Interpreting protein expression in multiplexed tissue imaging data presents a significant challenge due to the high dimensionality of the resulting images, the variety of intracellular structures, cell shapes resulting from 2-D tissue sectioning, and the presence of technological noise and imaging artifacts. Here, we introduce the Information-Controlled Variational Autoencoder (IC-VAE), a deep generative model designed to tackle this challenge. The contribution of IC-VAE to the VAE framework is the ability to control the shared information among latent subspaces. We use IC-VAE to factorize each cell's image into its true protein expression, various cellular components, and background noise, while controlling the shared information among some of these components. Compared with other normalization methods, this approach leads to superior results in downstream analysis, such as analyzing the expression of biomarkers, classification for cell types, or visualizing cell clusters using t-SNE/UMAP techniques.

Most widely used tools for finding marker genes in single cell data (SeuratT/NegBinom/Poisson, CellRanger, EdgeR, limmatrend) use a conventional definition of differentially expressed genes: genes with different mean expression values. However, in single-cell data, a cell population can be a mixture of many cell types/cell states, hence the mean expression of genes cannot represent the whole population. In addition, these tools assume that gene expression of a population belongs to a specific family of distribution. This assumption is often violated in single-cell data. In this work, we define marker genes of a cell population as genes that can be used to distinguish cells in the population from cells in other populations. Besides log-fold change, we devise a new metric to classify genes into up-regulated, down-regulated, and transitional states. In a benchmark for finding up-regulated and down-regulated genes, our tool outperforms all compared methods, including Seurat, ROTS, scDD, edgeR, MAST, limma, normal t-test, Wilcoxon and Kolmogorov–Smirnov test. Our method is much faster than all compared methods, therefore, enables interactive analysis for large single-cell data sets in BioTuring Browser. Venice algorithm is available within Signac package: https://github.com/bioturing/signac).

Annotating cell types using single-cell transcriptome data usually requires binarizing the expression data to distinguish between the background noise vs. real expression or low expression vs. high expression cases. A common approach is choosing a “reasonable” cutoff value, but it remains unclear how to choose it. In this work, we describe a simple yet effective approach for finding this threshold value.
A common procedure to annotate cell types in a single-cell RNA-seq study is to first perform graph-based clustering, and further check the expression of some marker genes in each cluster. In some cases, scientists need to distinguish between the real expression of a gene vs the background expression. In other cases, they need to know if a gene expresses highly in one cluster and lowly in another cluster (e.g., NK CD56bright and NK CD56dim). This requires choosing a threshold to binarize the expression data. It remains unclear how to choose this threshold. Here, we propose to binarize the data in a way that minimizes the clustering information loss. Below, we describe the formulation in detail.

An important but rarely discussed phenomenon in single cell data generated by the 10X-Chromium protocol is that the fraction of non-exonic reads is very high. This number usually exceeds 30% of the total reads. Without aligning them to a complete genome reference, non-exonic reads can be erroneously aligned to the transcriptome reference with higher error rates. To tackle this problem, Cell Ranger chooses to firstly align reads against the whole genome, and at a later step, uses a genome annotation to select reads that align to the transcriptome. Despite its high running time and large memory consumption, Cell Ranger remains the most widely used tool to quantify 10XGenomics single cell RNA-Seq data for its accuracy.
In this work, we introduce Hera-T, a fast and accurate tool for estimating gene abundances in single cell data generated by the 10X-Chromium protocol. By devising a new strategy for aligning reads to both transcriptome and genome references, Hera-T reduces both running time and memory consumption from 10 to 100 folds while giving similar results compared to Cell Ranger’s. Hera-T also addresses some difficult splicing alignment scenarios that Cell Ranger fails to address, and therefore, obtains better accuracy compared to Cell Ranger. Excluding the reads in those scenarios, Hera-T and Cell Ranger results have correlation scores > 0.99.
For a single-cell data set with 49 million of reads, Cell Ranger took 3 hours (179 minutes) while Hera-T took 1.75 minutes; for another single-cell data set with 784 millions of reads, Cell Ranger took about 25 hours while Hera-T took 32 minutes. For those data sets, Cell Ranger completely used all 32 GB of memory while Hera-T consumed at most 8 GB. Hera-T package is available for download at: https://bioturing.com/product/hera-t

