DriverDBv3: a multi-omics database for cancer driver gene research.

An integrative multi-omics database is needed urgently, because focusing only on analysis of one-dimensional data falls far short of providing an understanding of cancer. Previously, we presented DriverDB, a cancer driver gene database that applies published bioinformatics algorithms to identify driver genes/mutations. The updated DriverDBv3 database (http://ngs.ym.edu.tw/driverdb) is designed to interpret cancer omics' sophisticated information with concise data visualization. To offer diverse insights into molecular dysregulation/dysfunction events, we incorporated computational tools to define CNV and methylation drivers. Further, four new features, CNV, Methylation, Survival, and miRNA, allow users to explore the relations from two perspectives in the 'Cancer' and 'Gene' sections. The 'Survival' panel offers not only significant survival genes, but gene pairs synergistic effects determine. A fresh function, 'Survival Analysis' in 'Customized-analysis,' allows users to investigate the co-occurring events in user-defined gene(s) by mutation status or by expression in a specific patient group. Moreover, we redesigned the web interface and provided interactive figures to interpret cancer omics' sophisticated information, and also constructed a Summary panel in the 'Cancer' and 'Gene' sections to visualize the features on multi-omics levels concisely. DriverDBv3 seeks to improve the study of integrative cancer omics data by identifying driver genes and contributes to cancer biology.

Decoding the differentiation of mesenchymal stem cells into mesangial cells at the transcriptomic level.

BACKGROUND: Mesangial cells play an important role in the glomerulus to provide mechanical support and maintaine efficient ultrafiltration of renal plasma. Loss of mesangial cells due to pathologic conditions may lead to impaired renal function. Mesenchymal stem cells (MSC) can differentiate into many cell types, including mesangial cells. However transcriptomic profiling during MSC differentiation into mesangial cells had not been studied yet. The aim of this study is to examine the pattern of transcriptomic changes during MSC differentiation into mesangial cells, to understand the involvement of transcription factor (TF) along the differentiation process, and finally to elucidate the relationship among TF-TF and TF-key gene or biomarkers during the differentiation of MSC into mesangial cells. RESULTS: Several ascending and descending monotonic key genes were identified by Monotonic Feature Selector. The identified descending monotonic key genes are related to stemness or regulation of cell cycle while ascending monotonic key genes are associated with the functions of mesangial cells. The TFs were arranged in a co-expression network in order of time by Time-Ordered Gene Co-expression Network (TO-GCN) analysis. TO-GCN analysis can classify the differentiation process into three stages: differentiation preparation, differentiation initiation and maturation. Furthermore, it can also explore TF-TF-key genes regulatory relationships in the muscle contraction process. CONCLUSIONS: A systematic analysis for transcriptomic profiling of MSC differentiation into mesangial cells has been established. Key genes or biomarkers, TFs and pathways involved in differentiation of MSC-mesangial cells have been identified and the related biological implications have been discussed. Finally, we further elucidated for the first time the three main stages of mesangial cell differentiation, and the regulatory relationships between TF-TF-key genes involved in the muscle contraction process. Through this study, we have increased fundamental understanding of the gene transcripts during the differentiation of MSC into mesangial cells.

Application of a deep learning algorithm for detection and visualization of hip fractures on plain pelvic radiographs.

OBJECTIVE: To identify the feasibility of using a deep convolutional neural network (DCNN) for the detection and localization of hip fractures on plain frontal pelvic radiographs (PXRs). Hip fracture is a leading worldwide health problem for the elderly. A missed diagnosis of hip fracture on radiography leads to a dismal prognosis. The application of a DCNN to PXRs can potentially improve the accuracy and efficiency of hip fracture diagnosis. METHODS: A DCNN was pretrained using 25,505 limb radiographs between January 2012 and December 2017. It was retrained using 3605 PXRs between August 2008 and December 2016. The accuracy, sensitivity, false-negative rate, and area under the receiver operating characteristic curve (AUC) were evaluated on 100 independent PXRs acquired during 2017. The authors also used the visualization algorithm gradient-weighted class activation mapping (Grad-CAM) to confirm the validity of the model. RESULTS: The algorithm achieved an accuracy of 91%, a sensitivity of 98%, a false-negative rate of 2%, and an AUC of 0.98 for identifying hip fractures. The visualization algorithm showed an accuracy of 95.9% for lesion identification. CONCLUSIONS: A DCNN not only detected hip fractures on PXRs with a low false-negative rate but also had high accuracy for localizing fracture lesions. The DCNN might be an efficient and economical model to help clinicians make a diagnosis without interrupting the current clinical pathway. KEY POINTS: • Automated detection of hip fractures on frontal pelvic radiographs may facilitate emergent screening and evaluation efforts for primary physicians. • Good visualization of the fracture site by Grad-CAM enables the rapid integration of this tool into the current medical system. • The feasibility and efficiency of utilizing a deep neural network have been confirmed for the screening of hip fractures.

