Search bioRxiv⌕ Search

Biology subjects

Lai, Z.

Publications and source records attributed to Lai, Z..

At least 37 records · Page 2Linked to original sources
bioRxiv2023

Differential response of Senna occidentalis L. to arsenic and cadmium contaminated soil

We investigated the phytoremediation potential of Senna occidentalis L., a pantropical plant that has been associated with tolerance to heavy metal-contaminated soils around mining sites. Seedlings of S. occidentalis were exposed to cadmium chloride (CdCl2) and sodium arsenate (Na3AsO4) at concentrations of 200, 300, and 400 mg L-1 under greenhouse conditions. Heavy metal tolerance was assessed by comparing biomass and stress indicators such as chlorophyll, proline, and hydrogen peroxide content. Arsenic treatment had more toxic effects than cadmium on Senna physiology. Regardless of concentration of arsenic applied, the biomass decreased by 50% as compared to control and cadmium-treated plants. Chlorophyll content decreased with exposure to both heavy metals. Higher concentration of Cd and As (400 mg L-1) resulted in 50% reduction in chlorophyll content. Proline and hydrogen peroxide levels were higher in arsenic-treated plants compared to controls and cadmium-treated plants, indicating an enhanced stress response when exposed to arsenic. When heavy metal content was measured, there was a significant accumulation of arsenic in the leaves, stems, and roots, indicating that arsenic in these tissues was responsible for the profound changes in biomass, proline, and hydrogen peroxide content. In contrast, although significant, there was less cadmium uptake by Senna and tolerance can be seen, which was reflected by normal biomass, proline, and hydrogen peroxide levels. High translocation of metals from soil into roots and low translocation from root to shoot tissues suggests the potential for S. occidentalis to be used for phytostabilization of arsenic- and cadmium-contaminated soils.

plant biology↗
bioRxiv2023

Spatial-Live: A lightweight and versatile tool for single cell spatial-omics data visualization

Single cell spatial-omics data visualization plays a pivotal role in unraveling the intricate spatial organization and heterogeneity of cellular systems. Although various software tools and packages have been developed for this purpose, challenges persist in terms of user-friendly accessibility, data integration, and interactivity. In this study, we introduce Spatial-Live, a lightweight and versatile viewer tool designed for flexible single-cell spatial-omics data visualization. Spatial-Live overcomes the fundamental limitations of two-dimensional (2D) orthographic modes by employing a layer-stacking strategy, enabling efficient rendering of diverse data types with interactive features, and enhancing visualization with richer information in a unified three-dimensional (3D) space.

bioinformatics↗
bioRxiv2023

Transcriptional Profiles of Non-neuronal and Immune Cells in Mouse Trigeminal Ganglia

Non-neuronal cells constitute 90-95% of sensory ganglia. These cells play critical roles in modulation of nociceptive signal transmissions by sensory neurons. Accordingly, the aim of this review-study was to identify, profile and summarize TG non-neuronal cell types in naive male mice using published and our own data generated by single-cell RNA sequencing (scRNA-seq), flow cytometry (FC) and immunohistochemistry (IHC). TG contains 5 types of non-neuronal cells: glial, fibroblasts, smooth muscle, endothelial and immune cells. There is agreement among publications for glial, fibroblasts, smooth muscle and endothelial cells. Based on gene profiles, glial cells were classified as Schwann cells and satellite glial cells (SGC). Mpz had dominant expression in Schwann cells, and Fabp7 is specific for SCG. Two types of Col1a2+ fibroblasts located throughout TG were distinguished using gene profiles. TG smooth muscle and endothelial cells representing blood vessels were detected with well recognized markers. Our study split reported single TG immune cell group into 3 types of macrophages and 4 types of neutrophils. Macrophages were located among neuronal bodies and nerve fibers, and were sub-grouped by unique transcriptomic profiles and using Ccr2, Cx3cr1 and Iba1 as markers. S100a8+ neutrophils were located in dura surrounding TG and were sub-grouped by clustering and expressions of Csf3r, Ly6G, Ngp, Elane and Mpo. Overall, generated and summarized here dataset on non-neuronal TG cells could provide essential and fundamental information for studies on cell plasticity, interactomic network between neurons and non-neuronal cells and function during variety of pain conditions in the head and neck region.

