Posters

Poster presentations at ISMB 2020 will be presented virtually. Authors will pre-record their poster talk (5-7 minutes) and will upload it to the virtual conference platform site along with a PDF of their poster. All registered conference participants will have access to the poster and presentation through the conference and content until October 31, 2020. There are Q&A opportunities through a chat function to allow interaction between presenters and participants.

Preliminary information on preparing your poster and poster talk are available at: https://www.iscb.org/ismb2020-general/presenterinfo#posters

Ideally authors should be available for interactive chat during the times noted below:

Search:

View Posters By Category

Poster Session A: July 13 & July 14 7:45 am - 9:15 am Eastern Daylight Time	Session B: July 15 and July 16 between 7:45 am - 9:15 am Eastern Daylight Time
Bioinfo-core CAMDA COSI COVID-19 Education COSI EvoCompGen COSI Function / CAFA 4 MLCSB COSI SCANGEN (Special Session) SysMod COSI Systems Immunology (Special Session) Text Mining July 14 between 10:40 am - 2:00 pm EDT iRNA COSI	3DSIG COSI Bio-Ontologies COSI BioVis COSI CompMS COSI COVID-19 EvoCompGen COSI HitSeq COSI MICROBIOME COSI NetBio COSI RegSys COSI TransMed COSI VarI COSI General Comp Bio

Short Abstract: While mitochondrial genetics have been extensively studied, the role of epigenetic regulation of the mitochondrial genome remains largely unexplored. DNA methylation plays a major role in regulating gene expression in the nucleus. The existence and potential role of mitochondrial DNA (mtDNA) methylation remain, however, controversial. Here, we employ a combination of dry and wet lab methodologies to assess mtDNA methylation in human prefrontal cortex from neurological healthy controls (n =27) and individuals with Parkinson’s disease (PD) (n = 27).
A specific double-mapping approach to analyse whole-genome bisulfite sequencing data efficiently removes false-positive misalignments of nuclear mitochondrial pseudogene sequences (NUMT).The ultra-deep mtDNA coverage (16842±4905) reveals an overall lack of mtDNA methylation with the exception of a single CpG hotspot at position m.545 on the heavy-strand-promoter. Furthermore, the m.545 methylation hotspot is validated using a combination of methylation-sensitive DNA restriction and quantitative PCR. In addition, we show that the methylation levels of m.545 positively correlate with mtDNA gene expression. We detect no differences in m.545 methylation between PD cases and controls.
Our findings suggest that, mtDNA methylation occurs in human prefrontal cortex exclusively on the m.545 position of the heavy-strand-promoter and it may play a role in mitochondrial gene expression.

Short Abstract: Amplicon sequencing has now been widely applied to explore genetic heterogeneity in populations and identify rare somatic mutations. The main challenge is how to confidently distinguish rare biological variants from noise sequences that can ultimately confound genetic analysis. One solution attaches Unique Molecular Identifiers (UMIs) to sample sequences before Polymerase Chain Reaction (PCR) amplification to increase the sensitivity for identifying ultra-low frequency mutations. Current algorithms that take into account UMI information either ignore errors within UMIs or cluster UMIs with an arbitrary threshold on the number of tolerated errors. Here we introduce a novel probabilistic framework for clustering amplicon sequences with unique molecular identifiers. A generative model is proposed not only for identifying biological variants but also for deduplicating PCR replicates. We test our model on both simulated data and a real Human Immunodeficiency Virus (HIV) dataset. The results show that our model achieves better accuracy than competing methods in identifying sequence variants and abundance estimation. Moreover, we show that clustering with UMIs can help correct possible biases in sequencing experiments.

Short Abstract: Recent techniques on single-cell RNA sequencing have boosted transcriptome-wide observation of gene expression dynamics of a heterogeneous cell population at a single-cell scale. Typical examples of such analysis include inference of a pseudotime cell trajectory, and comparison of pseudotime trajectories between different experimental conditions will tell us how feature genes regulate a dynamic cellular process. Existing methods for comparing pseudotime trajectories, however, force users to select trajectories to be compared because most of them can deal only with simple linear trajectories, leading to the possibility of making a biased interpretation. Here we present CAPITAL, a method for comparing complex pseudotime trajectories with tree alignment whereby trajectories including branching can be compared without any knowledge of paths in the trajectories to be compared. Computational tests on public data indicate that CAPITAL can align non-linear pseudotime trajectories at a sufficient level of known annotations, and reveal expression dynamics of several marker genes.

Short Abstract: While Next Generation Sequencing with short reads can achieve high accuracy in most of the human genome, specialized sequencing and/or analysis methods are required for calling structural variants (SVs) and variants in repetitive regions such as tandem repeats, homologous genes, and pseudogenes. The Genome In A Bottle (GIAB) consortium was established to authoritatively characterize genomes for use in method validation and benchmarking. Here, we use linked- and long-reads to establish new benchmarks for structural variants and small variants in challenging genomic regions. The new SV benchmark set includes ~10,000 insertions and deletions >50bp in size inside 2.66 Gbp benchmark regions supported by diploid assemblies. Many of these SVs were sequence-resolved by long reads and de novo assembly. For the new small variant benchmark, long and linked-read data added >300,000 new variants and 189 million bp, over half in regions that are difficult to map with short reads, including coding regions in medically important genes (e.g., PMS2 and CYP21A2). To demonstrate the utility of these benchmark sets, we have compared variant calls from a variety of technologies to the benchmark sets, and we have shown that they reliably identify false positives and false negatives.

Short Abstract: We recently demonstrated that splitting large sequencing datasets into sequences of length k (a procedure called k-mer decomposition) and exploring these with an adaptive genetic algorithm allowed a more accurate classification and prediction of disease outcome. In addition, this analysis could be performed without a reference genome or transcriptome and was applicable to any sequencing technology. We now demonstrate through advanced machine learning approaches that our k-mer approaches can help explore RNA-seq data more thoroughly, much faster and with much less computational resources than classical bioinformatics approaches. In addition, the use of our approach can help protect patients identity and allows for better exchange of sensitive genomic data.

