RESUMO
Deep learning models have been used for a variety of image processing tasks. However, most of these models are developed through supervised learning approaches, which rely heavily on the availability of large-scale annotated datasets. Developing such datasets is tedious and expensive. In the absence of an annotated dataset, synthetic data can be used for model development; however, due to the substantial differences between simulated and real data, a phenomenon referred to as domain gap, the resulting models often underperform when applied to real data. In this research, we aim to address this challenge by first computationally simulating a large-scale annotated dataset and then using a generative adversarial network (GAN) to fill the gap between simulated and real images. This approach results in a synthetic dataset that can be effectively utilized to train a deep-learning model. Using this approach, we developed a realistic annotated synthetic dataset for wheat head segmentation. This dataset was then used to develop a deep-learning model for semantic segmentation. The resulting model achieved a Dice score of 83.4% on an internal dataset and Dice scores of 79.6% and 83.6% on two external datasets from the Global Wheat Head Detection datasets. While we proposed this approach in the context of wheat head segmentation, it can be generalized to other crop types or, more broadly, to images with dense, repeated patterns such as those found in cellular imagery.
RESUMO
During endochondral ossification, chondrocytes secrete a proteoglycan (PG)-rich extracellular matrix that can inhibit the process of cartilage maturation, including expression of Ihh and Col10a1. Because bone morphogenetic proteins (BMPs) can promote cartilage maturation, we hypothesized that cartilage PGs normally inhibit BMP signalling. Accordingly, BMP signalling was evaluated in chondrocytes of wild-type and PG mutant (fam20b-/-) zebrafish and inhibited with temporal control using the drug DMH1 or an inducible dominant-negative BMP receptor transgene (dnBMPR). Compared with wild type, phospho-Smad1/5/9, but not phospho-p38, was increased in fam20b-/- chondrocytes, but only after they secreted PGs. Phospho-Smad1/5/9 was decreased in DMH1-treated or dnBMPR-activated wild-type chondrocytes, and DMH1 also decreased phospho-p38 levels. ihha and col10a1a were decreased in DMH1-treated or dnBMPR-activated chondrocytes, and less perichondral bone formed. Finally, early ihha and col10a1a expression and early perichondral bone formation of fam20b mutants were rescued with DMH1 treatment or dnBMPR activation. Therefore, PG inhibition of canonical BMP-dependent cartilage maturation delays endochondral ossification, and these results offer hope for the development of growth factor therapies for skeletal defects of PG diseases.
Assuntos
Osteogênese , Proteoglicanas , Animais , Osteogênese/genética , Proteoglicanas/genética , Proteoglicanas/metabolismo , Peixe-Zebra/genética , Cartilagem/metabolismo , Condrócitos/metabolismo , Proteínas Morfogenéticas Ósseas/metabolismoRESUMO
BACKGROUND: Mouse, chick, and zebrafish undergo a highly conserved program of cartilage maturation during endochondral ossification (bone formation via a cartilage template). Standard histological and molecular features of cartilage maturation are chondrocyte hypertrophy, downregulation of the chondrogenic markers Sox9 and Col2a1, and upregulation of Col10a1. We tested whether cartilage maturation is conserved in an amphibian, the western clawed frog Xenopus tropicalis, using in situ hybridization for standard markers and a novel laser-capture microdissection RNAseq data set. We also functionally tested whether thyroid hormone drives cartilage maturation in X tropicalis, as it does in other vertebrates. RESULTS: The developing frog humerus mostly followed the standard progression of cartilage maturation. Chondrocytes gradually became hypertrophic as col2a1 and sox9 were eventually down-regulated, but col10a1 was not up-regulated. However, the expression levels of several genes associated with the early formation of cartilage, such as acan, sox5, and col9a2, remained highly expressed even as humeral chondrocytes matured. Greater deviances were observed in head cartilages, including the ceratohyal, which underwent hypertrophy within hours of becoming cartilaginous, maintained relatively high levels of col2a1 and sox9, and lacked col10a1 expression. Interestingly, treating frog larvae with thyroid hormone antagonists did not specifically reduce head cartilage hypertrophy, resulting rather in a global developmental delay. CONCLUSION: These data reveal that basic cartilage maturation features in the head, and to a lesser extent in the limb, are not conserved in X tropicalis. Future work revealing how frogs deviate from the standard cartilage maturation program might shed light on both evolutionary and health studies.