neuroscience↗
bioRxiv2023

Requirement of GrgA for Chlamydia infectious progeny production, optimal growth, and efficient plasmid maintenance

Chlamydia, an obligate intracellular bacterial pathogen, has a unique developmental cycle involving the differentiation of invading elementary bodies (EBs) to noninfectious reticulate bodies (RBs), replication of RBs, and redifferentiation of RBs into progeny EBs. Progression of this cycle is regulated by three sigma factors, which direct the RNA polymerase to their respective target gene promoters. We hypothesized that the Chlamydia-specific transcriptional regulator GrgA, previously shown to activate {sigma}66 and {sigma}28, plays an essential role in chlamydial development and growth. To test this hypothesis, we applied a novel genetic tool known as dependence on plasmid-mediated expression (DOPE) to create Chlamydia trachomatis with conditional GrgA-deficiency. We show that GrgA-deficient C. trachomatis RBs have a growth rate that is approximately half of the normal rate and fail to transition into progeny EBs. In addition, GrgA-deficient C. trachomatis fail to maintain its virulence plasmid. Results of RNA-seq analysis indicate that GrgA promotes RB growth by optimizing tRNA synthesis and expression of nutrient-acquisition genes, while it enables RB-to-EB conversion by facilitating the expression of a histone and outer membrane proteins required for EB morphogenesis. GrgA also regulates numerous other late genes required for host cell exit and subsequent EB invasion into host cells. Importantly, GrgA stimulates the expression of {sigma}54, the third and last sigma factor, and its activator AtoC, and thereby indirectly upregulating the expression of {sigma}54-dependent genes. In conclusion, our work demonstrates that GrgA is a master transcriptional regulator in Chlamydia and plays multiple essential roles in chlamydial pathogenicity. IMPORTANCEHallmarks of the developmental cycle of the obligate intracellular pathogenic bacterium Chlamydia are the primary differentiation of the infectious elementary body (EB) into the proliferative reticulate body (RB) and the secondary differentiation of RBs back into EBs. The mechanisms regulating these transitions remain unclear. In this report, we developed an effective novel strategy termed DOPE that allows for the knockdown of essential genes in Chlamydia. We demonstrate that GrgA, a Chlamydia-specific transcription factor, is essential for the secondary differentiation and optimal growth of RBs. We also show that GrgA, a chromosome-encoded regulatory protein, controls the maintenance of the chlamydial virulence plasmid. Transcriptomic analysis further indicates that GrgA functions as a critical regulator of all three sigma factors that recognize different promoter sets at developmental stages. The DOPE strategy outlined here should provide a valuable tool for future studies examining chlamydial growth, development, and pathogenicity.

microbiology↗
bioRxiv2023

Association of inflammation and tissue damage inducedbiological processes in masseter muscle with the resolution of chronic myalgia

Biological processes linked to intramuscular inflammation during myogenous temporomandibular disorder (TMDM) are largely unknown. We mimicked this inflammation by intra-masseteric muscle (MM) injections of complete Freunds adjuvant (CFA) or collagenase type 2 (Col), which emulates tissue damage. CFA triggered mechanical hypersensitivity at 1d post-injection was mainly linked to processes controlling chemotactic activity of monocytes and neutrophils. At 5d post-CFA, when hypersensitivity was resolved, there was minimal inflammation whereas tissue repair processes were pronounced. Low dose Col (0.2U) also produced acute orofacial hypersensitivity that was linked to tissue repair, but not inflammatory processes. High dose Col (10U) triggered prolonged orofacial hypersensitivity with inflammatory processes dominating at 1d post-injection. At pre-resolution time point (6d), tissue repair processes were underway and a significant increase in pro-inflammatory gene expressions compared to 1d post-injection were detected. RNA-seq and flow cytometry showed that immune processes in MM were linked to accumulation of macrophages, natural killer and natural killer T cells, dendritic cells and T-cells. Altogether, CFA and Col treatments induced different immune processes in MM. Importantly, orofacial hypersensitivity resolution was preceded with repairs of muscle cell and extracellular matrix, an elevation in immune system gene expression and accumulation of distinct immune cells in MM.