Short Abstract: Although a massive ChIP-seq data is publicly available, current ChIP-seq databases have been established with a limited number of samples or insufficient quality assessment due to difficulty in large scale public ChIP-seq data analysis. To aim at providing ChIP-seq analysis results for comprehensive public data with quality assessable information, we launched C4S DB, a Comprehensive Collection and Comparison for ChIP-Seq database. We designed C4S DB to treat human ChIP-seq data released at the ENCODE portal and GEO. To cover as many as possible public datasets, we developed a semi-automated workflow to extract metadata for GEO ChIP-seq data. Since making researchers possible to choose data with suitable quality is crucial, our analysis pipeline includes steps to calculate several QC metrics which are comparable to the ENCODE project. Moreover, we also provide our novel QC metric, which does not require peak calling to evaluate ChIP noise level. C4S DB shares these statistics. It also provides the functions to visualise the orchestration of transcriptional regulatory elements around a gene and to map the similarity between experiments to infer regulatory elements' relationships. C4S DB will be a powerful platform to work with large scale public ChIP-seq data.

Short Abstract: c-Myc is critical for the positive selection of germinal centre (GC) B cells. It indirectly regulates the alternative splicing (AS) of transcripts (e.g. Pkm) via induction of the RNA-binding protein PTBP1. There are still unrevealed alternative isoforms of most transcripts which might mediate functional roles in GC B cells. The changes in the AS in the positively selected GC B cells have been previously addressed using the short-read Illumina sequencing. However, the majority of the transcript variants generated by AS and alternative polyadenylation events were not detected. The long-read sequencing platforms emerged out of the need for full-length sequencing to resolve the complexity of isoforms. In our study, we adapted the long-read sequencing, Oxford Nanopore Technology (ONT), using a Smart-seq2 approach to understand the post-transcriptional regulation in positively selected GC B cells. Using 1ng total RNA, we reached a depth of approximately 2-million reads per sample and detected transcripts from over 9500 genes with at least 5 supporting reads. Our findings support that Smart-seq2 adapted ONT RNA sequencing is a powerful workflow for the identification and quantification of complex isoforms in positively selected GC B cells and for the development of a single-cell long-read sequencing platform for rare cell populations.

Short Abstract: Single-cell RNA-seq analysis is an effective tool for revealing differences in various cell types and states. Analyzing gene expression data is considered to be an effective approach for discovering biological mechanisms, especially cellular heterogeneity. In particular, understanding how cells change under different conditions is a very important issue for the further effective use of stem cells. However, a robust approach to express the process of state change based on gene expression data has not been established. In this presentation, we propose a method for identifying cell lineages for a data set containing cells in two different conditions: mutant and wild type. First, cluster cells based on the gene expression level, then infer a cell trajectory by applying partition-based graph abstraction (PAGA). We used Seurat version 3.1 and Scanpy version 1.4.4 for clustering and constructing PAGA graphs. We applied our method to the trajectory inference of the differentiation process of mouse embryonic stem cells with two time points, days 0 and 5 . We will present the results of comparing the trajectories between mutant and wild-type cells, and identify the effect of the mutation in the differentiation process.

Short Abstract: High-Throughput Transcriptomics (HTTr) is emerging as a cost-effective method for broadly assessing chemical toxicity across many target pathways and modes of action in a single assay. US EPA has designed an in vitro chemical screening protocol using targeted RNA-seq wherein batches of chemicals are tested in 8-point concentration series in triplicate using a 384-well plate format. Using a pilot set of 42 chemicals exposed to MCF7 cells, we have demonstrated the utility of this platform for predicting both biological target and overall point of departure for each chemical. US EPA has subsequently scaled this screening platform to several thousand chemicals across multiple cell lines. Our recent work on this growing collection of data has included: 1) development of an open-source pipeline for rapid and robust processing of targeted RNA-seq data; 2) comparison of multiple analytical methods for estimating differential expression and dose-response models; 3) quantification of reproducibility across platforms and methods; and 4) signature-level analysis methods to summarize results and link findings to interpretable biology and known hazards. Notably, the results show that aggregating signal at the signature level improves reproducibility and reduces uncertainty in screening results. This abstract does not necessarily reflect US EPA policy.

Short Abstract: Clinical Variant Detection and Phasing through Automated Long Read Clustering
Qian Zeng*, Lax Iyer*, Ingrid Chen*, Stan Letovsky & Brian Krueger, Laboratory Corporation of America Holdings
John Harting & Kristina Weber, Pacific Biosciences
*Equal Contribution

Short Abstract: The changing landscape of genomics research and clinical practice warrants computational pipelines that efficiently orchestrate complex analyses of large data volumes across heterogeneous environments.
We performed a systematic evaluation of the popular bioinformatics Workflow Management Systems (WfMS): Nextflow, CWL and WDL and some of their executors, along with Swift/T, a WfMS commonly used in high-scale physics applications. A genomic variant calling pipeline and a scalability-testing framework were developed in each and run locally, on an HPC cluster and in the cloud. This allowed contrasting the four in terms of language expressiveness, modularity, scalability, robustness, reproducibility, interoperability, ease of development, and adoptability.
This evaluation helps guide the choice of the appropriate WfMs as driven by the interplay of the workflow language, executor and computational ecosystem, as well as the collaborative practices of our field. We also arrived at a set of design considerations to further improve WfMS deployment in research and clinical settings.
As the community and its needs continue to evolve along with computational infrastructure, WfMS will evolve, especially those with permissive licenses. Just as the dataflow paradigm and containerization are now widely used, this field will continue to see innovations elsewhere, such as big data technologies and portability.