Assuntos
Cartilagem , Peixe-Zebra , Camundongos , Animais , Camundongos Transgênicos , Condrócitos/metabolismo , Anfíbios , Hipertrofia , Diferenciação CelularRESUMO
Introduction: The intracellular Ca2+ sensor stromal interaction molecule 1 (STIM1) is thought to play a critical role in enamel development, as its mutations cause Amelogenesis Imperfecta (AI). We recently established an ameloblast-specific (AmelX-iCre) Stim1 conditional deletion mouse model to investigate the role of STIM1 in controlling ameloblast function and differentiation in vivo (Stim1 cKO). Our pilot data (Said et al., J. Dent. Res., 2019, 98, 1002-1010) support our hypothesis for a broad role of Stim1 in amelogenesis. This paper aims to provide an in-depth characterization of the enamel phenotype observed in our Stim1 cKO model. Methods: We crossed AmelX-iCre mice with Stim1-floxed animals to develop ameloblast-specific Stim1 cKO mice. Scanning electron microscopy, energy dispersive spectroscopy, and micro- CT were used to study the enamel phenotype. RNAseq and RT-qPCR were utilized to evaluate changes in the gene expression of several key ameloblast genes. Immunohistochemistry was used to detect the amelogenin, matrix metalloprotease 20 and kallikrein 4 proteins in ameloblasts. Results: Stim1 cKO animals exhibited a hypomineralized AI phenotype, with reduced enamel volume, diminished mineral density, and lower calcium content. The mutant enamel phenotype was more severe in older Stim1 cKO mice compared to younger ones and changes in enamel volume and mineral content were more pronounced in incisors compared to molars. Exploratory RNAseq analysis of incisors' ameloblasts suggested that ablation of Stim1 altered the expression levels of several genes encoding enamel matrix proteins which were confirmed by subsequent RT-qPCR. On the other hand, RT-qPCR analysis of molars' ameloblasts showed non-significant differences in the expression levels of enamel matrix genes between control and Stim1-deficient cells. Moreover, gene expression analysis of incisors' and molars' ameloblasts showed that Stim1 ablation caused changes in the expression levels of several genes associated with calcium transport and mitochondrial kinetics. Conclusions: Collectively, these findings suggest that the loss of Stim1 in ameloblasts may impact enamel mineralization and ameloblast gene expression.
RESUMO
Purpose: To investigate the impact of the following three methodological pitfalls on model generalizability: (a) violation of the independence assumption, (b) model evaluation with an inappropriate performance indicator or baseline for comparison, and (c) batch effect. Materials and Methods: The authors used retrospective CT, histopathologic analysis, and radiography datasets to develop machine learning models with and without the three methodological pitfalls to quantitatively illustrate their effect on model performance and generalizability. F1 score was used to measure performance, and differences in performance between models developed with and without errors were assessed using the Wilcoxon rank sum test when applicable. Results: Violation of the independence assumption by applying oversampling, feature selection, and data augmentation before splitting data into training, validation, and test sets seemingly improved model F1 scores by 71.2% for predicting local recurrence and 5.0% for predicting 3-year overall survival in head and neck cancer and by 46.0% for distinguishing histopathologic patterns in lung cancer. Randomly distributing data points for a patient across datasets superficially improved the F1 score by 21.8%. High model performance metrics did not indicate high-quality lung segmentation. In the presence of a batch effect, a model built for pneumonia detection had an F1 score of 98.7% but correctly classified only 3.86% of samples from a new dataset of healthy patients. Conclusion: Machine learning models developed with these methodological pitfalls, which are undetectable during internal evaluation, produce inaccurate predictions; thus, understanding and avoiding these pitfalls is necessary for developing generalizable models.Keywords: Random Forest, Diagnosis, Prognosis, Convolutional Neural Network (CNN), Medical Image Analysis, Generalizability, Machine Learning, Deep Learning, Model Evaluation Supplemental material is available for this article. Published under a CC BY 4.0 license.