neuroscience↗
bioRxiv2023

A new tractable method for generating Human Alveolar Macrophage Like cells in vitro to study lung inflammatory processes and diseases

Alveolar macrophages (AMs) are unique lung resident cells that contact airborne pathogens and environmental particulates. The contribution of human AMs (HAM) to pulmonary diseases remains poorly understood due to difficulty in accessing them from human donors and their rapid phenotypic change during in vitro culture. Thus, there remains an unmet need for cost-effective methods for generating and/or differentiating primary cells into a HAM phenotype, particularly important for translational and clinical studies. We developed cell culture conditions that mimic the lung alveolar environment in humans using lung lipids, i.e., Infasurf (calfactant, natural bovine surfactant) and lung-associated cytokines (GM-CSF, TGF-{beta}, and IL-10) that facilitate the conversion of blood-obtained monocytes to an AM-Like (AML) phenotype and function in tissue culture. Similar to HAM, AML cells are particularly susceptible to both Mycobacterium tuberculosis and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections. This study reveals the importance of alveolar space components in the development and maintenance of HAM phenotype and function, and provides a readily accessible model to study HAM in infectious and inflammatory disease processes, as well as therapies and vaccines. IMPORTANCEMillions die annually from respiratory disorders. Lower respiratory track gas-exchanging alveoli maintain a precarious balance between fighting invaders and minimizing tissue damage. Key players herein are resident AMs. However, there are no easily accessible in vitro models of HAMs, presenting a huge scientific challenge. Here we present a novel model for generating AML cells based on differentiating blood monocytes in a defined lung component cocktail. This model is non-invasive, significantly less costly than performing a bronchoalveolar lavage, yields more AML cells than HAMs per donor and retains their phenotype in culture. We have applied this model to early studies of M. tuberculosis and SARS-CoV-2. This model will significantly advance respiratory biology research.

immunology↗
bioRxiv2023

Nicotinic acetylcholine receptor signaling maintains epithelial barrier integrity

Disruption of epithelial barriers is a common disease manifestation in chronic degenerative diseases of the airways, lung and intestine. Extensive human genetic studies have identified risk loci in such diseases, including in chronic obstructive pulmonary disease (COPD) and inflammatory bowel diseases (IBD). The genes associated with these loci have not fully been determined, and functional characterization of such genes requires extensive studies in model organisms. Here, we report the results of a screen in Drosophila melanogaster that allowed for rapid identification, validation and prioritization of COPD risk genes that were selected based on risk loci identified in human genome-wide association studies (GWAS) studies. Using intestinal barrier dysfunction in flies as a readout, our results validate the impact of candidate gene perturbations on epithelial barrier function in 56% of the cases, resulting in a prioritized target gene list. We further report the functional characterization in flies of one family of these genes, encoding for nicotinic acetylcholine receptor subunits (nAchR). We find that nAchR signaling in enterocytes of the fly gut promotes epithelial barrier function and epithelial homeostasis by regulating the production of the peritrophic matrix. Our findings identify COPD associated genes critical for epithelial barrier maintenance, and provide insight into the role of epithelial nAchR signaling for homeostasis.

genetics↗
bioRxiv2023

Understanding Painful versus Non-Painful Dental Pain in Male and Female Patients: A Transcriptomic Analysis of Human Biopsies

