Biology subjects

Kannan, S.

Publications and source records attributed to Kannan, S..

53 records · Page 3Linked to original sources

bioRxiv2020

Microglia-targeted dendrimer-2PMPA therapy robustly inhibits GCPII and improves cognition in a mouse model of multiple sclerosis

Roughly half of all individuals with multiple sclerosis (MS) experience cognitive impairment, but there are no approved treatments that target this aspect of the disease. Recent studies link reduced brain N-acetylaspartylglutamate (NAAG) levels to impaired cognition in various neurological diseases, including MS. NAAG levels are regulated by glutamate carboxypeptidase II (GCPII), which hydrolyzes the neuropeptide to N-acetyl-aspartate (NAA) and glutamate. Although several GCPII inhibitors, such as 2-(phosphonomethyl)-pentanedioic acid (2-PMPA), elevate brain NAAG levels and restore cognitive function in preclinical studies when given at high systemic doses or via direct brain injection, no GCPII inhibitors are clinically available due to poor bioavailability and limited brain penetration. Systemic hydroxyl dendrimers (~4 nm) have been successfully used to enhance brain delivery of drugs selectively to activated glia. We recently discovered that GCPII is highly upregulated in activated microglia after brain injury. To determine if dendrimer conjugation could enhance the brain delivery of GCPII inhibitors, specifically in the context of MS, we attached 2-PMPA to hydroxyl polyamidoamicne (PAMAM) dendrimers (D-2PMPA) using a highly efficient click chemistry approach. Targeted uptake of D-2PMPA into activated glia was subsequently confirmed in glial cultures where it showed robust anti-inflammatory activity, including an elevation in TGF{beta} and a reduction in TNF. Given these positive effects, D-2PMPA (20mg/kg) or vehicle dendrimer were dosed twice weekly to experimental autoimmune encephalomyelitis (EAE)-immunized mice starting at disease onset (therapeutic paradigm). D-2PMPA significantly improved cognition in EAE as assessed by Barnes maze performance, even though physical severity was not impacted. Glial target engagement was confirmed, as CD11b+ enriched cells isolated from hippocampi in D-2PMPA-treated mice exhibited almost complete loss of GCPII activity. These data demonstrate the utility of hydroxyl dendrimers to enhance brain penetration and support the development of D-2PMPA to treat cognitive impairment in MS. FundingThis work was funded by the National Multiple Sclerosis Society (RG-1507-05403 to BSS), the National Institute of Health NINDS (R01NS093416 to SK, RM and BSS), and Ashvattha Therapeutics. We would also like to acknowledge support for the statistical analysis from the National Center for Research Resources and NIH NCATS (1UL1TR001079). HighlightsO_LIThe GCPII inhibitor 2-PMPA was conjugated to hydroxyl PAMAM dendrimers (D-2PMPA) C_LIO_LID-2PMPA targeted activated glia in culture and displayed anti-inflammatory activity C_LIO_LIWhen dosed systemically to EAE mice, D-2PMPA inhibited CD11b+ cell GCPII activity C_LIO_LIWhen dosed systemically to EAE mice, D-2PMPA improved cognitive function C_LI

neuroscience↗

bioRxiv2020

UNCURL-App: Interactive Database-Driven Analysis of Single Cell RNA Sequencing Data

MotivationAnalysis of single cell RNA sequencing (scRNA-seq) datasets is a complex and time-consuming process, requiring both biological knowledge and technical skill. With the rapid growth in scRNA-seq datasets, there is a need to simplify and systematize this process. ResultsWe introduce UNCURL-App, an online GUI-based interactive tool which integrates the scRNA-seq analysis pipeline with prior knowledge. UNCURL-App introduces two key innovations: First, cell type databases are integrated directly with the rest of the analysis process, allowing the user to identify potential cell types or pathways directly from the data. Second, tools for interactive re-analysis allow the user to create, merge, or delete clusters. In addition, UNCURL-App integrates multiple stages of the data analysis pipeline into a single interface, including dimensionality reduction, clustering, differential expression, and cell type identification. AvailabilityThe web tool is available at https://uncurl.cs.washington.edu/. Source code is available at https://github.com/yjzhang/uncurl_app Contactgseelig@uw.edu