RSEM has been mainly known for its accuracy in transcript abundance quantification. However, its quantification time is extremely high compared to that of recent quantification tools. In this paper, we revised the RSEM’s EM algorithm. In particular, we derived accurate M-step updates to eliminate incorrect heuristic updates in RSEM. We also implement some optimizations that reduce the quantification time about a hundred times while still have better accuracy compared to RSEM. In particular, we noticed that different parameters have different convergence rates, therefore we identified and removed early converged parameters to significantly reduce the model complexity in further iterations, and we also use SQUAREM method to further speed up the convergence rate. We implemented these revisions in a packaged named Hera-EM, with source code available at: https://github.com/bioturing/hera/tree/master/hera-EM

The lion's share of bacteria in various environments cannot be cloned in the laboratory and thus cannot be sequenced using existing technologies. A major goal of single-cell genomics is to complement gene-centric metagenomic data with whole-genome assemblies of uncultivated organisms. Assembly of single-cell data is challenging because of highly non-uniform read coverage as well as elevated levels of sequencing errors and chimeric reads. We describe SPAdes, a new assembler for both single-cell and standard (multicell) assembly, and demonstrate that it improves on the recently released E+V−SC assembler (specialized for single-cell data) and on popular assemblers Velvet and SoapDeNovo (for multicell data). SPAdes generates single-cell assemblies, providing information about genomes of uncultivatable bacteria that vastly exceeds what may be obtained via traditional metagenomics studies. SPAdes is available online (http://bioinf.spbau.ru/spades). It is distributed as open source software.

The lion's share of bacteria in various environments cannot be cloned in the laboratory and thus cannot be sequenced using existing technologies. A major goal of single-cell genomics is to complement gene-centric metagenomic data with whole-genome assemblies of uncultivated organisms. Assembly of single-cell data is challenging because of highly non-uniform read coverage as well as elevated levels of sequencing errors and chimeric reads. We describe SPAdes, a new assembler for both single-cell and standard (multicell) assembly, and demonstrate that it improves on the recently released E+V−SC assembler (specialized for single-cell data) and on popular assemblers Velvet and SoapDeNovo (for multicell data). SPAdes generates single-cell assemblies, providing information about genomes of uncultivatable bacteria that vastly exceeds what may be obtained via traditional metagenomics studies. SPAdes is available online (http://bioinf.spbau.ru/spades). It is distributed as open source software.

A protocol is described for sequencing the transcriptome of a cell nucleus. Nuclei are isolated from specimens and sorted by FACS, cDNA libraries are constructed and RNA-seq is performed, followed by data analysis. Some steps follow published methods (Smart-seq2 for cDNA synthesis and Nextera XT barcoded library preparation) and are not described in detail here. Previous single-cell approaches for RNA-seq from tissues include cell dissociation using protease treatment at 30 °C, which is known to alter the transcriptome. We isolate nuclei at 4 °C from tissue homogenates, which cause minimal damage. Nuclear transcriptomes can be obtained from postmortem human brain tissue stored at -80 °C, making brain archives accessible for RNA-seq from individual neurons. The method also allows investigation of biological features unique to nuclei, such as enrichment of certain transcripts and precursors of some noncoding RNAs. By following this procedure, it takes about 4 d to construct cDNA libraries that are ready for sequencing.

Bacterial genomes are simpler than mammalian ones, and yet assembling the former from the data currently generated by high-throughput short-read sequencing machines still results in hundreds of contigs. To improve assembly quality, recent studies have utilized longer Pacific Biosciences (PacBio) reads or jumping libraries to connect contigs into larger scaffolds or help assemblers resolve ambiguities in repetitive regions of the genome. However, their popularity in contemporary genomic research is still limited by high cost and error rates. In this work, we explore the possibility of improving assemblies by using complete genomes from closely related species/strains. We present Ragout, a genome rearrangement approach, to address this problem. In contrast with most reference-guided algorithms, where only one reference genome is used, Ragout uses multiple references along with the evolutionary relationship among these references in order to determine the correct order of the contigs. Additionally, Ragout uses the assembly graph and multi-scale synteny blocks to reduce assembly gaps caused by small contigs from the input assembly. In simulations as well as real datasets, we believe that for common bacterial species, where many complete genome sequences from related strains have been available, the current high-throughput short-read sequencing paradigm is sufficient to obtain a single high-quality scaffold for each chromosome.