YM500v3: a database for small RNA sequencing in human cancer research.

We previously presented the YM500 database, which contains >8000 small RNA sequencing (smRNA-seq) data sets and integrated analysis results for various cancer miRNome studies. In the updated YM500v3 database (http://ngs.ym.edu.tw/ym500/) presented herein, we not only focus on miRNAs but also on other functional small non-coding RNAs (sncRNAs), such as PIWI-interacting RNAs (piRNAs), tRNA-derived fragments (tRFs), small nuclear RNAs (snRNAs) and small nucleolar RNAs (snoRNAs). There is growing knowledge of the role of sncRNAs in gene regulation and tumorigenesis. We have also incorporated >10 000 cancer-related RNA-seq and >3000 more smRNA-seq data sets into the YM500v3 database. Furthermore, there are two main new sections, 'Survival' and 'Cancer', in this updated version. The 'Survival' section provides the survival analysis results in all cancer types or in a user-defined group of samples for a specific sncRNA. The 'Cancer' section provides the results of differential expression analyses, miRNA-gene interactions and cancer miRNA-related pathways. In the 'Expression' section, sncRNA expression profiles across cancer and sample types are newly provided. Cancer-related sncRNAs hold potential for both biotech applications and basic research.

DriverDBv2: a database for human cancer driver gene research.

We previously presented DriverDB, a database that incorporates ∼ 6000 cases of exome-seq data, in addition to annotation databases and published bioinformatics algorithms dedicated to driver gene/mutation identification. The database provides two points of view, 'Cancer' and 'Gene', to help researchers visualize the relationships between cancers and driver genes/mutations. In the updated DriverDBv2 database (http://ngs.ym.edu.tw/driverdb) presented herein, we incorporated >9500 cancer-related RNA-seq datasets and >7000 more exome-seq datasets from The Cancer Genome Atlas (TCGA), International Cancer Genome Consortium (ICGC), and published papers. Seven additional computational algorithms (meaning that the updated database contains 15 in total), which were developed for driver gene identification, are incorporated into our analysis pipeline, and the results are provided in the 'Cancer' section. Furthermore, there are two main new features, 'Expression' and 'Hotspot', in the 'Gene' section. 'Expression' displays two expression profiles of a gene in terms of sample types and mutation types, respectively. 'Hotspot' indicates the hotspot mutation regions of a gene according to the results provided by four bioinformatics tools. A new function, 'Gene Set', allows users to investigate the relationships among mutations, expression levels and clinical data for a set of genes, a specific dataset and clinical features.

Comprehensive evaluation of fusion transcript detection algorithms and a meta-caller to combine top performing methods in paired-end RNA-seq data.

BACKGROUND: Fusion transcripts are formed by either fusion genes (DNA level) or trans-splicing events (RNA level). They have been recognized as a promising tool for diagnosing, subtyping and treating cancers. RNA-seq has become a precise and efficient standard for genome-wide screening of such aberration events. Many fusion transcript detection algorithms have been developed for paired-end RNA-seq data but their performance has not been comprehensively evaluated to guide practitioners. In this paper, we evaluated 15 popular algorithms by their precision and recall trade-off, accuracy of supporting reads and computational cost. We further combine top-performing methods for improved ensemble detection. RESULTS: Fifteen fusion transcript detection tools were compared using three synthetic data sets under different coverage, read length, insert size and background noise, and three real data sets with selected experimental validations. No single method dominantly performed the best but SOAPfuse generally performed well, followed by FusionCatcher and JAFFA. We further demonstrated the potential of a meta-caller algorithm by combining top performing methods to re-prioritize candidate fusion transcripts with high confidence that can be followed by experimental validation. CONCLUSION: Our result provides insightful recommendations when applying individual tool or combining top performers to identify fusion transcript candidates.

YM500v2: a small RNA sequencing (smRNA-seq) database for human cancer miRNome research.