bioinformatics↗

bioRxiv2020

Transcriptomic entropy quantifies cardiomyocyte maturation at single cell level

The immaturity of pluripotent stem cell (PSC)-derived tissues has emerged as a universal problem for their biomedical applications. While efforts have been made to generate adult-like cells from PSCs, direct benchmarking of PSC-derived tissues against in vivo development has not been established. Thus, maturation status is often assessed on an ad-hoc basis. Single cell RNA-sequencing (scRNA-seq) offers a promising solution, though cross-study comparison is limited by dataset-specific batch effects. Here, we developed a novel approach to quantify PSC-derived cardiomyocyte (CM) maturation through transcriptomic entropy. Transcriptomic entropy is robust across datasets regardless of differences in isolation protocols, library preparation, and other potential batch effects. With this new model, we analyzed over 45 scRNA-seq datasets and over 52,000 CMs, and established a cross-study, cross-species CM maturation reference. This reference enabled us to directly compare PSC-CMs with the in vivo developmental trajectory and thereby to quantify PSC-CM maturation status. We further found that our entropy-based approach can be used for other cell types, including pancreatic beta cells and hepatocytes. Our study presents a biologically relevant and interpretable metric for quantifying PSC-derived tissue maturation, and is extensible to numerous tissue engineering contexts. Significance StatementThere is significant interest in generating mature cardiomyocytes from pluripotent stem cells. However, there are currently few effective metrics to quantify the maturation status of a single cardiomyocyte. We developed a new metric for measuring cardiomyocyte maturation using single cell RNA-sequencing data. This metric, called entropy score, uses the gene distribution to estimate maturation at the single cell level. Entropy score enables comparing pluripotent stem cell-derived cardiomyocytes directly against endogenously-isolated cardiomyocytes. Thus, entropy score can better assist in development of approaches to improve the maturation of pluripotent stem cell-derived cardiomyocytes.

bioinformatics↗

bioRxiv2020

Critical role for cold shock protein YB-1 in cytokinesis

High levels of the cold shock protein Y-box-binding protein-1, YB-1, are tightly correlated with increased cell proliferation and cancer progression. However, the precise mechanism by which YB-1 regulates proliferation is unknown. Here, we found that YB-1 depletion in several cell lines resulted in cytokinesis failure, multinucleation and an increase in G1 transit time. Rescue experiments indicated that YB-1 was required for completion of cytokinesis. Using confocal imaging of cells undergoing cytokinesis both in vitro and in zebrafish embryos, we found that YB-1 was critical for microtubule organization during cytokinesis. Using mass spectrometry we identified multiple novel phosphorylation sites on YB-1. We show that phosphorylation of YB-1 at multiple serine residues was essential for its function during cytokinesis. Using atomistic modelling we show how multiple phosphorylations alter YB-1 conformation, allowing it to interact with protein partners. Our results establish phosphorylated YB-1 as a critical regulator of cytokinesis, defining for the first time precisely how YB-1 regulates cell division. SUMMARYY-box-binding protein-1, YB-1, is essential for cell division, but it is not clear how it functions. Using live imaging and confocal microscopy we show that YB-1 functions only in the last step of division, specifically being required to initiate cytokinesis.