RESUMO
OBJECTIVES: Nucleic acid amplification tests (NAATs) are highly sensitive for the detection of Chlamydia trachomatis (CT) and Neisseria gonorrhoeae (NG) DNA/ribosomal RNA (rRNA). Previous studies have demonstrated contamination of surfaces in sexual health clinics (SHCs) with CT/NG. False positive results can occur if patient samples are contaminated by environmental DNA/rRNA. This can have a dramatic impact on patients' lives and relationships. Previous attempts to reduce contamination, through staff training alone, have been unsuccessful. We aimed to investigate environmental contamination levels in SHCs and to assess a two-armed intervention aimed at reducing surface contamination. METHODS: Questionnaires were sent to 10 SHCs. Six clinics, with differing characteristics, were selected to participate in sample collection. Each clinic followed standardised instructions to sample surfaces using a CT/NG NAAT swab. Clinics were invited to introduce the two-armed intervention. The first arm was cleaning with a chlorine-based cleaning solution once daily. The second arm involved introducing clinic-specific changes to reduce contamination. RESULTS: 7/10 (70%) clinics completed the questionnaire. Overall, 88/263 (33%) swabs were positive for CT/NG. Clinics 1, 3 and 4 had high levels of contamination, with 28/64 (44%), 17/33 (52%) and 30/52 (58%) swabs testing positive, respectively. Clinics 2 and 6 had lower levels of contamination, with 7/46 (15%) and 6/35 (17%), respectively. 0/33 (0%) of swabs were positive at clinic 5 and this was the only clinic already using a chlorine-based solution to clean all surfaces and delivering twice-yearly clinic-specific infection control training. Following both intervention arms at clinic 1, 2/50 (4%, p<0.0001) swabs tested positive for CT/NG. Clinic 4 introduced each arm separately. After the first intervention, 13/52 (25%, p=0.003) swabs tested positive and following the second arm 4/50 (8%, p<0.0001) swabs were positive. CONCLUSIONS: Environmental contamination is a concern in SHCs. We recommend that all SHCs monitor contamination levels and, if necessary, consider using chlorine-based cleaning products and introduce clinic-specific changes to address environmental contamination.
Assuntos
Infecções por Chlamydia , Gonorreia , Humanos , Neisseria gonorrhoeae/genética , Chlamydia trachomatis/genética , Gonorreia/diagnóstico , Gonorreia/prevenção & controle , Cloro , Infecções por Chlamydia/diagnóstico , Infecções por Chlamydia/prevenção & controle , Sensibilidade e Especificidade , Técnicas de Amplificação de Ácido Nucleico/métodosRESUMO
To explain how cartilage appeared in different parts of the vertebrate body at discrete times during evolution, we hypothesize that different embryonic populations co-opted expression of a core gene regulatory network (GRN) driving chondrocyte differentiation. To test this hypothesis, laser-capture microdissection coupled with RNA-seq was used to reveal chondrocyte transcriptomes in the developing chick humerus and ceratobranchial, which are mesoderm- and neural crest-derived, respectively. During endochondral ossification, two general types of chondrocytes differentiate. Immature chondrocytes (IMM) represent the early stages of cartilage differentiation, while mature chondrocytes (MAT) undergo additional stages of differentiation, including hypertrophy and stimulating matrix mineralization and degradation. Venn diagram analyses generally revealed a high degree of conservation between chondrocyte transcriptomes of the limb and head, including SOX9, COL2A1, and ACAN expression. Typical maturation genes, such as COL10A1, IBSP, and SPP1, were upregulated in MAT compared to IMM in both limb and head chondrocytes. Gene co-expression network (GCN) analyses of limb and head chondrocyte transcriptomes estimated the core GRN governing cartilage differentiation. Two discrete portions of the GCN contained genes that were differentially expressed in limb or head chondrocytes, but these genes were enriched for biological processes related to limb/forelimb morphogenesis or neural crest-dependent processes, respectively, perhaps simply reflecting the embryonic origin of the cells. A core GRN driving cartilage differentiation in limb and head was revealed that included typical chondrocyte differentiation and maturation markers, as well as putative novel "chondrocyte" genes. Conservation of a core transcriptional program during chondrocyte differentiation in both the limb and head suggest that the same core GRN was co-opted when cartilage appeared in different regions of the skeleton during vertebrate evolution.