We previously presented YM500, which is an integrated database for miRNA quantification, isomiR identification, arm switching discovery and novel miRNA prediction from 468 human smRNA-seq datasets. Here in this updated YM500v2 database (http://ngs.ym.edu.tw/ym500/), we focus on the cancer miRNome to make the database more disease-orientated. New miRNA-related algorithms developed after YM500 were included in YM500v2, and, more significantly, more than 8000 cancer-related smRNA-seq datasets (including those of primary tumors, paired normal tissues, PBMC, recurrent tumors, and metastatic tumors) were incorporated into YM500v2. Novel miRNAs (miRNAs not included in the miRBase R21) were not only predicted by three independent algorithms but also cleaned by a new in silico filtration strategy and validated by wetlab data such as Cross-Linked ImmunoPrecipitation sequencing (CLIP-seq) to reduce the false-positive rate. A new function 'Meta-analysis' is additionally provided for allowing users to identify real-time differentially expressed miRNAs and arm-switching events according to customer-defined sample groups and dozens of clinical criteria tidying up by proficient clinicians. Cancer miRNAs identified hold the potential for both basic research and biotech applications.

Cell death caused by the synergistic effects of zinc and dopamine is mediated by a stress sensor gene Gadd45b - implication in the pathogenesis of Parkinson's disease.

The pathogenesis of Parkinson's disease (PD) is not completely understood, Zinc (Zn(2+) ) and dopamine (DA) have been shown to involve in the degeneration of dopaminergic cells. By microarray analysis, we identified Gadd45b as a candidate molecule that mediates Zn(2+) and DA-induced cell death; the mRNA and protein levels of Gadd45b are increased by Zn(2+) treatment and raised to an even higher level by Zn(2+) plus DA treatment. Zn(2+) plus DA treatment-induced PC12 cell death was enhanced when there was over-expression of Gadd45b and was decreased by knock down of Gadd45b. MAPK p38 and JNK signaling was able to cross-talk with Gadd45b during Zn(2+) and DA treatment. The synergistic effects of Zn(2+) and DA on PC12 cell death can be accounted for by an activation of the Gadd45b-induced cell death pathway and an inhibition of p38/JNK survival pathway. Furthermore, the in vivo results show that the levels of Gadd45b protein expression and phosphorylation of p38 were increased in the substantia nigra by the infusion of Zn(2+) /DA in the mouse brain and the level of Gadd45b mRNA is significantly higher in the substantia nigra of male PD patients than normal controls. The novel role of Gadd45b and its interactions with JNK and p38 will help our understanding of the pathogenesis of PD and help the development of future treatments for PD. Zinc and dopamine are implicated in the degeneration of dopaminergic neurons. We previously demonstrated that zinc and dopamine induced synergistic effects on PC12 cell death. Results from this study show that these synergistic effects can be accounted for by activation of the Gadd45b-induced cell death pathway and inhibition of the p38/JNK survival pathway. We provide in vitro and in vivo evidence to support a novel role for Gadd45b in the pathogenesis of Parkinson's disease.

DriverDB: an exome sequencing database for cancer driver gene identification.

Exome sequencing (exome-seq) has aided in the discovery of a huge amount of mutations in cancers, yet challenges remain in converting oncogenomics data into information that is interpretable and accessible for clinical care. We constructed DriverDB (http://ngs.ym.edu.tw/driverdb/), a database which incorporates 6079 cases of exome-seq data, annotation databases (such as dbSNP, 1000 Genome and Cosmic) and published bioinformatics algorithms dedicated to driver gene/mutation identification. We provide two points of view, 'Cancer' and 'Gene', to help researchers to visualize the relationships between cancers and driver genes/mutations. The 'Cancer' section summarizes the calculated results of driver genes by eight computational methods for a specific cancer type/dataset and provides three levels of biological interpretation for realization of the relationships between driver genes. The 'Gene' section is designed to visualize the mutation information of a driver gene in five different aspects. Moreover, a 'Meta-Analysis' function is provided so researchers may identify driver genes in customer-defined samples. The novel driver genes/mutations identified hold potential for both basic research and biotech applications.

Discovering monotonic stemness marker genes from time-series stem cell microarray data.