cell biology↗

bioRxiv2020

PGC1/PPAR Drive Cardiomyocyte Maturation through Regulation of Yap1 and SF3B2

Cardiomyocytes undergo significant levels of structural and functional changes after birth--fundamental processes essential for the heart to produce the volume and contractility to pump blood to the growing body. However, due to the challenges in isolating single postnatal/adult myocytes, how individual newborn cardiomyocytes acquire multiple aspects of mature phenotypes remains poorly understood. Here we implemented large-particle sorting and analyzed single myocytes from neonatal to adult hearts. Early myocytes exhibited a wide-ranging transcriptomic and size heterogeneity, maintained until adulthood with a continuous transcriptomic shift. Gene regulatory network analysis followed by mosaic gene deletion revealed that peroxisome proliferator-activated receptor coactivator-1 signaling--activated in vivo but inactive in pluripotent stem cell-derived cardiomyocytes--mediates the shift. The signaling regulated key aspects of cardiomyocyte maturation simultaneously through previously unrecognized regulators, including Yap1 and SF3B2. Our study provides a single-cell roadmap of heterogeneous transitions coupled to cellular features and unveils a multifaceted regulator controlling cardiomyocyte maturation. Significance StatementHow the individual single myocytes achieve full maturity remains a black box, largely due to the challenges with the isolation of single mature myocytes. Understanding this process is particularly important as the immaturity and early developmental arrest of pluripotent stem cell-derived myocytes has emerged a major concern in the field. Here we present the first study of high-quality single-cell transcriptomic analysis of cardiac muscle cells from neonatal to adult hearts. We identify a central transcription factor and its novel targets that control key aspects of myocyte maturation, including cellular hypertrophy, contractility, and mitochondrial activity.

developmental biology↗

bioRxiv2020

Proteogenomic single cell analysis of skeletal muscle myocytes

Skeletal muscle myocytes have evolved into slow and fast-twitch types. These types are functionally distinct as a result of differential gene and protein expression. However, an understanding of the complexity of gene and protein variation between myofibers is unknown. We performed deep, whole cell, single cell RNA-seq on intact and fragments of skeletal myocytes from the mouse flexor digitorum brevis muscle. We compared the genomic expression data of 171 of these cells with two human proteomic datasets. The first was a spatial proteomics survey of mosaic patterns of protein expression utilizing the Human Protein Atlas (HPA) and the HPASubC tool. The second was a mass-spectrometry (MS) derived proteomic dataset of single human muscle fibers. Immunohistochemistry and RNA-ISH were used to understand variable expression. scRNA-seq identified three distinct clusters of myocytes (a slow/fast 2A cluster and two fast 2X clusters). Utilizing 1,605 mosaic patterned proteins from visual proteomics, and 596 differentially expressed proteins by MS methods, we explore this fast 2X division. Only 36 genes/proteins had variable expression across all three studies, of which nine are newly described as variable between fast/slow twitch myofibers. An additional 414 genes/proteins were identified as variable by two methods. Immunohistochemistry and RNA-ISH generally validated variable expression across methods presumably due to species-related differences. In this first integrated proteogenomic analysis of mature skeletal muscle myocytes we confirm the main fiber types and greatly expand the known repertoire of twitch-type specific genes/proteins. We also demonstrate the importance of integrating genomic and proteomic datasets.

genomics↗

bioRxiv2019

QAlign: Aligning nanopore reads accurately using current-level modeling

MotivationEfficient and accurate alignment of DNA / RNA sequence reads to each other or to a reference genome / transcriptome is an important problem in genomic analysis. Nanopore sequencing has emerged as a major sequencing technology and many long-read aligners have been designed for aligning nanopore reads. However, the high error rate makes accurate and efficient alignment difficult. Utilizing the noise and error characteristics inherent in the sequencing process properly can play a vital role in constructing a robust aligner. In this paper, we design QAlign, a pre-processor that can be used with any long-read aligner for aligning long reads to a genome / transcriptome or to other long reads. The key idea in QAlign is to convert the nucleotide reads into discretized current levels that capture the error modes of the nanopore sequencer before running it through a sequence aligner. ResultsWe show that QAlign is able to improve alignment rates from around 80% up to 90% with nanopore reads when aligning to the genome. We also show that QAlign improves the average overlap quality by 9.2%, 2.5% and 10.8% in three real datasets for read-to-read alignment. Read-to-transcriptome alignment rates are improved from 51.6% to 75.4% and 82.6% to 90% in two real datasets. Availabilityhttps://github.com/joshidhaivat/QAlign.git

bioinformatics↗

bioRxiv2019

Listing candidate diagnostic markers and exploring the molecular basis of human spermatogenesis and a type of male infertility (Non-Obstructive Azoospermia), with transcriptomics via the next generation sequencing methods