Short Abstract: Performing comprehensive quality control is necessary to remove poor quality cells in single-cell RNA sequencing (scRNA-seq) data. Artifacts in the scRNA-seq data, such as doublets or ambient RNA, can also hinder downstream clustering and marker selection and need to be assessed. While several algorithms have been developed to perform various quality control tasks, they are only available in different packages across various programming environments. We have built an easy-to-use pipeline in the singleCellTK package that generates a comprehensive summary of quality control metrics from several existing tools. Our pipeline is able to import data from various preprocessing tools, compute general quality control metrics, detect doublets, and predict and correct for ambient RNA contamination for each cell. Further, to make the data accessible in down-stream analysis workflows, it exports the results to common data structures including the SingleCellExperiment object for analysis in R, the AnnData object for analysis in Python, as well as flat text files for use in any generic workflow. Overall, this pipeline will help streamline and standardize quality control analyses for scRNA-seq data.

Short Abstract: Assembling reference-quality, chromosome-level genomes for both model and novel eukaryotic organisms is an increasingly achievable task for single research teams. However, the broad variety of sequencing technologies, assembly algorithms, and post-assembly processing tools currently available means that there is no clear consensus on a best-practice computational protocol for eukaryotic de novo genome assembly. Here, we provide a comprehensive benchmark of 28 state-of-the-art assembly and polishing packages, in various combinations, when assembling four eukaryotic genomes using both next generation (Illumina HiSeq) and third generation (Oxford Nanopore Technologies R9 and PacBio RSII) sequencing data. Recommendations are made for the most effective tools given particular genome constraints, such as repeat content, GC%, sequencing depth, and genome size, and against common assessment metrics such as contiguity, computational performance, gene completeness, and reference reconstruction. We also present a Snakemake-based pipeline for eukaryotic genome assembly, pyro, to further assist future de novo assembly and package benchmarking projects.

Short Abstract: Studies combining PCR techniques and next-generation DNA sequencing (NGS) to characterize the HIV genetic diversity and understand the reservoir dynamics have been extensively reported. In such investigation thousands of sequences are generated, requiring automation in the characterization process. Here we present the Defective and Intact HIV genome Assembler (DIHIVA), a pipeline to annotate and classify HIV genomes based on NGS data. First, we remove PCR amplification and perform error correction using BBtools. After, a quality-control check is carried out by Trim Galore package to trim Illumina adapters and low-quality bases. BBtools is also used to remove possible contaminant reads using HIV genome sequences obtained from Los Alamos HIV database. We then use SPAdes, a k-mer based assembler, to reconstruct the HIV-1 sequences. The longest assembled contig is aligned via BLAST to the Los Alamos HIV genome database, in order to set it in the forward orientation. Finally, the HIV genome is annotated against Hxb2 using BLAST. In the end, sequences are classified as intact or defective, which could be Major Splicing Donor (MSD) Mutation, Non-functional, Missing Internal Genes or Weird. In order to provide access to non-computer specialists we developed a user-friendly interface for the pipeline using R/Shiny.

Short Abstract: We present distinct, a statistical method to perform, via hierarchical permutation tests, differential analyses between groups of densities. distinct is a general and flexible tool: due to its fully non-parametric nature, which makes no assumptions on how the data was generated, it can be applied to a variety of datasets. It is particularly suitable to perform differential analyses on single cell data, such as single cell RNA sequencing (scRNA-seq) and high-dimensional flow or mass cytometry (HDCyto) data.
While most methods for differential expression target differences in the mean abundance between conditions, single-cell data can show more complex variations. distinct, by comparing full distributions, identifies, both, differential patterns involving changes in the mean, as well as more subtle variations that do not involve the mean (e.g., unimodal vs. bi-modal distributions with the same mean).
distinct explicitly models the variability between samples (i.e., biological replicates), can adjust for covariates (e.g., batch effects) and allows multi-group (i.e., >2 groups) comparisons.
We will present results, based on scRNA-seq and HDCyto simulated and experimental datasets, where distinct outperforms several competitors and is able to detect more patterns of differential expression compared to canonical differential methods.
distinct is freely available as a Bioconductor R package.

Short Abstract: RNA sequencing has become a standard technique for investigating the transcriptome of eukaryotes, bacteria, and archaea. The Dual RNA-seq protocol extends the RNA-seq technique to measure transcript expression in a host-pathogen system. However, simultaneous profiling of the gene expression of a pathogen and its host raises additional complications, particularly in read quantification.

Here, we present a Nextflow pipeline for quantifying Dual RNA-seq data. Host and pathogen RNA-seq reads are processed together, and, after quality control and trimming, are mapped onto two different references, the bacterial and eukaryotic genomes. We have implemented two read quantification strategies. The first is alignment-based mapping of reads onto the genomes with STAR followed by quantification with HTSeq, estimating gene expression using uniquely mapped reads. For the second we used Salmon with Selective Alignment, a fast transcriptome quantification method that handles multi-mapped reads. This allows us to investigate the importance of multi-mapped reads that may originate from different gene isoforms in the host and repetitive elements that are highly abundant in some bacterial genomes, as well as cross-mapping reads that may originate from either transcriptome. Using simulations of a variety of host-pathogen systems, we provide initial guidance towards optimal read quantification strategies for Dual RNA-seq experiments.

Short Abstract: The low cost of sequencing has made epigenomics a powerful instrument for the study of chromatin accessibility, histone modifications, and other epigenetic mechanisms.

Several tools have been developed to tackle the main steps of the computational and statistical analyses. The ENCODE consortium has developed a set of best practices and experimental guidelines for transcriptomic and epigenomic experiments. However, these guidelines focus on signal discovery and there is still no consensus on the best practices for the detection of differential signal, e.g., when comparing tissues under different conditions or following different treatments.