BACKGROUND: Identification of genes with ascending or descending monotonic expression patterns over time or stages of stem cells is an important issue in time-series microarray data analysis. We propose a method named Monotonic Feature Selector (MFSelector) based on a concept of total discriminating error (DEtotal) to identify monotonic genes. MFSelector considers various time stages in stage order (i.e., Stage One vs. other stages, Stages One and Two vs. remaining stages and so on) and computes DEtotal of each gene. MFSelector can successfully identify genes with monotonic characteristics. RESULTS: We have demonstrated the effectiveness of MFSelector on two synthetic data sets and two stem cell differentiation data sets: embryonic stem cell neurogenesis (ESCN) and embryonic stem cell vasculogenesis (ESCV) data sets. We have also performed extensive quantitative comparisons of the three monotonic gene selection approaches. Some of the monotonic marker genes such as OCT4, NANOG, BLBP, discovered from the ESCN dataset exhibit consistent behavior with that reported in other studies. The role of monotonic genes found by MFSelector in either stemness or differentiation is validated using information obtained from Gene Ontology analysis and other literature. We justify and demonstrate that descending genes are involved in the proliferation or self-renewal activity of stem cells, while ascending genes are involved in differentiation of stem cells into variant cell lineages. CONCLUSIONS: We have developed a novel system, easy to use even with no pre-existing knowledge, to identify gene sets with monotonic expression patterns in multi-stage as well as in time-series genomics matrices. The case studies on ESCN and ESCV have helped to get a better understanding of stemness and differentiation. The novel monotonic marker genes discovered from a data set are found to exhibit consistent behavior in another independent data set, demonstrating the utility of the proposed method. The MFSelector R function and data sets can be downloaded from: http://microarray.ym.edu.tw/tools/MFSelector/.

YM500: a small RNA sequencing (smRNA-seq) database for microRNA research.

MicroRNAs (miRNAs) are small RNAs ∼22 nt in length that are involved in the regulation of a variety of physiological and pathological processes. Advances in high-throughput small RNA sequencing (smRNA-seq), one of the next-generation sequencing applications, have reshaped the miRNA research landscape. In this study, we established an integrative database, the YM500 (http://ngs.ym.edu.tw/ym500/), containing analysis pipelines and analysis results for 609 human and mice smRNA-seq results, including public data from the Gene Expression Omnibus (GEO) and some private sources. YM500 collects analysis results for miRNA quantification, for isomiR identification (incl. RNA editing), for arm switching discovery, and, more importantly, for novel miRNA predictions. Wetlab validation on >100 miRNAs confirmed high correlation between miRNA profiling and RT-qPCR results (R = 0.84). This database allows researchers to search these four different types of analysis results via our interactive web interface. YM500 allows researchers to define the criteria of isomiRs, and also integrates the information of dbSNP to help researchers distinguish isomiRs from SNPs. A user-friendly interface is provided to integrate miRNA-related information and existing evidence from hundreds of sequencing datasets. The identified novel miRNAs and isomiRs hold the potential for both basic research and biotech applications.

Deep learning-based glomerulus detection and classification with generative morphology augmentation in renal pathology images.

Glomerulus morphology on renal pathology images provides valuable diagnosis and outcome prediction information. To provide better care, an efficient, standardized, and scalable method is urgently needed to optimize the time-consuming and labor-intensive interpretation process by renal pathologists. This paper proposes a deep convolutional neural network (CNN)-based approach to automatically detect and classify glomeruli with different stains in renal pathology images. In the glomerulus detection stage, this paper proposes a flattened Xception with a feature pyramid network (FX-FPN). The FX-FPN is employed as a backbone in the framework of faster region-based CNN to improve glomerulus detection performance. In the classification stage, this paper considers classifications of five glomerulus morphologies using a flattened Xception classifier. To endow the classifier with higher discriminability, this paper proposes a generative data augmentation approach for patch-based glomerulus morphology augmentation. New glomerulus patches of different morphologies are generated for data augmentation through the cycle-consistent generative adversarial network (CycleGAN). The single detection model shows the F1 score up to 0.9524 in H&E and PAS stains. The classification result shows that the average sensitivity and specificity are 0.7077 and 0.9316, respectively, by using the flattened Xception with the original training data. The sensitivity and specificity increase to 0.7623 and 0.9443, respectively, by using the generative data augmentation. Comparisons with different deep CNN models show the effectiveness and superiority of the proposed approach.

Predicting long-term time to cardiovascular incidents using myocardial perfusion imaging and deep convolutional neural networks.

