Non-small-cell lung cancer classification via RNA-Seq and histology imaging probability fusion.
Ontology highlight
ABSTRACT: Background
Adenocarcinoma and squamous cell carcinoma are the two most prevalent lung cancer types, and their distinction requires different screenings, such as the visual inspection of histology slides by an expert pathologist, the analysis of gene expression or computer tomography scans, among others. In recent years, there has been an increasing gathering of biological data for decision support systems in the diagnosis (e.g. histology imaging, next-generation sequencing technologies data, clinical information, etc.). Using all these sources to design integrative classification approaches may improve the final diagnosis of a patient, in the same way that doctors can use multiple types of screenings to reach a final decision on the diagnosis. In this work, we present a late fusion classification model using histology and RNA-Seq data for adenocarcinoma, squamous-cell carcinoma and healthy lung tissue.Results
The classification model improves results over using each source of information separately, being able to reduce the diagnosis error rate up to a 64% over the isolate histology classifier and a 24% over the isolate gene expression classifier, reaching a mean F1-Score of 95.19% and a mean AUC of 0.991.Conclusions
These findings suggest that a classification model using a late fusion methodology can considerably help clinicians in the diagnosis between the aforementioned lung cancer cancer subtypes over using each source of information separately. This approach can also be applied to any cancer type or disease with heterogeneous sources of information.
SUBMITTER: Carrillo-Perez F
PROVIDER: S-EPMC8456075 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:We propose a nanomedical device for the classification of lung cancer (LC) histology. The device profiles volatile organic compounds (VOCs) in the headspace of (subtypes of) LC cells, using gold nanoparticle (GNP) sensors that are suitable for detecting LC-specific patterns of VOC profiles, as determined by gas chromatography-mass spectrometry analysis. Analyzing the GNP sensing signals by support vector machine allowed significant discrimination between (i) LC and healthy cells; (ii) small cell LC and non-small cell LC; and between (iii) two subtypes of non-small cell LC: adenocarcinoma and squamous cell carcinoma. The discriminative power of the GNP sensors was then linked with the chemical nature and composition of the headspace VOCs of each LC state. These proof-of-concept findings could totally revolutionize LC screening and diagnosis, and might eventually allow early and differential diagnosis of LC subtypes with detectable or unreachable lung nodules.In this study, a nanomedical device that profiles volatile organic compounds (VOCs) in lung cancer cells is investigated, using a matrix of gold nanoparticle (GNP) sensors that are suitable for detecting lung cancer (LC) specific patterns of VOC profiles. This device might eventually allow early differential diagnosis of LC subtypes including unreachable lung nodules.
| S-EPMC4745892 | biostudies-literature
Project description:Tumor histology is an important predictor of therapeutic response and outcomes in lung cancer. Tissue sampling for pathologist review is the most reliable method for histology classification, however, recent advances in deep learning for medical image analysis allude to the utility of radiologic data in further describing disease characteristics and for risk stratification. In this study, we propose a radiomics approach to predicting non-small cell lung cancer (NSCLC) tumor histology from non-invasive standard-of-care computed tomography (CT) data. We trained and validated convolutional neural networks (CNNs) on a dataset comprising 311 early-stage NSCLC patients receiving surgical treatment at Massachusetts General Hospital (MGH), with a focus on the two most common histological types: adenocarcinoma (ADC) and Squamous Cell Carcinoma (SCC). The CNNs were able to predict tumor histology with an AUC of 0.71(p = 0.018). We also found that using machine learning classifiers such as k-nearest neighbors (kNN) and support vector machine (SVM) on CNN-derived quantitative radiomics features yielded comparable discriminative performance, with AUC of up to 0.71 (p = 0.017). Our best performing CNN functioned as a robust probabilistic classifier in heterogeneous test sets, with qualitatively interpretable visual explanations to its predictions. Deep learning based radiomics can identify histological phenotypes in lung cancer. It has the potential to augment existing approaches and serve as a corrective aid for diagnosticians.
| S-EPMC7943565 | biostudies-literature
Project description:BACKGROUND:Understanding cellular and molecular heterogeneity in glioblastoma (GBM), the most common and aggressive primary brain malignancy, is a crucial step towards the development of effective therapies. Besides the inter-patient variability, the presence of multiple cell populations within tumors calls for the need to develop modeling strategies able to extract the molecular signatures driving tumor evolution and treatment failure. With the advances in single-cell RNA Sequencing (scRNA-Seq), tumors can now be dissected at the cell level, unveiling information from their life history to their clinical implications. RESULTS:We propose a classification setting based on GBM scRNA-Seq data, through sparse logistic regression, where different cell populations (neoplastic and normal cells) are taken as classes. The goal is to identify gene features discriminating between the classes, but also those shared by different neoplastic clones. The latter will be approached via the network-based twiner regularizer to identify gene signatures shared by neoplastic cells from the tumor core and infiltrating neoplastic cells originated from the tumor periphery, as putative disease biomarkers to target multiple neoplastic clones. Our analysis is supported by the literature through the identification of several known molecular players in GBM. Moreover, the relevance of the selected genes was confirmed by their significance in the survival outcomes in bulk GBM RNA-Seq data, as well as their association with several Gene Ontology (GO) biological process terms. CONCLUSIONS:We presented a methodology intended to identify genes discriminating between GBM clones, but also those playing a similar role in different GBM neoplastic clones (including migrating cells), therefore potential targets for therapy research. Our results contribute to a deeper understanding on the genetic features behind GBM, by disclosing novel therapeutic directions accounting for GBM heterogeneity.