RESUMO
Ancestors of the Antarctic icefishes (family Channichthyidae) were benthic and had no swim bladder, making it energetically expensive to rise from the ocean floor. To exploit the water column, benthopelagic icefishes were hypothesized to have evolved a skeleton with "reduced bone," which gross anatomical data supported. Here, we tested the hypothesis that changes to icefish bones also occurred below the level of gross anatomy. Histology and micro-CT imaging of representative craniofacial bones (i.e., ceratohyal, frontal, dentary, and articular) of extant Antarctic fish species specifically evaluated two features that might cause the appearance of "reduced bone": bone microstructure (e.g., bone volume fraction and structure linear density) and bone mineral density (BMD, or mass of mineral per volume of bone). Measures of bone microstructure were not consistently different in bones from the icefishes Chaenocephalus aceratus and Champsocephalus gunnari, compared to the related benthic notothenioids Notothenia coriiceps and Gobionotothen gibberifrons. Some quantitative measures, such as bone volume fraction and structure linear density, were significantly increased in some icefish bones compared to homologous bones of non-icefish. However, such differences were rare, and no microstructural measures were consistently different in icefishes across all bones and species analyzed. Furthermore, BMD was similar among homologous bones of icefish and non-icefish Antarctic notothenioids. In summary, "reduced bone" in icefishes was not due to systemic changes in bone microstructure or BMD, raising the prospect that "reduced bone" in icefish occurs only at the gross anatomic level (i.e., smaller or fewer bones). Given that icefishes exhibit delayed skeletal development compared to non-icefish Antarctic fishes, combining these phenotypic data with genomic data might clarify genetic changes driving skeletal heterochrony.
Assuntos
Densidade Óssea , Perciformes , Animais , Regiões Antárticas , Peixes/anatomia & histologia , Perciformes/anatomia & histologiaRESUMO
Gene set analysis has been widely used to gain insight from high-throughput expression studies. Although various tools and methods have been developed for gene set analysis, there is no consensus among researchers regarding best practice(s). Most often, evaluation studies have reported contradictory recommendations of which methods are superior. Therefore, an unbiased quantitative framework for evaluations of gene set analysis methods will be valuable. Such a framework requires gene expression datasets where enrichment status of gene sets is known a priori. In the absence of such gold standard datasets, artificial datasets are commonly used for evaluations of gene set analysis methods; however, they often rely on oversimplifying assumptions that make them biased in favor of or against a given method. In this paper, we propose a quantitative framework for evaluation of gene set analysis methods by synthesizing expression datasets using real data, without relying on oversimplifying or unrealistic assumptions, while preserving complex gene-gene correlations and retaining the distribution of expression values. The utility of the quantitative approach is shown by evaluating ten widely used gene set analysis methods. An implementation of the proposed method is publicly available. We suggest using Silver to evaluate existing and new gene set analysis methods. Evaluation using Silver provides a better understanding of current methods and can aid in the development of gene set analysis methods to achieve higher specificity without sacrificing sensitivity.