Myocardial perfusion imaging (MPI) is a clinical tool which can assess the heart's perfusion status, thereby revealing impairments in patients' cardiac function. Within the MPI modality, the acquired three-dimensional signals are typically represented as a sequence of two-dimensional grayscale tomographic images. Here, we proposed an end-to-end survival training approach for processing gray-scale MPI tomograms to generate a risk score which reflects subsequent time to cardiovascular incidents, including cardiovascular death, non-fatal myocardial infarction, and non-fatal ischemic stroke (collectively known as Major Adverse Cardiovascular Events; MACE) as well as Congestive Heart Failure (CHF). We recruited a total of 1928 patients who had undergone MPI followed by coronary interventions. Among them, 80% (n = 1540) were randomly reserved for the training and 5- fold cross-validation stage, while 20% (n = 388) were set aside for the testing stage. The end-to-end survival training can converge well in generating effective AI models via the fivefold cross-validation approach with 1540 patients. When a candidate model is evaluated using independent images, the model can stratify patients into below-median-risk (n = 194) and above-median-risk (n = 194) groups, the corresponding survival curves of the two groups have significant difference (P < 0.0001). We further stratify the above-median-risk group to the quartile 3 and 4 group (n = 97 each), and the three patient strata, referred to as the high, intermediate and low risk groups respectively, manifest statistically significant difference. Notably, the 5-year cardiovascular incident rate is less than 5% in the low-risk group (accounting for 50% of all patients), while the rate is nearly 40% in the high-risk group (accounting for 25% of all patients). Evaluation of patient subgroups revealed stronger effect size in patients with three blocked arteries (Hazard ratio [HR]: 18.377, 95% CI 3.719-90.801, p < 0.001), followed by those with two blocked vessels at HR 7.484 (95% CI 1.858-30.150; p = 0.005). Regarding stent placement, patients with a single stent displayed a HR of 4.410 (95% CI 1.399-13.904; p = 0.011). Patients with two stents show a HR of 10.699 (95% CI 2.262-50.601; p = 0.003), escalating notably to a HR of 57.446 (95% CI 1.922-1717.207; p = 0.019) for patients with three or more stents, indicating a substantial relationship between the disease severity and the predictive capability of the AI for subsequent cardiovascular inciidents. The success of the MPI AI model in stratifying patients into subgroups with distinct time-to-cardiovascular incidents demonstrated the feasibility of proposed end-to-end survival training approach.

Deep Learning for Automated Detection and Localization of Traumatic Abdominal Solid Organ Injuries on CT Scans.

Computed tomography (CT) is the most commonly used diagnostic modality for blunt abdominal trauma (BAT), significantly influencing management approaches. Deep learning models (DLMs) have shown great promise in enhancing various aspects of clinical practice. There is limited literature available on the use of DLMs specifically for trauma image evaluation. In this study, we developed a DLM aimed at detecting solid organ injuries to assist medical professionals in rapidly identifying life-threatening injuries. The study enrolled patients from a single trauma center who received abdominal CT scans between 2008 and 2017. Patients with spleen, liver, or kidney injury were categorized as the solid organ injury group, while others were considered negative cases. Only images acquired from the trauma center were enrolled. A subset of images acquired in the last year was designated as the test set, and the remaining images were utilized to train and validate the detection models. The performance of each model was assessed using metrics such as the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value based on the best Youden index operating point. The study developed the models using 1302 (87%) scans for training and tested them on 194 (13%) scans. The spleen injury model demonstrated an accuracy of 0.938 and a specificity of 0.952. The accuracy and specificity of the liver injury model were reported as 0.820 and 0.847, respectively. The kidney injury model showed an accuracy of 0.959 and a specificity of 0.989. We developed a DLM that can automate the detection of solid organ injuries by abdominal CT scans with acceptable diagnostic accuracy. It cannot replace the role of clinicians, but we can expect it to be a potential tool to accelerate the process of therapeutic decisions for trauma care.

The three-dimensional weakly supervised deep learning algorithm for traumatic splenic injury detection and sequential localization: an experimental study.