Here, we focus in particular on ATAC-seq/SONO-seq and Histone Modification ChIP-seq. We present three complete workflows for the differential analysis of epigenomic data, based on the Bioconductor packages DiffBind, DEScan2, and CSAW, respectively. We compare the approaches on real data and provide guidelines on how to best apply each pipeline to one’s data. Moreover, we provide some indication on lower-level, and often neglected, steps of the analysis, such as the impact of the aligner, read de-duplication, and normalization on the final results.

Short Abstract: A vast amount of genome sequencing data has been produced over the past decades. This genomic data has exploded with the advent of single-cell analysis. Especially long-read sequencing becomes dominant in genomics, the genomic data explosion requires an enormous level of storage capacity. It costs a lot of time and resources to store and transfer the data and causes a bottleneck in genome sequencing analysis. To resolve this issue, various compression techniques have been proposed to reduce the size of original raw sequencing data in FASTQ, but they have focused on short-read sequencing only.
In this study, we design a compression algorithm based on a read reordering using a scoring model to increase the compression ratio for the long-read sequencing data. We evaluate our FastqCLS tool using benchmark datasets that have been commonly used in other FASTQ compression studies. We also include new long-read sequencing data in this validation. We compare our method with existing major FASTQ compression tools. As a result, our FastqCLS outperforms in terms of a compression ratio for storing long-read sequencing data supporting a variety of the read-lengths. FastqCLS can be downloaded from github.com/krlucete/FastqCLS.

Short Abstract: Accurate high-throughput sequencing (HTS) of genetic variants is vital for research and medicine. Certain genomic positions can be prone to higher rates of systematic sequencing and alignment bias, resulting in false-positive variant calls. Current standard practices to identify loci that cannot be sequenced accurately utilize consensus between different sequencing methods as a proxy for sequencing confidence. These practices have significant limitations that remain unaddressed.

We have developed a novel statistical method, summarizing sequenced reads from whole genome clinical samples in “Incremental Databases” maintaining patient confidentiality. Allele statistics were cataloged for each genomic position that consistently showed systematic biases with the corresponding HTS pipeline.

We found systematic biases at ~1-3% of the human autosomal genome across five patient cohorts. We identified how susceptible different genomic regions were to systematic biases, including large homopolymer flanks (odds ratio=23.29-33.69) and the NIST high-confidence genomic regions (odds ratio=0.154-0.191). We confirmed our predictions on gold-standard reference genomes and demonstrated these systematic biases lead to suspect variant calls within clinical panels.

This study implements a novel method to enhance quality control of sequenced samples, by flagging variants displaying systematic sequencing biases for further analysis or exclusion. This work has recently been accepted for publication in Genome Research.

Short Abstract: Motivation: Although chromosome conformation capture techniques have become more standardized over the recent years, the computational analyses downstream are still hard to navigate. While some workflows have been widely adopted, it is clear that each of these algorithms only answers a subset of questions one can ask of this kind of data. Therefore, we think it is crucial to leverage several of these approaches to gain a more complete picture.

Methods: Using the Snakemake workflow engine, we are combining the strengths of packages such as the Juicer pipeline, HiCExplorer tools and HOMER. This workflow is modular and can be easily adjusted for different input data or by adding new algorithms. With this workflow, we were able to explore how different read depths impact the detection of chromosome loops at higher resolutions. Additionally, we benchmarked how the number of replicates effect the reproducibility of the results.

Conclusion: We created a workflow that combines reproducibility, quality control, ease of use and comprehensive visualization to address aspects important to any next generation sequencing analysis. We show best practices using publicly available data to help guide researchers when planning their experiments.

Short Abstract: Very large deep sequencing datasets have become ubiquitous in biomedical research, and efficient algorithms are continuously developed to handle them. Many of them use minimizers to obtain speed-up and save memory usage. Recently, we suggested the use of universal hitting sets (UHS) to potentially improve over minimizers. Here, we showcase the benefit of UHS in the de Bruijn graph construction step of genome assembly by the MSP algorithm. We show that using UHS instead of minimizers leads to faster runtime, more balanced bin partitions and less overall memory.

Short Abstract: Single-cell data is now being collected across many patients in varying conditions. However, data is still relatively expensive. This opens up the opportunity for computational methods to decrease overall cost by inferring a single-cell measurement based on similarity to the meta-data of other similar samples. We examine this problem with an optimal transport perspective. This allows us to leverage a variant of the Sinkhorn algorithm for extremely computationally efficient approximations of transport along discrete manifolds. Our method first constructs the manifold between samples, then aligns this to the manifold of patients, and finally applies this to interpolate a barycenter sample along this manifold. We show first that we are able to better interpolate samples between timepoints than existing methods e.g. Waddington-OT (Schiebinger et al. 2019 Cell) by accounting for structure between multiple timepoints instead of pairs. We then show when the relationship between patients is an inferred manifold, how to impute a patient’s single-cell measurements based on other similar single-cell samples by aligning the manifold of patients with that of single-cell measurements. When the manifold of patients exhibits non-linear but intrinsically low-dimensional structure, we are able to more accurately infer a single-cell measurement.

Short Abstract: High-throughput single cell transcriptomics analysis is widely used to study human tumors, however a major challenge is distinguishing the stromal cells from the malignant cancer cells, as well as clonal substructure within tumors. To address this challenge, we developed an integrative Bayesian segmentation approach, COPYKAT to estimate genomic copy numbers at 5MB resolution from high-throughput single cell RNA-seq data. We applied COPYKAT to 39,709 single cells from 16 tumors across 4 cancer types, including premalignant and triple-negative breast cancers, pancreatic ductal adenocarcinomas, and anaplastic thyroid cancer. From these data we could accurately (98%) classify tumor cells from stromal cells. In three TNBC tumors COPYKAT resolved multiple clonal subpopulations of genotypes that differed in expression of breast cancer genes and enrichment of cancer hallmarks including EMT and hypoxia. These data show that COPYKAT can accurately resolve clonal copy number substructure in tumors and classify tumor and normal cells in a variety of human cancers.