Dental pain from apical periodontitis is an infection induced-orofacial pain condition that presents with diversity in pain phenotypes among patients. While 60% of patients with a full-blown disease present with the hallmark symptom of mechanical allodynia, nearly 40% of patients experience no pain. Furthermore, a sexual dichotomy exists, with females exhibiting lower mechanical thresholds under basal and diseased states. Finally, the prevalence of post-treatment pain refractory to commonly used analgesics ranges from 7-19% (~2 million patients), which warrants a thorough investigation of the cellular changes occurring in different patient cohorts. We, therefore, conducted a transcriptomic assessment of periapical biopsies (peripheral diseased tissue) from patients with persistent apical periodontitis. Surgical biopsies from symptomatic male (SM), asymptomatic male (AM), symptomatic female (SF), and asymptomatic female (AF) patients were collected and processed for bulk RNA sequencing. Using strict selection criteria, our study found several unique differentially regulated genes (DEGs) between symptomatic and asymptomatic patients, as well as novel candidate genes between sexes within the same pain group. Specifically, we found the role of cells of the innate and adaptive immune system in mediating nociception in symptomatic patients and the role of genes involved in tissue homeostasis in potentially inhibiting nociception in asymptomatic patients. Furthermore, sex-related differences appear to be tightly regulated by macrophage activity, its secretome, and/or migration. Collectively, we present, for the first time, a comprehensive assessment of peripherally diseased human tissue after a microbial insult and shed important insights into the regulation of the trigeminal system in female and male patients. SummaryClinical diversity paves the way for translational research. We conducted a comprehensive transcriptomic assessment of peripherally diseased human tissue from patients with apical periodontitis to understand better clinical diversity observed in pain phenotypes between patients as well as perform in-depth analyses of sex-related dimorphism in patients with apical periodontitis.

genomics↗
bioRxiv2023

Pan-cancer Proteomics Analysis to Identify Tumor-Enriched and Highly Expressed Cell Surface Antigens as Potential Targets for Cancer Therapeutics

The National Cancer Institutes Clinical Proteomic Tumor Analysis Consortium (CPTAC) provides unique opportunities for cancer target discovery using protein expression. Proteomics data from CPTAC tumor types have been primarily generated using a multiplex tandem mass tag (TMT) approach, which is designed to provide protein quantification relative to reference samples. However, relative protein expression data is suboptimal for prioritization of targets within a tissue type, which requires additional reprocessing of the original proteomics data to derive absolute quantitation estimation. We evaluated the feasibility of using differential protein analysis coupled with intensity-based absolute quantification (iBAQ) to identify tumor-enriched and highly expressed cell surface antigens, employing tandem mass tag (TMT) proteomics data from CPTAC. Absolute quantification derived from TMT proteomics data was highly correlated with that of label-free proteomics data from the CPTAC colon adenocarcinoma cohort, which contains proteomics data measured by both approaches. We validated the TMT-iBAQ approach by comparing the iBAQ value to the receptor density value of HER2 and TROP2 measured by flow cytometry in about 30 selected breast and lung cancer cell lines from the Cancer Cell Line Encyclopedia. Collections of these tumor-enriched and highly expressed cell surface antigens could serve as a valuable resource for the development of cancer therapeutics, including antibody-drug conjugates and immunotherapeutic agents.

bioinformatics↗
bioRxiv2023

Plant selection and ecological microhabitat drive domestications of shrub-associated microbiomes in a revegetated shrub ecosystem