We report full-length draft de novo genome assemblies for 16 widely used inbred mouse strains and find extensive strain-specific haplotype variation. We identify and characterize 2,567 regions on the current mouse reference genome exhibiting the greatest sequence diversity. These regions are enriched for genes involved in pathogen defence and immunity and exhibit enrichment of transposable elements and signatures of recent retrotransposition events. Combinations of alleles and genes unique to an individual strain are commonly observed at these loci, reflecting distinct strain phenotypes. We used these genomes to improve the mouse reference genome, resulting in the completion of 10 new gene structures. Also, 62 new coding loci were added to the reference genome annotation. These genomes identified a large, previously unannotated, gene (Efcab3-like) encoding 5,874 amino acids. Mutant Efcab3-like mice display anomalies in multiple brain regions, suggesting a possible role for this gene in the regulation of brain development.

Comparing strains within the same microbial species has proven effective in the identification of genes and genomic regions responsible for virulence, as well as in the diagnosis and treatment of infectious diseases. In this paper, we present Sibelia, a tool for finding synteny blocks in multiple closely related microbial genomes using iterative de Bruijn graphs. Unlike most other tools, Sibelia can find synteny blocks that are repeated within genomes as well as blocks shared by multiple genomes. It represents synteny blocks in a hierarchy structure with multiple layers, each of which representing a different granularity level. Sibelia has been designed to work efficiently with a large number of microbial genomes; it finds synteny blocks in 31 S. aureus genomes within 31 minutes and in 59 E.coli genomes within 107 minutes on a standard desktop. Sibelia software is distributed under the GNU GPL v2 license and is available at: https://github.com/bioinf/Sibelia. Sibelia’s web-server is available at: http://etool.me/software/sibelia.

The recent proliferation of next generation sequencing with short reads has enabled many new experimental opportunities but, at the same time, has raised formidable computational challenges in genome assembly. One of the key advances that has led to an improvement in contig lengths has been mate pairs, which facilitate the assembly of repeating regions. Mate pairs have been algorithmically incorporated into most next generation assemblers as various heuristic post-processing steps to correct the assembly graph or to link contigs into scaffolds. Such methods have allowed the identification of longer contigs than would be possible with single reads; however, they can still fail to resolve complex repeats. Thus, improved methods for incorporating mate pairs will have a strong effect on contig length in the future. Here, we introduce the paired de Bruijn graph, a generalization of the de Bruijn graph that incorporates mate pair information into the graph structure itself instead of analyzing mate pairs at a post-processing step. This graph has the potential to be used in place of the de Bruijn graph in any de Bruijn graph based assembler, maintaining all other assembly steps such as error-correction and repeat resolution. Through assembly results on simulated perfect data, we argue that this can effectively improve the contig sizes in assembly.

Next-generation sequencing (NGS) technologies have raised a challenging de novo genome assembly problem that is further amplified in recently emerged single-cell sequencing projects. While various NGS assemblers can use information from several libraries of read-pairs, most of them were originally developed for a single library and do not fully benefit from multiple libraries. Moreover, most assemblers assume uniform read coverage, condition that does not hold for single-cell projects where utilization of read-pairs is even more challenging. We have developed an exSPAnder algorithm that accurately resolves repeats in the case of both single and multiple libraries of read-pairs in both standard and single-cell assembly projects.

Motivation: The rapidly increasing set of sequenced genomes highlights the importance of identifying the synteny blocks in multiple and/or highly duplicated genomes. Most synteny block reconstruction algorithms use genes shared over all genomes to construct the synteny blocks for multiple genomes. However, the number of genes shared among all genomes quickly decreases with the increase in the number of genomes.
Results: We propose the Duplications and Rearrangements In Multiple Mammals (DRIMM)-Synteny algorithm to address this bottleneck and apply it to analyzing genomic architectures of yeast, plant and mammalian genomes. We further combine synteny block generation with rearrangement analysis to reconstruct the ancestral preduplicated yeast genome.