Short Abstract: Double minute chromosomes are highly amplified oncogenic, acentric, extrachromosomal DNA that are frequently observed in the cells of numerous cancer types. Algorithmic discovery of double minutes (DM) can potentially improve bench-derived therapies for cancer treatment. A hindrance to this task is that DMs evolve, yielding circular chromatin that shares segments from progenitor double minutes. This creates multiple double minutes in the tumor genome that are distinct, but that share loci for overlapping amplicon coordinates. Existing DM discovery algorithms (that use only whole genome sequencing data) can potentially misclassify DMs that share overlapping coordinates. In this study, we describe a method called HolistIC that predicts double minutes in tumor genomes by integrating whole genome shotgun sequencing (WGS) and Hi-C sequencing data. This resolves ambiguity in double minute prediction that exists when using WGS data alone, a limitation of existing methods for this problem. We implemented and tested our algorithm on the tandem Hi-C and WGS datasets of two cancer datasets and a simulated dataset. Our results show that HolistIC can distinguish between double minutes that share amplicon coordinates, an advance over current methods for this problem.

Short Abstract: With the advent of long read sequencing (LRS) platforms, there has been an increase in precision and a higher throughput sequencing of transcripts. Software SQANTI has been previously developed by the Conesa lab to qualitatively analyze LRS data. However, a quantifying method relying solely on LRS has not been developed. Such method requires a deeper understanding of the underlying distribution of LRS. Our goal is to evaluate how sequencing depth affects transcript distribution, as a first step in developing an LRS methodology for transcript quantification. To assess this, saturation function in NOIseq package was re-designed to evaluate LRS and applied to the Iso-seq Melanoma and Alzheimer (Sequel2) datasets published by PacBio. A total of 9K detected transcripts ( >10 counts) were found in Melanoma, from 700K reads. While 24K detected transcripts (>10 counts) were found in Alzheimer, from a total of 2M reads. Additionally, data showed transcript length quantification bias, with longer transcripts having lower number of reads. Finally, saturation is achieved with Sequel2 levels of coverage (Alzheimer), but not in those generated by previous technologies (Melanoma). This indicates that at Sequel2 level sequencing depths, a significant fraction of the transcriptome can be detected with a significant read coverage.

Short Abstract: Whole-genome shotgun sequencing enables the analysis of microbial communities in unprecedented detail, with important implications in medicine and ecology. Predicting the presence and relative abundances of microbes in a sample, known as “metagenomic profiling”, is a critical step in microbiome analysis. Existing profiling methods have been shown to suffer from poor false positive or false negative rates, while alignment-based approaches are often considered accurate but computationally infeasible. Here we present a novel method, Metalign, that addresses these concerns by performing efficient alignment-based metagenomic profiling.
Metalign employs a high-speed, high-recall pre-filtering method based on the mathematical concept of Containment Min Hash to reduce the reference database size dramatically before alignment, followed by a method to estimate organism relative abundances in the sample by handling reads aligned to multiple genomes. We show that Metalign achieves significantly improved results over existing methods on simulated datasets (Figure 1) from a large benchmarking study, CAMI, and performs well on in vitro mock community data and environmental data from the Tara Oceans project. Metalign is freely available at github.com/nlapier2/Metalign, and via bioconda.

Short Abstract: Taxonomic classification is an important step in the analysis of samples found in metagenomic studies. Conventional mapping-based methods trade off between high memory and low recall, with recent deep learning methods suffering from very large model sizes. We aim to develop a more memory-efficient technique for taxonomic classification. A task of particular interest is abundance estimation. Current methods initially classify reads independently and are agnostic to the co-occurence patterns between taxa. In this work, we also attempt to take these patterns into account. We develop a novel memory-efficient read classification technique, combining deep learning and locality-sensitive hashing. We show that this approach outperforms conventional mapping-based and other deep learning methods for taxonomic classification when restricting all methods to a fixed memory footprint. Moreover, we formulate the task of abundance estimation as a Multiple Instance Learning problem and we extend current deep learning architectures with two types of permutation-invariant MIL pooling layers: a) deepsets and b) attention-based pooling. We illustrate that our architectures can exploit the co-occurrence of species in metagenomic read sets and outperform the single-read architectures in predicting the distribution over taxa at higher taxonomic ranks.

Short Abstract: We introduce a statistical model for the estimation of sample index-hopping rate in multiplexed droplet-based single-cell RNA sequencing data and probabilistic inference of the true sample of origin of the hopped reads. Across the datasets we analyzed, we estimate the sample index hopping probability to range between 0.003–0.009, a small number that counter-intuitively gives rise to a large fraction of "phantom molecules" – in more than 25% of samples, the fraction of phantom molecules exceeds 8%, with this fraction reaching as high as 85% in low-complexity samples. These phantom molecules lead to widespread complications in downstream analyses, including transcriptome mixing across cells, emergence of phantom copies of cells from other samples, and misclassification of empty droplets as cells. We demonstrate that our model-based approach can correct for these artifacts by accurately purging the majority of phantom molecules from the data. An R package implementing our approach along with reproducible R markdown notebooks of the analyses are available at
csglab.github.io/PhantomPurgeR/.