| S-EPMC7029554 | biostudies-literature
Project description:MotivationPIWI-interacting RNAs (piRNAs), 23-36 nt small silencing RNAs, repress transposon expression in the metazoan germ line, thereby protecting the genome. Although high-throughput sequencing has made it possible to examine the genome and transcriptome at unprecedented resolution, extracting useful information from gigabytes of sequencing data still requires substantial computational skills. Additionally, researchers may analyze and interpret the same data differently, generating results that are difficult to reconcile. To address these issues, we developed a coordinated set of pipelines, 'piPipes', to analyze piRNA and transposon-derived RNAs from a variety of high-throughput sequencing libraries, including small RNA, RNA, degradome or 7-methyl guanosine cap analysis of gene expression (CAGE), chromatin immunoprecipitation (ChIP) and genomic DNA-seq. piPipes can also produce figures and tables suitable for publication. By facilitating data analysis, piPipes provides an opportunity to standardize computational methods in the piRNA field.Supplementary informationSupplementary information, including flowcharts and example figures for each pipeline, are available at Bioinformatics online.Availability and implementationpiPipes is implemented in Bash, C++, Python, Perl and R. piPipes is free, open-source software distributed under the GPLv3 license and is available at http://bowhan.github.io/piPipes/.ContactPhillip.Zamore@umassmed.edu or Zhiping.Weng@umassmed.eduSupplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC4325541 | biostudies-literature
Project description:RNA-seq is a well-established method for studying the transcriptome. Popular methods for library preparation in RNA-seq such as Illumina TruSeq® RNA v2 kit use a poly-A pulldown strategy. Such methods can cause loss of coverage at the 5' end of genes, impacting the ability to detect fusions when used on degraded samples. The goal of this study was to quantify the effects RNA degradation has on fusion detection when using poly-A selected mRNA and to identify the variables involved in this process.Using both artificially and naturally degraded samples, we found that there is a reduced ability to detect fusions as the distance of the breakpoint from the 3' end of the gene increases. The median transcript coverage decreases exponentially as a function of the distance from the 3' end and there is a linear relationship between the coverage decay rate and the RNA integrity number (RIN). Based on these findings we developed plots that show the probability of detecting a gene fusion ("sensitivity") as a function of the distance of the fusion breakpoint from the 3' end.This study developed a strategy to assess the impact that RNA degradation has on the ability to detect gene fusions by RNA-seq.
| S-EPMC5072325 | biostudies-literature
Project description:The performance evaluation of fusion detection algorithms from high-throughput sequencing data crucially relies on the availability of data with known positive and negative cases of gene rearrangements. The use of simulated data circumvents some shortcomings of real data by generation of an unlimited number of true and false positive events, and the consequent robust estimation of accuracy measures, such as precision and recall. Although a few simulated fusion datasets from RNA Sequencing (RNA-Seq) are available, they are of limited sample size. This makes it difficult to systematically evaluate the performance of RNA-Seq based fusion-detection algorithms. Here, we present SimFuse to address this problem. SimFuse utilizes real sequencing data as the fusions' background to closely approximate the distribution of reads from a real sequencing library and uses a reference genome as the template from which to simulate fusions' supporting reads. To assess the supporting read-specific performance, SimFuse generates multiple datasets with various numbers of fusion supporting reads. Compared to an extant simulated dataset, SimFuse gives users control over the supporting read features and the sample size of the simulated library, based on which the performance metrics needed for the validation and comparison of alternative fusion-detection algorithms can be rigorously estimated.
| S-EPMC4709598 | biostudies-literature
Project description:Single-cell sequencing, which is used to detect clinically important tumor subpopulations, is necessary for understanding tumor heterogeneity. Here, we analyzed transcriptomic data obtained from 34 single cells from human lung adenocarcinoma (LADC) patient-derived xenografts (PDXs). To focus on the intrinsic transcriptomic signatures of these tumors, we filtered out genes that displayed extensive expression changes following xenografting and cell culture. Then, we performed clustering analysis using co-regulated gene modules rather than individual genes to minimize read drop-out errors associated with single-cell sequencing. This combined approach revealed two distinct intra-tumoral subgroups that were primarily distinguished by the gene module G64. The G64 module was predominantly composed of cell-cycle genes. E2F1 was found to be the transcription factor that most likely mediates the expression of the G64 module in single LADC cells. Interestingly, the G64 module also indicated inter-tumoral heterogeneity based on its association with patient survival and other clinical variables such as smoking status and tumor stage. Taken together, these results demonstrate the feasibility of single-cell RNA sequencing and the strength of our analytical pipeline for the identification of tumor subpopulations.