Mitochondria are a major target for aging and are instrumental in the age-dependent deterioration of the human brain, but studying mitochondria in aging human neurons has been challenging. Direct fibroblast-to-induced neuron (iN) conversion yields functional neurons that retain important signs of aging, in contrast to iPSC differentiation. Here, we analyzed mitochondrial features in iNs from individuals of different ages. iNs from old donors display decreased oxidative phosphorylation (OXPHOS)-related gene expression, impaired axonal mitochondrial morphologies, lower mitochondrial membrane potentials, reduced energy production, and increased oxidized proteins levels. In contrast, the fibroblasts from which iNs were generated show only mild age-dependent changes, consistent with a metabolic shift from glycolysis-dependent fibroblasts to OXPHOS-dependent iNs. Indeed, OXPHOS-induced old fibroblasts show increased mitochondrial aging features similar to iNs. Our data indicate that iNs are a valuable tool for studying mitochondrial aging and support a bioenergetic explanation for the high susceptibility of the brain to aging.

Motivation: de Bruijn graphs have been proposed as a data structure to facilitate the analysis of related whole genome sequences, in both a population and comparative genomic settings. However, current approaches do not scale well to many genomes of large size (such as mammalian genomes).
Results: In this article, we present TWOPACO, a simple and scalable low memory algorithm for the direct construction of the compacted de Bruijn graph from a set of complete genomes. We demonstrate that it can construct the graph for 100 simulated human genomes in less than a day and eight real primates in < 2 h, on a typical shared-memory machine. We believe that this progress will enable novel biological analyses of hundreds of mammalian-sized genomes.
Availability and Implementation: Our code and data is available for download from github.com/medvedevgroup/TwoPaCo.

Understanding the mechanisms driving lineage-specific evolution in both primates and rodents has been hindered by the lack of sister clades with a similar phylogenetic structure having high-quality genome assemblies. Here, we have created chromosome-level assemblies of the Mus caroli and Mus pahari genomes. Together with the Mus musculus and Rattus norvegicus genomes, this set of rodent genomes is similar in divergence times to the Hominidae (human-chimpanzee-gorilla-orangutan). By comparing the evolutionary dynamics between the Muridae and Hominidae, we identified punctate events of chromosome reshuffling that shaped the ancestral karyotype of Mus musculus and Mus caroli between 3 and 6 million yr ago, but that are absent in the Hominidae. Hominidae show between four- and sevenfold lower rates of nucleotide change and feature turnover in both neutral and functional sequences, suggesting an underlying coherence to the Muridae acceleration. Our system of matched, high-quality genome assemblies revealed how specific classes of repeats can play lineage-specific roles in related species. Recent LINE activity has remodeled protein-coding loci to a greater extent across the Muridae than the Hominidae, with functional consequences at the species level such as reproductive isolation. Furthermore, we charted a Muridae-specific retrotransposon expansion at unprecedented resolution, revealing how a single nucleotide mutation transformed a specific SINE element into an active CTCF binding site carrier specifically in Mus caroli, which resulted in thousands of novel, species-specific CTCF binding sites. Our results show that the comparison of matched phylogenetic sets of genomes will be an increasingly powerful strategy for understanding mammalian biology.

Despite the rapid development of sequencing technologies, the assembly of mammalian-scale genomes into complete chromosomes remains one of the most challenging problems in bioinformatics. To help address this difficulty, we developed Ragout 2, a reference-assisted assembly tool that works for large and complex genomes. By taking one or more target assemblies (generated from an NGS assembler) and one or multiple related reference genomes, Ragout 2 infers the evolutionary relationships between the genomes and builds the final assemblies using a genome rearrangement approach. By using Ragout 2, we transformed NGS assemblies of 16 laboratory mouse strains into sets of complete chromosomes, leaving < 5% of sequence unlocalized per set. Various benchmarks, including PCR testing and realigning of long Pacific Biosciences (PacBio) reads, suggest only a small number of structural errors in the final assemblies, comparable with direct assembly approaches. We applied Ragout 2 to the Mus caroli and Mus pahari genomes, which exhibit karyotype-scale variations compared with other genomes from the Muridae family. Chromosome painting maps confirmed most large-scale rearrangements that Ragout 2 detected. We applied Ragout 2 to improve draft sequences of three ape genomes that have recently been published. Ragout 2 transformed three sets of contigs (generated using PacBio reads only) into chromosome-scale assemblies with accuracy comparable to chromosome assemblies generated in the original study using BioNano maps, Hi-C, BAC clones, and FISH.