Short Abstract: Immunotherapy has emerged as one of the main treatments for patients with cancer. Understanding the complexity of tumor microenvironment will enable the discovery of new mechanisms and biomarkers. Studies have used spatial profiling technologies such as NanoString Digital Spatial Profiler (DSP) to discover predictors of therapy response. However, lack of analysis tools and proper statistical approaches hinders the effective transformation of spatial omics data into biologically meaningful results. Here, we developed an open-source R/Bioconductor package, nDSPA, specially designed to leverage the spatial elements for data analysis and hypothesis testing. It implements QC, data preprocessing, visualization, and comparison of transcript or protein expression levels from defined regions of interest (ROIs). By evaluating the performance of differential expression identification using various statistical models, we recommend linear mixed effect models incorporating multiple ROIs, the distance between ROIs, and technical batches. nDSPA will be released on GitHub and accompanied by an R Shiny application for interactive data exploration. While built on DSP data, our statistical algorithm can be generalized to other platforms with spatial compartments. We hope our method will provide new insights into the rapidly developing landscape of data-driven biomarker discovery and enable the broad application of spatial omics technologies in translational research.

Short Abstract: The ability to generate high-quality genome sequences is cornerstone to modern biological research. Even with recent advancements in sequencing technologies, many genome assemblies are still not achieving reference-grade. Recently, we have developed ntJoin, a tool that leverages structural synteny between a draft assembly and reference sequence(s) to contiguate and correct the former with respect to the latter. Instead of alignments, ntJoin uses a lightweight mapping approach based on a graph data structure generated from ordered minimizer sketches. The tool can be used in a variety of different applications, including (but not limited to) improving a draft assembly with a reference-grade genome, a short read assembly with a draft long read assembly, and a draft assembly with an assembly from a closely-related species. When scaffolding a human short read assembly using the reference human genome or a long read assembly, ntJoin improves the NGA50 length 23- and 13-fold, respectively, in under 13 min, using less than 11 GB of RAM. Compared to existing reference-guided scaffolders, ntJoin generates highly contiguous assemblies faster and using less memory. ntJoin is written in C++ and Python, and is freely available at github.com/bcgsc/ntjoin.

Short Abstract: High-throughput transcriptomics (HTTr) shows promise as a useful method in chemical safety assessment due to its broad biological coverage and potential for identifying diverse mechanisms of action for many chemicals. Concentration response profiling also provides information on chemical hazard and potency, which can inform chemical prioritization. US EPA has screened ~2,200 chemicals in 8-point concentration response in MCF-7 cells using the TempO-Seq whole transcriptome assay. Screening was performed in 384-well plate format and included multiple reference samples to assess the biological and technical variability of the platform. We developed several quality control metrics based on mapping rate and read count distributions across probes to flag low-quality samples and estimate that ~94% of samples had sufficient quality. Analysis of reference samples demonstrated high reproducibility across plates and experimental groups, with a median correlation near 0.93. Further refinement of the concentration-response modeling suggests signature-level analyses have predictive power for identifying points of departure for specific mechanisms of bioactivity. This work represents one of the largest transcriptomic chemical safety screens to date, shows promise as a first-tier screening method for hazard identification, and may inform the prioritization of chemicals for additional study. This abstract does not necessarily reflect US EPA policy

Short Abstract: The advent of new generation sequencing for the first time made it possible to significantly speed up and reduce the cost of determining the full sequence of millions of genomes of organisms, from bacteria to humans. It is clear from the bioinformatics side that as the cost of sequencing decreases, the amount of data to be processed will increase. It is necessary to identify areas of analysis that are routine and to automate them.
We created Pannopi - an open-source, scalable, easy-to-install, and use an assembly and annotation pipeline based on a hierarchical pan-genome graph. The program performs a large-scale analysis of the nucleic acid sequence of bacteria from preparation to functional annotation. The process runs from the preparation of sequence reads to genome assembly, through cleaning up the genome from external contamination to structural and functional annotation. Quality control is carried out throughout the process. Pannopi allows users to select the taxons for pan-genome graph construction and required modules; it can be used on a separate command-line program or through a web interface.

Short Abstract: Map-based (aka hierarchical shotgun) sequencing and assembly approaches from the Sanger sequencing era tend to generate highly contiguous assemblies, essential for a broad range of downstream analysis. However, these methods are expensive, labor-intensive, and time-consuming compared to methods based exclusively on newer high-throughput sequencing (HiTSeq) technologies. Herein, we transfer the concept of physical maps from the earlier map-based assembly approaches into the domain of next-generation sequencing data. We introduce Physlr, a tool that utilizes long-range information provided by linked-read technologies to construct a chromosome-scale physical map de novo, which can then be used to scaffold a draft assembly of any sequencing data. We benchmark Physlr using stLFR linked reads of three human individuals (NA12878, NA24695, and NA24143). For all datasets, Physlr improves across all studied metrics compared to state-of-the-art linked-read scaffolders, and yields chromosome-level contiguity. For NA12878, Physlr results in 56 Mb NG50 (8.3- and 1.19-fold increase over baseline and best comparator (ARCS), respectively), 7.32 Mb NGA50 (0.78- and 0.08-fold increase, respectively), and fewer misassemblies (-6.5%).
We also apply Physlr on a long-read Shasta assembly scaffolded with ARCS and find the NG50 and NGA50 increase by 187% (95.5 Mb) and 10% (20.7 Mb), and misassemblies decrease by 28%.

Short Abstract: Transcriptome profiling is one of the most frequently used technologies and key to interpreting the function of the genome in human diseases. However, quantification of transcript expression with short read RNA-sequencing remains challenging as different transcripts from the same gene are often highly similar. Nanopore RNA Sequencing reduces the complexity of transcriptome profiling with ultra-long reads that can cover the full length of the isoforms. The technology has a high sequencing error rate and often generates shorter, fragmented reads due to RNA degradation, however, currently no specific transcript quantification method exists for such data. Here, we present bambu, a long read isoform discovery and quantification method. Bambu performs probabilistic assignment of reads to annotated and novel transcripts across samples to improve the accuracy of transcript expression estimates. We apply our method to cancer cell line data with spike-in controls, and compare the results with estimates obtained from short read data. Bambu recovered annotated isoforms from spike-ins and showed consistency in gene expression estimation with existing methods for short read RNA-Sequencing data, but improved accuracy in transcript expression estimation. The method is implemented in R (github.com/GoekeLab/bambu), enabling simple, fast, and accurate analysis of long read transcriptome profiling data.