Studying the molecular basis of Non-Obstructive Azoospermia (NOA), a type of male infertility with failed spermatogenesis at various stages, can also help in exploring molecular basis of human spermatogenesis and possibly pave way to identify new targets for male contraceptive development. Hence, we initiated a functional genomics study by applying RNA-seq. Testicular biopsies collected from donors with Non-Obstructive Azoospermia (NOA), Obstructive Azoospermia (OA), Congenital Bilateral Absence of the Vas Deferens (CBAVD), and Varicocele (VA) conditions. Strong association of 100+ genes with human spermatogenesis and NOA has been detected via NGS-based transcriptomic analysis. In addition, 20 RNA molecules have been short-listed for potential diagnostic applications (non-obstructive azoospermia vs. obstructive azoospermia, varicocele or normal). A hierarchical list of several genes and alternatively spliced mRNAs, transcribed differentially in NOA, is reported - based on a strength of association. Such association with NOA, spermatogenesis or both is a new finding for many genes as revealed by a comparison with a newly prepared comprehensive list of genes having such association with human spermatogenesis/NOA. Many top-ranking genes involved in viral gene expression were up-regulated in testes from NOA-patients, while those associated with an antiviral mechanism were down-regulated. A tangential finding: while most well-established control mRNAs did not qualify, two new ones worked best in RT-qPCR experiments. Needle-aspiration of testicular biopsies, followed by the use of short-listed promising candidate biomarkers (i.e., 16 mRNA & 4 chimeric transcripts) and control mRNAs in RT-qPCR-based diagnostic assays, may help to avoid open surgeries in future.

developmental biology↗

bioRxiv2019

Macrocyclization of an all-D linear peptide improves target affinity and imparts cellular activity: A novel stapled α-helical peptide modality

Peptide-based inhibitors hold great potential for targeted modulation of intracellular protein-protein interactions (PPIs) by leveraging vast chemical space relative to primary structure via sequence diversity as well as conformationally through varying secondary and tertiary structures. However, the development of peptide therapeutics has been hindered because of their limited conformational stability, proteolytic sensitivity and cell permeability. Several contemporary peptide design strategies address these issues to varying degrees. Strategic macrocyclization through optimally placed chemical braces such as olefinic hydrocarbon crosslinks, commonly referred to as staples, may address these issues by i) restricting conformational freedom to improve target affinities, ii) improving proteolytic resistance, and iii) enhancing cell permeability. Conversely, molecules constructed entirely from D-amino acids are hyper-resistant to proteolytic cleavage, but generally lack conformational stability and membrane permeability. Since neither approach is a complete solution, we have combined these strategies to identify the first examples of all-D -helical stapled and stitched peptides. As a template, we used a recently reported all D-linear peptide that is a potent inhibitor of the p53-Mdm2 interaction, but is devoid of cellular activity. To design both stapled and stitched all-D-peptide analogues, we used computational modelling to predict optimal staple placement. The resultant novel macrocyclic all D-peptide was determined to exhibit increased -helicity, improved target binding, complete proteolytic stability and, most notably, cellular activity.

biochemistry↗

bioRxiv2019

The hunger games: sensing host arginine is essential for Leishmania parasite virulence

