep genome-wide sequencing of cffDNA from maternal plasma and a unique analysis approach. Here we suggest and discuss a method for identifying de novo mutations based on whole genome sequencing (WGS) of cell-free DNA (cfDNA) from maternal plasma samples. Our method consists of an augmented pipeline for analysis of de novo mutation candidates. It begins with an enhanced noninvasive fetal variant calling step, followed by a candidate de novo mutation filtration, and then finally, a supervised machine learning approach is utilized for reduction of false positive rates. Overall, this study provides a basis for genome-wide de novo mutation analysis in NIPD procedures, which could be used in any procedure where rare de novo mutations should be carefully picked out of a sea of data.

Noninvasive prenatal diagnosis (NIPD) is an emerging field, that enables testing for diseases in the fetus with no risk to the pregnancy, compared to invasive methods (e.g., amniocentesis). The procedure is based on the presence of fetal DNA within the mother's plasma cell-free DNA (cfDNA). Today, NIPD is performed for chromosomal abnormalities (e.g., Down syndrome) and some large deletions/duplications. It is also available for point mutations but is limited for one mutation or up to several genes simultaneously. Genome-wide detection of fetal point mutations was presented in a few studies, and the first software tool for this task, Hoobari, has recently become available. Here we describe the necessary steps in genome-wide noninvasive fetal genotyping, including examples using the Hoobari software. We discuss the various materials, software, computational infrastructure, and samples required for this analysis. Genome-wide analysis of point mutations in the fetus is not widely studied, albeit much space for algorithmic improvements exists. Here we suggest practical solutions for challenges along the process. Our work assists bioinformaticians in accessing NIPD data analysis and can eventually be utilized for other cfDNA-related fields.

MicroRNAs (miRNAs) regulate gene expression by binding to mRNAs. Consequently, they reduce target gene expression levels and expression variability, also known as "noise." Single-cell RNA sequencing (scRNA-seq) technology has been used to study miRNA and mRNA expression in single cells, and has demonstrated its strength in quantifying cell-to-cell variation. Here we describe how to investigate miRNA regulation using data with both mRNA and miRNA expression in single cell format. We show that miRNAs reduce the expression levels and also expression noise of target genes in single cells. Finally, we also discuss potential improvements in experimental design and computational analysis of scRNA-seq in order to reduce or partition the technical noise.

Since its inception, deep learning has revolutionized the field of machine learning and data-driven science. One such data-driven science to be transformed by deep learning is genomics. In the past decade, numerous genomics studies have adopted deep learning and its applications range from predicting regulatory elements to cancer classification. Despite its dominating efficacy in these applications, deep learning is not without drawbacks. A prominent shortcoming of deep learning is the lack of interpretability. Hence, the main objective of this study is to address this obstacle in the deep learning cancer classification. Here we adopt a feature importance scoring methodology (Gradient-based class activation mapping or Grad-CAM) on a quasi-recurrent neural network model that classify cancer based on FASTA sequencing data. In this study, we managed to formulate a nucleotide-to-genomic-region Grad-CAM scoring methodology, as well as, validate the use this methodology for the chosen model. Consequently, this allows for the utilization of the Grad-CAM scoring methodology for feature importance in deep learning cancer classification. The results from our study identify potential novel candidate genes, genomic elements, and mechanisms for future cancer research.

Recent scientific advancements in the field of genetics have fostered significant changes for the criminal justice system. Growing National DNA databases, public DNA databases, private direct-to-consumer (DTC) DNA testing companies, and improvements in next-generation sequencing (NGS) have resulted in effective methods for tracking down criminals and exonerating the innocent. While these recently discovered and profound techniques seem to provide benefits, their use in forensic detection has become subject to harsh legal opposition. Ultimately, should law enforcement be permitted to analyze DNA found at crime scenes and DNA that has accumulated in national, public, and private databases to aid in their investigations, or are individuals' privacy rights breached in the process?

Over the last decade, single cell RNA sequencing (scRNAseq) became an increasingly viable solution for analyzing cellular heterogeneity and cell-specific expression differences. While not as mature or fully realized as bulk sequencing, newly developed computational methods offer a solution to the challenges of scRNAseq data analysis, providing previously inaccessible biological insight at unprecedented levels of detail. Here, we go over the inherent challenges of single-cell data analysis and the computational methods used to overcome them. We cover current and future applications of scRNAseq in research of cellular dynamics and as an integrative component of biological research.

Deep learning is defined as the group of computational techniques allowing for the discovery of latent information within large amounts of data. Recently, many fields have seen the immense potential of deep learning to solve various tasks in ways which outperformed many other traditional methods. Genomic research could be the next frontier to take advantage of deep learning, as it has the perfect combination of vast amounts of data and diverse tasks. Here we present the platform we generated to combine deep learning and genomic sequencing data. We tested the platform on publicly available sequencing data from the gut microbiome of cancer patients. We showed that our platform is capable of classifying patients with higher accuracy than other methods, with some caveats. Overall, we believe genomic research is the next frontline for deep learning as there are exciting avenues waiting to be explored. We think that our platform, presented here, could serve as the basis for such future research.

Whole exome sequencing presents a powerful tool to study rare genetic disorders. The most challenging part of using exome sequencing for the purpose of disease-causing variant detection is analyzing, interpreting, and filtering the large number of detected variants. In this chapter we provide a comprehensive description of the various steps required for such an analysis. We address strategies in selecting samples to sequence, and technical considerations involved in exome sequencing. We then discuss how to identify variants, and methods for first annotating detected variants using characteristics such as allele frequency, location in the genome, and predicted severity, and then classifying and prioritizing the detected variants based on those annotations. Finally, we review possible gene annotations that may help to establish a relationship between genes carrying high-priority variants and the phenotype in question, in order to identify the most likely causative mutations.

Our ability to sequence the genomic data at our disposal is limited. At each experiment we can sequence reliably only a short fraction of even the smallest genome. We are then faced with the challenge of assembly-combining the short patches we have into a correct reconstruction of as large as possible a fragment of the original sample. The problem has been thoroughly researched and many commercial and academic tools exist to carry it out. However due to basic features of the problem the results of even our best efforts will be sometimes disappointing for the researcher. In this chapter we will try to explain why the assembly problem is so hard, what future directions may alleviate it in the near future, and what can be realistically expected from a current assembly experiment.