Shrubs are used for revegetation of degraded dryland ecosystem worldwide and could recruit large numbers of microbes from the soil; however, the plant-associated microbiome assembly and the effect of plant introduction on the soil microbiomes are not fully understood. We detected shrub-associated microbes from five ecological microhabitats, including the leaves, litter, roots, rhizosphere, and root zone, across four xeric shrub plantations (Artemisia ordosica, Caragana korshinskii, Hedysarum mongolicum, and Salix psammophila). To detect the patterns of shrub-associated microbiome assembly, 16S and ITS2 rRNA gene sequencing was performed. PERMANOVA and differential abundance analysis demonstrated that changes in the bacterial and fungal communities were more dependent on the microhabitats rather than on the plant species, with distinct niche differentiation. Moreover, source tracking and nestedness analysis showed that shrub-associated bacteria were primarily derived from bulk soils and slightly pruned in different microhabitats; however, a similar pattern was not found for fungi. Furthermore, the surrounding zone of roots was a hotpot for microbial recruitments of revegetated shrubs. Null model analysis indicated that homogeneous selection of determinism dominated the bacterial communities, whereas dispersal limitation and undominated process of stochasticity drove the assembly of fungal communities. Our findings indicate that ecological microhabitat of revegetated shrublands was the main predictor of the bacterial and fungal compositional variances. This study will help advance our understanding of the mechanism underlying the plant-soil microbiome feedbacks during the initial plant-establishment period in a dryland ecosystem. Further, this work provides theoretical reference for establishment and sustainable management of shrublands in drylands.

microbiology↗
bioRxiv2023

Sex-dependent Differences in the Genomic Profile of Lingual Sensory Neurons in Naive and Tongue-Tumor Bearing Mice

Mechanisms of sex-dependent orofacial pain are widely understudied. A significant gap in knowledge exists about comprehensive regulation of tissue-specific trigeminal sensory neurons in diseased state of both sexes. Using RNA sequencing of FACS sorted retro-labeled sensory neurons innervating tongue tissue, we determined changes in transcriptomic profiles in males and female mice under naive as well as tongue-tumor bearing conditions Our data revealed the following interesting findings: 1) Tongue tissue of female mice was innervated with higher number of trigeminal neurons compared to males; 2) Naive female neurons innervating the tongue exclusively expressed immune cell markers such as Csf1R, C1qa and others, that werent expressed in males. This was validated by Immunohistochemistry. 4) Accordingly, immune cell markers such as Csf1 exclusively sensitized TRPV1 responses in female TG neurons. 3) Male neurons were more tightly regulated than female neurons upon tumor growth and very few differentially expressed genes (DEGs) overlapped between the sexes, 5) Male DEGs contained higher number of transcription factors whereas female DEGs contained higher number of enzymes, cytokines and chemokines. Collectively, this is the first study to characterize the effect of sex as well as of tongue-tumor on global gene expression, pathways and molecular function of tongue-innervating sensory neurons.

neuroscience↗
bioRxiv2022

Discovery of thyrocyte heterogeneity reveals an essential role of Notch signaling in thyroid function and mammalian homeostasis

The thyroid functions at the apex of a web of endocrine organs that control cell growth, differentiation and metabolic homeostasis. Thyroid dysregulation significantly impacts human health in myriad ways with thyroid diseases standing as the most common endocrine disorder. Despite the essential role of the thyroid in human health, a high-resolution view of the cellular composition as well as molecular mechanisms that govern function of this crucial organ have been lacking. Employing the first single-cell analyses of adult mouse thyroid, we here report the discovery of unexpected thyrocyte heterogeneity, specifically three distinct thyrocyte subtypes marked by different metabolic and Notch signaling patterns. Using a battery of pharmacologic and genetic methods, we find that selective inhibition of Notch ligands and receptors disrupts thyrocyte mitochondrial activity and ROS production, thus decreasing levels of circulating thyroid hormones, inducing hypothyroidism and disrupting whole-body thermoregulation. We find an enriched frequency of hypothyroidism in children with Alagille Syndrome, a genetic disorder marked by Notch loss-of-function mutations, suggesting that our Notch-thyroid mechanisms are relevant in humans and directly account for Alagille hypothyroidism. Overall, our work reveals that Notch, although classically described as a developmental pathway that determines cell fate, controls homeostasis and thermoregulation in the adult through a mitochondria-based mechanism in a subset of thyrocytes. Our fine-grained picture of the thyroid unveils a novel understanding of this key metabolic organ and provides clinically impactful insights into its pathological dysfunctions.