Assuntos
Bases de Dados Genéticas/normas , Genômica/métodos , Software , Conjuntos de Dados como Assunto/normasRESUMO
Similarities and differences in the associations of biological entities among species can provide us with a better understanding of evolutionary relationships. Often the evolution of new phenotypes results from changes to interactions in pre-existing biological networks and comparing networks across species can identify evidence of conservation or adaptation. Gene co-expression networks (GCNs), constructed from high-throughput gene expression data, can be used to understand evolution and the rise of new phenotypes. The increasing abundance of gene expression data makes GCNs a valuable tool for the study of evolution in non-model organisms. In this paper, we cover motivations for why comparing these networks across species can be valuable for the study of evolution. We also review techniques for comparing GCNs in the context of evolution, including local and global methods of graph alignment. While some protein-protein interaction (PPI) bioinformatic methods can be used to compare co-expression networks, they often disregard highly relevant properties, including the existence of continuous and negative values for edge weights. Also, the lack of comparative datasets in non-model organisms has hindered the study of evolution using PPI networks. We also discuss limitations and challenges associated with cross-species comparison using GCNs, and provide suggestions for utilizing co-expression network alignments as an indispensable tool for evolutionary studies going forward.
RESUMO
Current radiomic studies of head and neck squamous cell carcinomas (HNSCC) are typically based on datasets combining tumors from different locations, assuming that the radiomic features are similar based on histopathologic characteristics. However, molecular pathogenesis and treatment in HNSCC substantially vary across different tumor sites. It is not known if a statistical difference exists between radiomic features from different tumor sites and how they affect machine learning model performance in endpoint prediction. To answer these questions, we extracted radiomic features from contrast-enhanced neck computed tomography scans (CTs) of 605 patients with HNSCC originating from the oral cavity, oropharynx, and hypopharynx/larynx. The difference in radiomic features of tumors from these sites was assessed using statistical analyses and Random Forest classifiers on the radiomic features with 10-fold cross-validation to predict tumor sites, nodal metastasis, and HPV status. We found statistically significant differences (p-value ≤ 0.05) between the radiomic features of HNSCC depending on tumor location. We also observed that differences in quantitative features among HNSCC from different locations impact the performance of machine learning models. This suggests that radiomic features may reveal biologic heterogeneity complementary to current gold standard histopathologic evaluation. We recommend considering tumor site in radiomic studies of HNSCC.
RESUMO
A 49-year-old man presented with a 1-week history of abdominal pain, distension, diarrhoea and fatigue. CT of the abdomen and pelvis revealed peritonitis with no identifiable cause. Diagnostic laparoscopy was performed, which excluded gastrointestinal perforation. Peritoneal fluid tested positive for Chlamydia trachomatis and rectal swabs were positive for C. trachomatis serovars consistent with lymphogranuloma venereum (LGV). Additional blood tests also revealed a diagnosis of syphilis. This is a rare documented case of LGV peritonitis in a male without associated immunodeficiency. The patient recovered well following laparoscopic washout and a course of appropriate antibiotics.
Assuntos
Linfogranuloma Venéreo , Peritonite , Antibacterianos/uso terapêutico , Chlamydia trachomatis , Homossexualidade Masculina , Humanos , Linfogranuloma Venéreo/complicações , Linfogranuloma Venéreo/diagnóstico , Linfogranuloma Venéreo/tratamento farmacológico , Masculino , Pessoa de Meia-Idade , Peritonite/diagnóstico , Peritonite/tratamento farmacológicoRESUMO
BACKGROUND: Gene co-expression networks (GCNs) are not easily comparable due to their complex structure. In this paper, we propose a tool, Juxtapose, together with similarity measures that can be utilized for comparative transcriptomics between a set of organisms. While we focus on its application to comparing co-expression networks across species in evolutionary studies, Juxtapose is also generalizable to co-expression network comparisons across tissues or conditions within the same species. METHODS: A word embedding strategy commonly used in natural language processing was utilized in order to generate gene embeddings based on walks made throughout the GCNs. Juxtapose was evaluated based on its ability to embed the nodes of synthetic structures in the networks consistently while also generating biologically informative results. Evaluation of the techniques proposed in this research utilized RNA-seq datasets from GTEx, a multi-species experiment of prefrontal cortex samples from the Gene Expression Omnibus, as well as synthesized datasets. Biological evaluation was performed using gene set enrichment analysis and known gene relationships in literature. RESULTS: We show that Juxtapose is capable of globally aligning synthesized networks as well as identifying areas that are conserved in real gene co-expression networks without reliance on external biological information. Furthermore, output from a matching algorithm that uses cosine distance between GCN embeddings is shown to be an informative measure of similarity that reflects the amount of topological similarity between networks. CONCLUSIONS: Juxtapose can be used to align GCNs without relying on known biological similarities and enables post-hoc analyses using biological parameters, such as orthology of genes, or conserved or variable pathways. AVAILABILITY: A development version of the software used in this paper is available at https://github.com/klovens/juxtapose.