Short Abstract: The development of high-throughput sequencing (HTS) technologies has created a supply of large volumes of sequence data. Genotyping by sequencing (GBS) in particular, is a cost-effective strategy to obtain a large number of samples through reduced representation. To use this data, traditionally, variant calling is performed against a reference genome. In the absence of such a reference, which is the case for most species, variant detection must be done de-novo. Several software solutions have been developed for this purpose, but unfortunately most are either inefficient or discover low numbers of polymorphisms. Here we present a new algorithm for efficient analysis of GBS reads without a reference genome. Validation was conducted using publicly available data of different species and different sequencing protocols as well as different population structures. GBS data of an interspecific rice (Oryza sativa) biparental population was analysed. The 10,000 SNVs discovered with an F1-score up to 5x better than other available technologies. DDrad data from Sea Bass was used to successfully validate diversity analysis, by replicating phylogenetic and population structures. This was done at a fraction of the time: processing 1 Mb of data in about ~3.3s. This represents half the time of the next fastest technology.

Short Abstract: De novo genome assembly is an important tool for modern molecular biology. It is also a useful method for studying genomes with high variation, such as in the context of cancer, as it is not biased by any reference. De novo short read assemblers commonly use de Bruijn graphs where sequences initially overlap by the number of bases specified with parameter k, followed by the identification of unambiguous walks on the graph. This parameter selection is a trade-off between connectivity and contiguity, but is inherently a crude approach that only works on average. Here, we present RResolver, an algorithm that takes a short read assembly with a fixed k as input, and then uses a k value close to the read size to resolve repeats in the graph. This is accomplished by a sliding window along possible paths to determine which ones are unsupported by reads and should be removed. When used with the popular short read assembler ABySS, RResolver takes 4 hours (6% of the whole pipeline) with 48 threads, and 40GB memory to run on a 150bp read human assembly. It improves contiguity (NGA50) by 5% while reducing misassemblies by 5% in the final result.

Short Abstract: Biomedical studies routinely generate ‘omics’ data representing marker profiles (genes, proteins, metabolites, etc.) of the experimental conditions under investigation. The corresponding raw data are usually deposited in public repositories, such as the NIH’s Gene Expression Omnibus (GEO) or EMBL-EBI’s ArrayExpress. However, extraction of the differential signatures associated with a phenotype often requires significant processing. Furthermore, not all studies make the raw data available and only publish finite sets of marker identifiers. A structured, efficient, and openly accessible platform for signature storage and analysis would thus address an unmet need. We have developed SigRepo, an R-based software system for the representation, storage, and interactive retrieval of omics signatures. The system includes the definition of R6 objects for the representation of signatures and signature collections; a MySQL Database for efficient storage, search, and retrieval of signatures; a suite of R command line functions; and an R Shiny Interface for interactive query and analysis of stored signatures, including signature comparison and pathway enrichment analysis. The system was developed as a Bioconductor package and an associated database, for installation on most architectures, to support both interactive and batch signature-based analyses.

Short Abstract: Disability, age-related diseases, and morbidity are common risk factors with advancing age. Recent studies have shown that age-related disability and diseases are delayed in people living to 100 and especially compressed in those surviving greater than 100 years. Changes in the immune system with age have also been demonstrated, including differences in proportion of immune cell populations and gene expression differences over age. In order to further explore immune cell populations during age progression, we investigated cell subpopulations of the peripheral blood immune system across age using publicly available single cell RNA-sequencing datasets of peripheral blood mononuclear cells from multiple subjects of extreme longevity and younger control ages ranging from 20-80 years. We performed integrated analyses of these datasets to gain higher resolution of subpopulations of immune cells across age and to generate corresponding gene expression signatures for extreme longevity subjects compared to controls. Early findings suggest changes in cell type composition are associated with age even in smaller subpopulations of cells.

Short Abstract: While animal model studies have extensively defined mechanisms controlling cell diversity in the developing mammalian lung, the limited data available from late stage human lung development represents a significant knowledge gap.
Single cell RNAseq generated transcriptional profiles of 5500 cells obtained from two newborn human lungs. Previously frozen single cell isolates were captured, and library preparation was completed on the Chromium 10X system. Data analysis was performed in Seurat, while cell type annotation was performed using the ToppGene. Single cell sequence data from 32000 mouse lung cells were used for comparison to the human data.
Transcriptional interrogation of newborn human lung cells readily identified distinct clusters including multiple populations of endothelial, mesenchymal and immune cells. Signature genes from each of these populations were identified. Computational integration of newborn human and postnatal mouse lung development cellular transcriptomes indicated they are highly comparable and facilitated the identification of distinct cellular subtypes. Comparison of the newborn human and mouse cellular transcriptomes also demonstrated cell type-specific differences in maturation states of newborn human lung cells. Matrix fibroblasts could be separated into younger or older cells. Cells with each molecular profile were spatially resolved within newborn human lung tissue.

Short Abstract: Aneuploidy plays critical roles in genome evolution.
Alleles, whose dosages affect the fitness of an ancestor, will have altered frequencies in the descendant populations upon perturbation.
Single-cell sequencing enables comprehensive genome-wide copy number profiling of thousands of cells at various evolutionary stage and lineage. That makes it possible to discover dosage effects invisible at tissue level, provided that the cell lineages can be accurately reconstructed.
Here, we present a Minimal Event Distance Aneuploidy Lineage Tree (MEDALT) algorithm that infers the evolution history of a cell population based on single-cell copy number (SCCN) profiles. We also present a statistical routine named lineage speciation analysis, which facilitates discovery of fitness-associated alterations and genes from SCCN lineage trees.
We assessed our approaches using a variety of single-cell datasets. Overall, MEDALT appeared more accurate than phylogenetics approaches in reconstructing copy number lineage. From the single-cell DNA-sequencing data of 20 triple-negative breast cancer patients, our approaches effectively prioritized genes that are essential for breast cancer cell fitness and are predictive of patient survival, including those implicating convergent evolution. Similar benefits were observed when applying our approaches on single-cell RNA sequencing data obtained from cancer patients.
The source code of our study is available at github.com/KChen-lab/MEDALT.