cell biology↗
bioRxiv2022

RNA Methylome Reveals the m6A-Mediated Regulation of Flavor Metabolites in Tea Leaves under Solar-Withering

Epitranscriptomic mark N6-methyladenosine (m6A) is the most predominant internal modification in RNAs, which plays pivotal roles in response to diverse stresses. Multiple environmental stresses caused by withering process can greatly influence the accumulation of specialized metabolites and the formation of tea flavor. However, little is known about the effects of m6A-mediated regulatory mechanism on flavor-related metabolisms in tea leaves. Here, we explored m6A-mediated regulatory mechanism and its impacts on flavonoid and terpenoid metabolisms under solar-withering using integrated RNA methylome and transcriptome. Dynamic changes in global m6A levels of tea leaves are mainly controlled by two m6A erasers (CsALKBH4A and CsALKBH4B) under solar-withering. Differentially methylated peak (DMP)-associated genes under different shading rates of solar-withering were identified and found to be enriched in terpenoid biosynthesis and spliceosome pathways. Further analyses indicated that CsALKBH4-driven RNA demethylation can not only directly affect the accumulation of volatile terpenoids by mediating the stability and abundance of terpenoid biosynthesis-related genes, but also indirectly influence the contents of flavonoids, catechins, and theaflavins via triggering the alternative splicing (AS)-mediated regulation. Our findings underscored a novel layer of epitranscriptomic gene regulation in tea flavor-related metabolic pathways and established a compelling link between m6A-mediated regulatory mechanism and the formation of high-quality flavor in tea leaves under solar-withering.

plant biology↗
bioRxiv2021

Novel aspects of iron homeostasis in pathogenic bloodstream form Trypanosoma brucei

Iron is an essential regulatory signal for virulence factors in many pathogens. Mammals and bloodstream form (BSF) Trypanosoma brucei obtain iron by receptor-mediated endocytosis of transferrin bound to receptors (TfR) but the mechanisms by which T. brucei subsequently handles iron remains enigmatic. Here, we analyse the transcriptome of T. brucei cultured in iron-rich and iron-poor conditions. We show that adaptation to iron-deprivation induces upregulation of TfR, a cohort of parasite-specific genes (ESAG3, PAGS), genes involved in glucose uptake and glycolysis (THT1 and hexokinase), endocytosis (Phosphatidic Acid Phosphatase, PAP2), and most notably a divergent RNA binding protein RBP5, indicative of a non-canonical mechanism for regulating intracellular iron levels. We show that cells depleted of TfR by RNA silencing import free iron as a compensatory survival strategy. The TfR and RBP5 iron response are reversible by genetic complementation, the response kinetics are similar, but the regulatory mechanisms are distinct. Increased TfR protein is due to increased mRNA. Increased RBP5 expression, however, occurs by a post-transcriptional feedback mechanism whereby RBP5 interacts with its own, and with PAP2 mRNAs. Further observations suggest that increased RBP5 expression in iron-deprived cells has a maximum threshold as ectopic overexpression above this threshold disrupts normal cell cycle progression resulting in an accumulation of anucleate cells and cells in G2/M phase. This phenotype is not observed with overexpression of RPB5 containing a point mutation (F61A) in its single RNA Recognition Motif. Our experiments shed new light on how T. brucei BSFs reorganise their transcriptome to deal with iron stress revealing the first iron responsive RNA binding protein that is co-regulated with TfR, is important for cell viability and iron homeostasis; two essential processes for successful proliferation. Author SummaryAfrican trypanosomes are single-celled extracellular parasites of humans and animals relying on essential host nutrients for survival. They satisfy their iron needs by capturing host transferrin-bound iron using a surface-localised transferrin receptor (TfR) that is structurally distinct from its host counterpart. Little is known about the trypanosome response to fluctuations in host iron availability, with the exception of modulated TfR expression. We show that unlike other eukaryotes, at the transcriptome level, trypanosomes do not regulate iron-dependent enzymes as a mechanism to cope with iron deprivation. Instead, we identify a group of novel iron responsive trypanosome-specific genes, particularly an RNA Binding Protein RBP5 that is responsive to iron levels, albeit mediated by a distinct mechanism from TfR. We show that although RBP5 expression is elevated at the mRNA and protein levels, increased abundance above a maximum threshold is toxic. The trypanosome TfR has been suggested as a therapeutic target, but whether it is essential for optimal host colonisation is unclear. Our data demonstrate that trypanosomes efficiently import free iron from their environment independent of TfR suggesting that alternative iron uptake pathways exist, and that any therapeutic interventions targeting TfR must be evaluated with caution.