Assuntos
Biologia Computacional , Redes Reguladoras de Genes , Algoritmos , SoftwareRESUMO
This article reviews the history of artificial intelligence and introduces the reader to major events that prompted interest in the field, as well as pitfalls and challenges that have slowed its development. The purpose of this article is to provide a high-level historical perspective on the development of the field over the past decades, highlighting the potential of the field for transforming health care, but also the importance of setting realistic expectations for artificial intelligence applications to avoid repeating historical cyclical trends and a third "artificial intelligence winter."
Assuntos
Inteligência Artificial , Neuroimagem/métodos , HumanosRESUMO
The deployment of machine learning (ML) models in the health care domain can increase the speed and accuracy of diagnosis and improve treatment planning and patient care. Translating academic research to applications that are deployable in clinical settings requires the ability to generalize and high reproducibility, which are contingent on a rigorous and sound methodology for the development and evaluation of ML models. This article describes the fundamental concepts and processes for ML model evaluation and highlights common workflows. It concludes with a discussion of the requirements for the deployment of ML models in clinical settings.
Assuntos
Interpretação de Imagem Assistida por Computador/métodos , Aprendizado de Máquina , Neuroimagem , Humanos , Reprodutibilidade dos TestesRESUMO
The extensive body of research and advances in machine learning (ML) and the availability of a large volume of patient data make ML a powerful tool for producing models with the potential for widespread deployment in clinical settings. This article provides an overview of the classic supervised and unsupervised ML methods as well as fundamental concepts required for understanding how to develop generalizable and high-performing ML applications. It also describes the important steps for developing a ML model and how decisions made in these steps affect model performance and ability to generalize.
Assuntos
Interpretação de Imagem Assistida por Computador/métodos , Aprendizado de Máquina , Neuroimagem/métodos , HumanosRESUMO
Gene set analysis methods are widely used to provide insight into high-throughput gene expression data. There are many gene set analysis methods available. These methods rely on various assumptions and have different requirements, strengths and weaknesses. In this paper, we classify gene set analysis methods based on their components, describe the underlying requirements and assumptions for each class, and provide directions for future research in developing and evaluating gene set analysis methods.
RESUMO
BACKGROUND: Mycoplasma genitalium (Mgen) causes non-gonococcal urethritis (NGU) and is believed to cause pelvic inflammatory disease (PID). High rates of macrolide resistance are well documented globally for Mgen. In Brighton, patients with NGU and PID are tested for Mgen and test of cure (TOC) offered post-treatment. METHODS: Demographic, clinical and treatment history data were collected over a 12-month period for all Mgen-positive patients in a Brighton-based genitourinary clinic. RESULTS: There were 114 patients with Mgen. 18% (61/339) of men with NGU and 9% (15/160) of women with PID had Mgen. 62/114 (54%) returned for first test TOC 4 weeks after treatment. 27/62 (44%) had a positive TOC; 25/27 (92.6%) had received azithromycin first line (500 mg stat then 250 mg OD for 4 days), 1/27 (3.7%) had received moxifloxacin first line (400 mg OD for 14 days) and 1/27 (3.7%) had received doxycycline first line (100 mg BD for 7 days). 20/27 (74%) returned for a second TOC 4 weeks later. 5/20 (25%) patients were positive on second TOC; 3/5 (60%) had received azithromycin second line and 2/5 (40%) had received moxifloxacin second line. Patients were more likely to have a positive TOC if they were at risk of reinfection (9/27 positive TOC vs 3/35 negative TOC; p=0.02). Patients given moxifloxacin were more likely to have a negative TOC (1/27 positive TOC vs 9/35 negative TOC; p=0.03) than those who received other antibiotic regimens. CONCLUSIONS: Treatment failure rates for Mgen following azithromycin use are substantial, raising concerns regarding resistance. However, reinfection risk may contribute, suggesting a requirement for improved public awareness and clinician knowledge.