Background: Bipolar disorder is a serious and common psychiatric disorder characterized by manic and depressive mood switches and a relapsing and remitting course. The cornerstone of clinical management is stabilization and prophylaxis using mood-stabilizing medications to reduce both manic and depressive symptoms. Lithium remains the gold standard of treatment with the strongest data for both efficacy and suicide prevention. However, many patients do not respond to this medication, and clinically there is a great need for tools to aid the clinician in selecting the correct treatment. Large genome wide association studies (GWAS) investigating retrospectively the effect of lithium response are in the pipeline; however, few large prospective studies on genetic predictors to of lithium response have yet been conducted. The purpose of this project is to identify genes that are associated with lithium response in a large prospective cohort of bipolar patients and to better understand the mechanism of action of lithium and the variation in the genome that influences clinical response.
Methods/design: This study is an 11-site prospective non-randomized open trial of lithium designed to ascertain a cohort of 700 subjects with bipolar I disorder who experience protocol-defined relapse prevention as a result of treatment with lithium monotherapy. All patients will be diagnosed using the Diagnostic Interview for Genetic Studies (DIGS) and will then enter a 2-year follow-up period on lithium monotherapy if and when they exhibit a score of 1 (normal, not ill), 2 (minimally ill) or 3 (mildly ill) on the Clinical Global Impressions of Severity Scale for Bipolar Disorder (CGI-S-BP Overall Bipolar Illness) for 4 of the 5 preceding weeks. Lithium will be titrated as clinically appropriate, not to exceed serum levels of 1.2 mEq/L. The sample will be evaluated longitudinally using a wide range of clinical scales, cognitive assessments and laboratory tests. On relapse, patients will be discontinued or crossed-over to treatment with valproic acid (VPA) or treatment as usual (TAU). Relapse is defined as a DSM-IV manic, major depressive or mixed episode or if the treating physician decides a change in medication is clinically necessary. The sample will be genotyped for GWAS. The outcome for lithium response will be analyzed as a time to event, where the event is defined as clinical relapse, using a Cox Proportional Hazards model. Positive single nucleotide polymorphisms (SNPs) from past genetic retrospective studies of lithium response, the Consortium on Lithium Genetics (ConLiGen), will be tested in this prospective study sample; a meta-analysis of these samples will then be performed. Finally, neurons will be derived from pluripotent stem cells from lithium responders and non-responders and tested in vivo for response to lithium by gene expression studies. SNPs in genes identified in these cellular studies will also be tested for association to response.
Discussion: Lithium is an extraordinarily important therapeutic drug in the clinical management of patients suffering from bipolar disorder. However, a significant proportion of patients, 30-40 %, fail to respond, and there is currently no method to identify the good lithium responders before initiation of treatment. Converging evidence suggests that genetic factors play a strong role in the variation of response to lithium, but only a few genes have been tested and the samples have largely been retrospective or quite small. The current study will collect an entirely unique sample of 700 patients with bipolar disorder to be stabilized on lithium monotherapy and followed for up to 2 years. This study will produce useful information to improve the understanding of the mechanism of action of lithium and will add to the development of a method to predict individual response to lithium, thereby accelerating recovery and reducing suffering and cost.

Clustering hypertext document collection is an important task in Information Retrieval. Most clustering methods are based on document content and do not take into account the hyper-text links. Here we propose a novel PageRank based clustering (PRC) algorithm which uses the hypertext structure. The PRC algorithm produces graph partitioning with high modularity and coverage. The comparison of the PRC algorithm with two content based clustering algorithms shows that there is a good match between PRC clustering and content based clustering.

Genome assembly using high throughput data with short reads, arguably, remains an unresolvable task in repetitive genomes, since when the length of a repeat exceeds the read length, it becomes difficult to unambiguously connect the flanking regions. The emergence of third generation sequencing (Pacific Biosciences) with long reads enables the opportunity to resolve complicated repeats that could not be resolved by the short read data. However, these long reads have high error rate and it is an uphill task to assemble the genome without using additional high quality short reads. Recently, Koren et al. 2012 [1] proposed an approach to use high quality short reads data to correct these long reads and, thus, make the assembly from long reads possible. However, due to the large size of both dataset (short and long reads), error-correction of these long reads requires excessively high computational resources, even on small bacterial genomes. In this work, instead of error correction of long reads, we first assemble the short reads and later map these long reads on the assembly graph to resolve repeats.