Arginine homeostasis in lysosomes is critical for growth and metabolism of mammalian cells. They employ a specific sensor (SLC38A9) that monitors intra-lysosome arginine sufficiency and subsequently up-regulates cellular mTORC1 activity. Lysosomes of macrophages (phagolysosomes) are the niche where the parasitic protozoan Leishmania resides and causes important human disease. Several years ago, we discovered that upon arginine starvation, cultured Leishmania parasites promptly activate a MAPK2-mediated Arginine Deprivation Response (ADR) pathway, resulting in up-regulation of the Leishmania arginine transporter (AAP3), as well as a small group of other transporters. Significantly, ADR is also activated during macrophage infection, implying that the intracellular parasite actively depletes arginine within the host phagolysosome, likely to prevent mTORC1 activation and enhance intracellular development. We hypothesize that ADR-mediated up-regulation of AAP3 activity is necessary to withstand the resultant arginine starvation. Both copies of the AAP3 genes are located (in tandem) on a tetrasomic chromosome (chr31), but only one (AAP3.2) is responsive to arginine deprivation. CRISPR/Cas9-mediated disruption of the AAP3 locus yielded mutants that retain a basal level of arginine transport (mediated by AAP3.1), but lack a functional copy of AAP3.2 and are therefore not responsive to arginine starvation. While these mutants grow normally in culture as promastigotes, they were impaired in their ability to develop inside THP1 macrophages grown under physiological concentrations of arginine (0.1 mM). However, flooding the macrophage growth medium with arginine (1.5 mM) restored parasite infectivity and intracellular growth to that of wild type. The results indicate that inside the host macrophage, Leishmania must overcome the arginine \"Hunger Games\" by up-regulating transport of arginine via the ADR. Furthermore, the AAP3.2 mutants were ~70-80% less virulent in Balb/C mice, showing, for the first time, that the ability to monitor and respond to changes in host metabolite levels is essential for pathogenesis.

cell biology↗

bioRxiv2019

Vaccine-induced ICOS+CD38+ cTfh are sensitive biosensors of age-related changes in inflammatory pathways

Humoral immune responses are dysregulated with aging but details remain incompletely understood. In particular, little is known about the effects of aging on T follicular helper (Tfh) CD4 cells, the subset that provides critical help to B cells for effective humoral immunity. We previously demonstrated that influenza vaccination increases a circulating Tfh (cTfh) subset that expresses ICOS and CD38, contains influenza-specific memory cells, and is correlated with antibody responses. To directly study the effects of aging on the cTfh response, we performed transcriptional profiling and cellular analysis before and after influenza vaccination in young and elderly adults. Several key differences in cTfh responses were revealed in the elderly. First, whole blood transcriptional profiling defined cross-validated genesets of youth versus aging and these genesets were, compared to other T cells, preferentially enriched in ICOS+CD38+ cTfh from young and elderly subjects, respectively, following vaccination. Second, vaccine-induced ICOS+CD38+ cTfh from the elderly were enriched for transcriptional signatures of inflammation including TNF-NFkB pathway activation. Indeed, we reveal a paradoxical positive effect of TNF signaling on Tfh providing help to B cells linked to survival circuits that may explain detrimental effects of TNF blockade on vaccine responses. Finally, vaccine-induced ICOS+CD38+ cTfh displayed strong enrichment for signatures of underlying age-associated biological changes. Thus, these data reveal key biological changes in cTfh during aging and also demonstrate the sensitivity of vaccine-induced cTfh to underlying changes in host physiology. This latter observation suggests that vaccine-induced cTfh could function as sensitive biosensors of underlying inflammatory and/or overall immune health.\n\nOne sentence summaryTranscriptional profiling of vaccine-induced circulating T follicular helper cell responding to influenza vaccination reveals age-associated effects on Tfh such as alterations in TNF-NFkB signaling.

immunology↗

bioRxiv2019

Large particle fluorescence-activated cell sorting enables high quality single cell RNA-sequencing and functional analysis of adult cardiomyocytes

RationaleSingle cell RNA sequencing (scRNA-seq) has emerged as a powerful tool to profile the transcriptome at single cell resolution, enabling comprehensive analysis of cellular trajectories and heterogeneity during development and disease. However, the use of scRNA-seq remains limited in cardiac pathology owing to technical difficulties associated with the isolation of single adult cardiomyocytes (CMs).\n\nObjectiveWe investigated the capability of large-particle fluorescence-activated cell sorting (LP-FACS) for isolation of viable single adult CMs.\n\nMethods and ResultsWe found that LP-FACS readily outperforms conventional FACS for isolation of struturally competent CMs, including large CMs. Additionally, LP-FACS enables isolation of fluorescent CMs from mosaic models. Importantly, the sorted CMs allow generation of high-quality scRNA-seq libraries. Unlike CMs isolated via previously utilized fluidic or manual methods, LP-FAC-isolated CMs generate libraries exhibiting normal levels of mitochondrial transcripts. Moreover, LP-FACS isolated CMs remain functionally competent and can be studied for contractile properties.\n\nConclusionsOur study enables high quality dissection of adult CM biology at single-cell resolution.