| S-EPMC4549254 | biostudies-literature
Project description:BackgroundLung cancer, one of the most common malignant tumors, exhibits high inter- and intra-tumor heterogeneity which contributes significantly to treatment resistance and failure. Single-cell RNA sequencing (scRNA-seq) has been widely used to dissect the cellular composition and characterize the molecular properties of cancer cells and their tumor microenvironment in lung cancer. However, the transcriptomic heterogeneity among various cancer cells in non-small cell lung cancer (NSCLC) warrants further illustration.MethodsTo comprehensively analyze the molecular heterogeneity of NSCLC, we performed high-precision single-cell RNA-seq analyses on 7364 individual cells from tumor tissues and matched normal tissues from 19 primary lung cancer patients and 1 pulmonary chondroid hamartoma patient.ResultsIn 6 of 16 patients sequenced, we identified a significant proportion of cancer cells simultaneously expressing classical marker genes for two or even three histologic subtypes of NSCLC-adenocarcinoma (ADC), squamous cell carcinoma (SCC), and neuroendocrine tumor (NET) in the same individual cell, which we defined as mixed-lineage tumor cells; this was verified by both co-immunostaining and RNA in situ hybridization. These data suggest that mixed-lineage tumor cells are highly plastic with mixed features of different types of NSCLC. Both copy number variation (CNV) patterns and mitochondrial mutations clearly showed that the mixed-lineage and single-lineage tumor cells from the same patient had common tumor ancestors rather than different origins. Moreover, we revealed that patients with high mixed-lineage features of different cancer subtypes had worse survival than patients with low mixed-lineage features, indicating that mixed-lineage tumor features were associated with poorer prognosis. In addition, gene signatures specific to mixed-lineage tumor cells were identified, including AKR1B1. Gene knockdown and small molecule inhibition of AKR1B1 can significantly decrease cell proliferation and promote cell apoptosis, suggesting that AKR1B1 plays an important role in tumorigenesis and can serve as a candidate target for tumor therapy of NSCLC patients with mixed-lineage tumor features.ConclusionsIn summary, our work provides novel insights into the tumor heterogeneity of NSCLC in terms of the identification of prevalent mixed-lineage subpopulations of cancer cells with combined signatures of SCC, ADC, and NET and offers clues for potential treatment strategies in these patients.
| S-EPMC9375433 | biostudies-literature
Project description:Genomic profiling efforts have revealed a rich diversity of oncogenic fusion genes. While there are many methods for identifying fusion genes from RNA-sequencing (RNA-seq) data, visualizing these transcripts and their supporting reads remains challenging. Clinker is a bioinformatics tool written in Python, R, and Bpipe that leverages the superTranscript method to visualize fusion genes. We demonstrate the use of Clinker to obtain interpretable visualizations of the RNA-seq data that lead to fusion calls. In addition, we use Clinker to explore multiple fusion transcripts with novel breakpoints within the P2RY8-CRLF2 fusion gene in B-cell acute lymphoblastic leukemia. Clinker is freely available software that allows visualization of fusion genes and the RNA-seq data used in their discovery.
| S-EPMC6065480 | biostudies-literature
Project description:Bulked segregant analysis (BSA) is an efficient method to rapidly and efficiently map genes responsible for mutant phenotypes. BSA requires access to quantitative genetic markers that are polymorphic in the mapping population. We have developed a modification of BSA (BSR-Seq) that makes use of RNA-Seq reads to efficiently map genes even in populations for which no polymorphic markers have been previously identified. Because of the digital nature of next-generation sequencing (NGS) data, it is possible to conduct de novo SNP discovery and quantitatively genotype BSA samples by analyzing the same RNA-Seq data using an empirical Bayesian approach. In addition, analysis of the RNA-Seq data provides information on the effects of the mutant on global patterns of gene expression at no extra cost. In combination these results greatly simplify gene cloning experiments. To demonstrate the utility of this strategy BSR-Seq was used to clone the glossy3 (gl3) gene of maize. Mutants of the glossy loci exhibit altered accumulation of epicuticular waxes on juvenile leaves. By subjecting the reference allele of gl3 to BSR-Seq, we were able to map the gl3 locus to an ? 2 Mb interval. The single gene located in the ? 2 Mb mapping interval whose expression was down-regulated in the mutant pool was subsequently demonstrated to be the gl3 gene via the analysis of multiple independent transposon induced mutant alleles. The gl3 gene encodes a putative myb transcription factor, which directly or indirectly affects the expression of a number of genes involved in the biosynthesis of very-long-chain fatty acids.
| S-EPMC3346754 | biostudies-literature