microbiology↗
bioRxiv2021

Genetic Diversity of Dianthus Plants Revealed by SRAP Markers

Dianthus is a valuable genetic resource for flower breeding. However, study on the genetic similarity of Dianthus plants is rare. In this work, 20 pairs of SRAP markers were used to analyze genetic diversity of 44 Dianthus plants, including 13 lines of wild Dianthus chinensis L., 7 lines of wild Dianthus superbus L. and 24 commercial varieties including Dianthus caryophyllus L., Dianthus plumarius L. and Dianthus barbatus L.. Results showed that precise interspecific and intraspecific genetic diversity was provided in Dianthus plants by using 20 pairs of SRAP molecular markers. The interspecific genetic diversity of Dianthus plants was much abundant and the intraspecific genetic difference of wild Dianthus species was related to their geographical distribution and habitats. In this work, theoretical basis and technical support were provided for crossbreeding and molecular mechanism research of Dianthus plants.

genetics↗
bioRxiv2020

Interpretable Segmentation of Amyloid-βStained Whole Slide Images of Brain Tissue

Neurodegenerative disease pathologies have been reported in both grey matter (GM) and white matter (WM) with different density distributions, an automated separation of GM/WM would be extremely advantageous for aiding in neuropathologic deep phenotyping. Standard segmentation methods typically involve manual annotations, where a trained researcher traces the delineation of GM/WM in ultra-high-resolution Whole Slide Images (WSIs). This method can be time-consuming and subjective, preventing the analysis of large amounts of WSIs at scale. This paper proposes an automated segmentation pipeline combining a Convolutional Neural Network (CNN) module for segmenting GM/WM regions and a post-processing module to remove artifacts/residues of tissues as well as generate XML annotations that can be visualized via Aperio ImageScope. First, we investigate two baseline models for medical image segmentation: FCN, and U-Net. Then we propose a patch-based approach, ResNet-Patch, to classify the GM/WM/background regions. In addition, we integrate a Neural Conditional Random Field (NCRF) module, ResNet-NCRF, to model and incorporate the spatial correlations among neighboring patches. Although their mechanisms are greatly different, both U-Net and ResNet-Patch/ResNet-NCRF achieve Intersection over Union (IoU) of more than 90% in GM and more than 80% in WM, while ResNet-Patch achieves 1% superior to U-Net with lower variance among various WSIs. ResNet-NCRF further improves the IoU by 3% for WM compared to ResNet-Patch before post-processing. We also apply gradient-weighted class activation mapping (Grad-CAM) to interpret the segmentation masks and provide relevant explanations and insights.

neuroscience↗
bioRxiv2019

Optimized data representation and convolutional neural network model for predicting tumor purity

Here we present a machine learning model, Deep Purity (DePuty) that leverages convolutional neural networks to accurately predict tumor purity from next-generation sequencing data from clinical samples without matched normals. As input, our model utilizes SNP-based copy number and minor allele frequency data formulated as a scatterplot image. With a representation matching that used by expert human annotators, we best an existing algorithm using only ~100 manually curated samples. Our simple, data-efficient approach can serve as a straightforward alternative to traditional, more complex statistical methods, for building performant purity prediction models that enable downstream bioinformatic analysis of tumor variants and absolute copy number alterations relevant to cancer genomics.

bioinformatics↗