Assuntos
Infecções por Mycoplasma/tratamento farmacológico , Mycoplasma genitalium/efeitos dos fármacos , Resultado do Tratamento , Uretrite/etiologia , Adolescente , Adulto , Antibacterianos/uso terapêutico , Azitromicina/uso terapêutico , Doxiciclina/uso terapêutico , Inglaterra/epidemiologia , Feminino , Humanos , Masculino , Programas de Rastreamento/métodos , Programas de Rastreamento/estatística & dados numéricos , Pessoa de Meia-Idade , Moxifloxacina/uso terapêutico , Infecções por Mycoplasma/epidemiologia , Mycoplasma genitalium/patogenicidade , Serviços de Saúde Reprodutiva/normas , Serviços de Saúde Reprodutiva/estatística & dados numéricos , Uretrite/epidemiologia , Uretrite/terapiaRESUMO
Gene set analysis is a quantitative approach for generating biological insight from gene expression datasets. The abundance of gene set analysis methods speaks to their popularity, but raises the question of the extent to which results are affected by the choice of method. Our systematic analysis of 13 popular methods using 6 different datasets, from both DNA microarray and RNA-Seq origin, shows that this choice matters a great deal. We observed that the overall number of gene sets reported by each method differed by up to 2 orders of magnitude, and there was a bias toward reporting large gene sets with some methods. Furthermore, there was substantial disagreement between the 20 most statistically significant gene sets reported by the methods. This was also observed when expanding to the 100 most statistically significant reported gene sets. For different datasets of the same phenotype/condition, the top 20 and top 100 most significant results also showed little to no agreement even when using the same method. GAGE, PAGE, and ORA were the only methods able to achieve relatively high reproducibility when comparing the 20 and 100 most statistically significant gene sets. Biological validation on a juvenile idiopathic arthritis (JIA) dataset showed wide variation in terms of the relevance of the top 20 and top 100 most significant gene sets to known biology of the disease, where GAGE predicted the most relevant gene sets, followed by GSEA, ORA, and PAGE.
Assuntos
Bases de Dados Genéticas , Perfilação da Expressão Gênica/estatística & dados numéricos , Artrite Juvenil/genética , Perfilação da Expressão Gênica/métodos , Perfilação da Expressão Gênica/normas , Humanos , Análise de Sequência com Séries de Oligonucleotídeos/estatística & dados numéricos , Fenótipo , Psoríase/genética , Reprodutibilidade dos TestesRESUMO
BACKGROUND: Gene set analysis is a well-established approach for interpretation of data from high-throughput gene expression studies. Achieving reproducible results is an essential requirement in such studies. One factor of a gene expression experiment that can affect reproducibility is the choice of sample size. However, choosing an appropriate sample size can be difficult, especially because the choice may be method-dependent. Further, sample size choice can have unexpected effects on specificity. RESULTS: In this paper, we report on a systematic, quantitative approach to study the effect of sample size on the reproducibility of the results from 13 gene set analysis methods. We also investigate the impact of sample size on the specificity of these methods. Rather than relying on synthetic data, the proposed approach uses real expression datasets to offer an accurate and reliable evaluation. CONCLUSION: Our findings show that, as a general pattern, the results of gene set analysis become more reproducible as sample size increases. However, the extent of reproducibility and the rate at which it increases vary from method to method. In addition, even in the absence of differential expression, some gene set analysis methods report a large number of false positives, and increasing sample size does not lead to reducing these false positives. The results of this research can be used when selecting a gene set analysis method from those available.