cell biology↗

bioRxiv2019

De-risking drug discovery of intracellular targeting peptides: screening strategies to eliminate false-positive hits

Discovery of false-positive target binding, due to assay interference or aggregation, presents a significant problem for drug discovery programs. These issues may often be unrealized and could lead researchers astray if not subject to independent verification of reproducibility and/or on-target mechanism of action. Although well-documented for small molecules, this issue has not been widely explored for peptide modality. As a case study, we demonstrate that two purported KRas inhibitors, stapled peptide SAH-SOS1A and macrocyclic peptide cyclorasin 9A5, exemplify false-positive molecules - both in terms of their sub-micromolar KRas binding affinities and their on-target cellular activities. We observed that the apparent binding of fluorescein-labeled SAH-SOS1A given by a fluorescence polarization assay is sensitive to detergent. False-positive readouts can arise from peptide adsorption to the surface of microplates. Hence, we used surface plasmon resonance and isothermal titration calorimetry to unambiguously show that both SAH-SOS1A and cyclorasin 9A5 are non-binders for KRas. Thermal shift assay and hydrogen-deuterium exchange mass spectrometry further demonstrate that both peptides destabilize KRas and induce unfolding of the protein. Furthermore, both peptides caused significant release of intracellular lactate dehydrogenase, suggesting that membrane rupture rather than on-target activity is accountable for their reported cytotoxicity. Finally, both peptides exhibited off-target activities by inhibiting the proliferation of U-2 OS and A549 cells, despite their independency of the KRas signaling pathway. Our findings demonstrate the critical need to employ orthogonal binding assays and cellular counter-screens to de-risk false-positive molecules. More rigorous workflows should lead to improved data and help obviate inadvertent scientific conclusions.\n\nSignificance statementFalse positive molecule hits occur frequently in high-throughput screens and can contaminate the scientific literature. This has become an increasingly serious issue in small molecule drug discovery and chemical probe development and it is not surprising that peptides may be similarly prone to assay interference. Using KRas as a target and two known macrocyclic peptide inhibitors as a case study, we clearly show that reporter-free biophysical assays and cellular counter-screens offer the solution to detect and de-risk the potential of false-positive compounds. We further discuss the advantages, limitations and overall strategic importance of such methods.

biochemistry↗

bioRxiv2019

Retromer repletion with AAV9-VPS35 restores endosomal function in the mouse hippocampus

Retromer has emerged as a master conductor of endosomal trafficking, and VPS35 and other retromer-related proteins are found to be deficient in late-onset Alzheimers disease (AD). Depleting VPS35 in neurons impairs retromer function, affecting for example the trafficking of the amyloid-precursor protein (APP) and the glutamate receptor GluA1. Whether VPS35 repletion, after chronic in vivo depletion, can rescue these impairments remains unknown. Here we set out to address this question by using a viral vector approach for VPS35 repletion. First, we completed a series of studies using neuronal cultures in order to optimize AAV9-VPS35 delivery, and to understand how exogenous VPS35 expression affects its endogenous levels as well as its binding to other retromer proteins. Next, we completed a series of studies in wildtype mice to determine the optimum protocol for in vivo delivery of AAV9-VPS35 to the hippocampus. We relied on this information to deliver AAV9-VPS35 to the hippocampus of mice genetically engineered to have chronic, neuronal-selective, VPS35 depletion. VPS35 repletion in the hippocampus was found to normalize APP cleavage and to restore glutamate receptor levels. Unexpectedly, chronic VPS35 depletion in neurons caused glial activation, similar to the pattern observed in AD, which was also partially normalized by VPS35 repletion. Taken together, these studies strengthen the mechanistic link between retromer and AD, and have therapeutic implications.