Short Abstract: Background
The Il17-producing T helper subset (Th17) is a well-established driver of multiple autoimmune diseases. At homeostasis most Th17 cells are found in the lamina propria of the intestine. Interestingly, recent studies strongly suggest an involvement of the intestinal Th17 population in extra-intestinal autoimmune diseases. However, the mechanism in which intestinal Th17 cells can drive autoimmune tissue inflammation in peripheral sites remains to be elucidated.
Results
We performed single-cell RNA sequencing of splenic, lymph node-derived, intestinal and central nervous system-infiltrating Th17 cells at homeostasis and during experimental autoimmune encephalomyelitis (EAE). Accordingly, we built an analysis pipeline for a comprehensive characterization of the Th17 population. In particular, we identified tissue-specific Th17 signatures and characterized the heterogeneity of Th17 cells in each tissue at homeostasis. We also described the changes in transcriptional profiles upon EAE in different tissues. Furthermore, we used single-cell TCR sequencing with cell hashing to characterize Th17 cell within-tissue clonal expansion as well as across-tissue migration at homeostasis and during EAE.
Conclusions
Our study provides an extensive single-cell survey of tissue Th17 cells during homeostasis and EAE. Th17 cells show transcriptomic heterogeneity within and across tissues, and there are evidences for cell migration to peripheral sites during EAE.

Short Abstract: Computational reconstruction of clonal evolution in cancers has become a crucial tool for understanding how tumors initiate and progress and how this process varies across patients. The field still struggles, however, with special challenges of applying phylogenetic methods to cancers, such as the prevalence and importance of copy number alteration (CNA) and structural variation (SV) events in tumor evolution, which are diffcult to profile accurately by prevailing sequencing methods in such a way that subsequent reconstruction by phylogenetic inference algorithms is accurate. In the present work, we develop computational methods to combine sequencing with multiplex interphase fluorescence in situ hybridization (miFISH) to exploit the complementary advantages of each technology in inferring accurate models of clonal CNA evolution accounting for both focal changes and aneuploidy at whole-genome scales. We demonstrate on simulated data that incorporation of FISH data substantially improves accurate inference of focal CNA and ploidy changes in clonal evolution from deconvolving bulk sequence data. Analysis of real glioblastoma data for which FISH, bulk sequence, and single cell sequence are all available confirms the power of FISH to enhance accurate reconstruction of clonal copy number evolution in conjunction with bulk and optionally single-cell sequence data.

Short Abstract: see long abstract

Short Abstract: SARS-CoV-2 is a single-stranded, 30 kb RNA virus from the coronavirus family that is currently causing a worldwide pandemic, with over 3 million confirmed cases and 212k deaths as of April 2020. Complete and accurate sequencing of the viral genome allows epidemiological tracing and discovery of mutations that may be important for antiviral and vaccine research.



PacBio’s SMRT Sequencing uses circular consensus sequencing that can generate long, highly accurate (HiFi) reads for SARS-Cov-2 genomes. We generated several in-house datasets using amplicon-based approaches with insert sizes ranging from 400 bp to several kb long, as well as hybridization-based approaches that can capture fragments of 5 kb or longer.



We tested several variant calling methods including PacBio’s own MinorVariant (originally developed for HIV quasi-species detection) and LoFreq. We show that regardless of amplicon size, reference-guided mapping approaches can call SNPs accurately at as low as 10-fold coverage under the assumption of a single, consensus genome. Higher coverages are required for identifying minor variants spiked in at 1%, 10%, and 50% frequency. Further, we compare the results on existing assemblers and show that even coverage and longer sequences are desired for non-guided, de novo assembly.

Short Abstract: The ability to efficiently query genomic variants from thousands of samples is critical to achieving the full potential of many medical and scientific applications such as personalized medicine. Performing variant queries based on positions in the reference or sample sequences is at the core of these applications. Efficiently supporting variant queries across thousands of samples is computationally challenging. Most solutions only support queries based on the reference sequence and the ones that support queries across multiple samples do not scale to data containing more than a few thousand samples. We present VariantStore, a system for efficiently indexing and querying genomic variants and their sequences in either the reference or sample-specific positions. We show the scalability of VariantStore by indexing genomic variants from the TCGA-BRCA project containing 8640 samples and 5M variants in 4 Hrs and the 1000 genomes project containing 2500 samples and 924M variants in 3 Hrs. Querying for variants in a gene takes between 0.002 – 3 seconds using memory only 10% of the size of the full representation.

Short Abstract: With an increasing number of patient-derived xenograft (PDX) models being created and subsequently sequenced to study tumor heterogeneity and to guide therapy decisions, there is a need for methods to separate reads originating from the graft (human) tumor and reads originating from the host species' (mouse) surrounding tissue.
Two kinds of methods are in use:
On the one hand, alignment-based tools like XenofilteR require that reads are mapped and aligned (by an external mapper/aligner) to the host and graft genomes separately first; the tool itself then processes the resulting alignments and quality metrics (typically BAM files) to assign each read or read pair.
On the other hand, alignment-free tools like xenome work directly on the raw read data (typically FASTQ files).
Recent studiescompare different approaches and tools, with partially conflicting results.

We here argue that a carefully engineered alignment-free approach using three-way bucketed Cuckoo hashing offers high read classification accuracy, fast running times and a reasonable memory footprint.
Software (`'xengsort'') is available (MIT license) at gitlab.com/genomeinformatics/xengsort.