neuroscience↗

bioRxiv2018

Towards inferring causal gene regulatory networks from single cell expression measurements

Single-cell transcriptome sequencing now routinely samples thousands of cells, potentially providing enough data to reconstruct causal gene regulatory networks from observational data. Here, we present Scribe, a toolkit for detecting and visualizing causal regulatory interactions between genes and explore the potential for single-cell experiments to power network reconstruction. Scribe employs Restricted Directed Information to determine causality by estimating the strength of information transferred from a potential regulator to its downstream target. We apply Scribe and other leading approaches for causal network reconstruction to several types of single-cell measurements and show that there is a dramatic drop in performance for \"pseudotime\" ordered single-cell data compared to true time series data. We demonstrate that performing causal inference requires temporal coupling between measurements. We show that methods such as \"RNA velocity\" restore some degree of coupling through an analysis of chromaffin cell fate commitment. These analyses therefore highlight an important shortcoming in experimental and computational methods for analyzing gene regulation at single-cell resolution and point the way towards overcoming it.

genomics↗

bioRxiv2018

Segmentation of Glomeruli Within Trichrome Images Using Deep Learning

IntroductionThe number of glomeruli and glomerulosclerosis evaluated on kidney biopsy slides constitute as standard components of a renal pathology report. Prevailing methods for glomerular assessment remain manual, labor intensive and non-standardized. We developed a deep learning framework to accurately identify and segment glomeruli from digitized images of human kidney biopsies. MethodsTrichrome-stained images (n=275) from renal biopsies of 171 chronic kidney disease patients treated at the Boston Medical Center from 2009-12 were analyzed. A sliding window operation was defined to crop each original image to smaller images. Each cropped image was then evaluated by three experts into three categories: (a) No glomerulus, (b) Normal or partially sclerosed glomerulus and (c) Globally sclerosed glomerulus. This led to identification of 751 unique images representing non{-}glomerular regions, 611 images with either normal or partially sclerosed (NPS) glomeruli and 134 images with globally sclerosed (GS) glomeruli. A convolutional neural network (CNN) was trained with cropped images as inputs and corresponding labels as output. Using this model, an image processing routine was developed to scan the test data images to segment the GS glomeruli. ResultsThe CNN model was able to accurately discriminate non-glomerular images from NPS and GS images (Performance on test data - Accuracy: 92.67{+/-}2.02% and Kappa: 0.8681{+/-}0.0392). The segmentation model that was based on the CNN multi-label classifier accurately marked the GS glomeruli on the test data (Matthews correlation coefficient = 0.628). ConclusionThis work demonstrates the power of deep learning for assessing complex histologic structures from digitized human kidney biopsies.

pathology↗

bioRxiv2017

Prior Knowledge And Sampling Model Informed Learning With Single Cell RNA-Seq Data

MotivationSingle cell RNA-seq (scRNA-seq) data contains a wealth of information which has to be inferred computationally from the observed sequencing reads. As the ability to sequence more cells improves rapidly, existing computational tools suffer from three problems. (1) The decreased reads-per-cell implies a highly sparse sample of the true cellular transcriptome. (2) Many tools simply cannot handle the size of the resulting datasets. (3) Prior biological knowledge such as bulk RNA-seq information of certain cell types or qualitative marker information is not taken into account. Here we present UNCURL, a preprocessing framework based on non-negative matrix factorization for scRNA-seq data, that is able to handle varying sampling distributions, scales to very large cell numbers and can incorporate prior knowledge.\n\nResultsWe find that preprocessing using UNCURL consistently improves performance of commonly used scRNA-seq tools for clustering, visualization, and lineage estimation, both in the absence and presence of prior knowledge. Finally we demonstrate that UNCURL is extremely scalable and parallelizable, and runs faster than other methods on a scRNA-seq dataset containing 1.3 million cells.\n\nAvailabilitySource code is available at https://github.com/yjzhang/uncurl_python\n\nContactksreeram@uw.edu, gseelig@uw.edu

bioinformatics↗