BACKGROUND: Splenic injury is the most common solid visceral injury in blunt abdominal trauma, and high-resolution abdominal computed tomography (CT) can adequately detect the injury. However, these lethal injuries sometimes have been overlooked in current practice. Deep learning (DL) algorithms have proven their capabilities in detecting abnormal findings in medical images. The aim of this study is to develop a three-dimensional, weakly supervised DL algorithm for detecting splenic injury on abdominal CT using a sequential localization and classification approach. MATERIAL AND METHODS: The dataset was collected in a tertiary trauma center on 600 patients who underwent abdominal CT between 2008 and 2018, half of whom had splenic injuries. The images were split into development and test datasets at a 4 : 1 ratio. A two-step DL algorithm, including localization and classification models, was constructed to identify the splenic injury. Model performance was evaluated using the area under the receiver operating characteristic curve (AUROC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Grad-CAM (Gradient-weighted Class Activation Mapping) heatmaps from the test set were visually assessed. To validate the algorithm, we also collected images from another hospital to serve as external validation data. RESULTS: A total of 480 patients, 50% of whom had spleen injuries, were included in the development dataset, and the rest were included in the test dataset. All patients underwent contrast-enhanced abdominal CT in the emergency room. The automatic two-step EfficientNet model detected splenic injury with an AUROC of 0.901 (95% CI: 0.836-0.953). At the maximum Youden index, the accuracy, sensitivity, specificity, PPV, and NPV were 0.88, 0.81, 0.92, 0.91, and 0.83, respectively. The heatmap identified 96.3% of splenic injury sites in true positive cases. The algorithm achieved a sensitivity of 0.92 for detecting trauma in the external validation cohort, with an acceptable accuracy of 0.80. CONCLUSIONS: The DL model can identify splenic injury on CT, and further application in trauma scenarios is possible.

Generating transcriptional regulatory networks from time-ordered stem cell differentiation RNA sequencing data.

We describe steps to 1) identify ascending and descending monotonic key genes from time-ordered stem cell differentiation expression data, 2) construct time-ordered transcriptional regulatory networks, and 3) infer the involvement of transcription factors along the differentiation process. For complete details on the use and execution of this protocol, please refer to Wong et al. (2020).

lncExplore: a database of pan-cancer analysis and systematic functional annotation for lncRNAs from RNA-sequencing data.

Over the past few years, with the rapid growth of deep-sequencing technology and the development of computational prediction algorithms, a large number of long non-coding RNAs (lncRNAs) have been identified in various types of human cancers. Therefore, it has become critical to determine how to properly annotate the potential function of lncRNAs from RNA-sequencing (RNA-seq) data and arrange the robust information and analysis into a useful system readily accessible by biological and clinical researchers. In order to produce a collective interpretation of lncRNA functions, it is necessary to integrate different types of data regarding the important functional diversity and regulatory role of these lncRNAs. In this study, we utilized transcriptomic sequencing data to systematically observe and identify lncRNAs and their potential functions from 5034 The Cancer Genome Atlas RNA-seq datasets covering 24 cancers. Then, we constructed the 'lncExplore' database that was developed to comprehensively integrate various types of genomic annotation data for collective interpretation. The distinctive features in our lncExplore database include (i) novel lncRNAs verified by both coding potential and translation efficiency score, (ii) pan-cancer analysis for studying the significantly aberrant expression across 24 human cancers, (iii) genomic annotation of lncRNAs, such as cis-regulatory information and gene ontology, (iv) observation of the regulatory roles as enhancer RNAs and competing endogenous RNAs and (v) the findings of the potential lncRNA biomarkers for the user-interested cancers by integrating clinical information and disease specificity score. The lncExplore database is to our knowledge the first public lncRNA annotation database providing cancer-specific lncRNA expression profiles for not only known but also novel lncRNAs, enhancer RNAs annotation and clinical analysis based on pan-cancer analysis. lncExplore provides a more complete pathway to highly efficient, novel and more comprehensive translation of laboratory discoveries into the clinical context and will assist in reinterpreting the biological regulatory function of lncRNAs in cancer research. Database URL http://lncexplore.bmi.nycu.edu.tw.

Comparison of Genetic Profiling between Primary Tumor and Circulating Tumor Cells Captured by Microfluidics in Epithelial Ovarian Cancer: Tumor Heterogeneity or Allele Dropout?

Epithelial ovarian cancer (EOC) is a leading cause of cancer mortality among women but unfortunately is usually not diagnosed until advanced stage. Early detection of EOC is of paramount importance to improve outcomes. Liquid biopsy of circulating tumor cells (CTCs) is emerging as one of the promising biomarkers for early detection of solid tumors. However, discrepancies in terms of oncogenomics (i.e., different genetic defects detected) between the germline, primary tumor, and liquid biopsy are a serious concern and may adversely affect downstream cancer management. Here, we illustrate the potential and pitfalls of CTCs by presenting two patients of Stage I EOC. We successfully isolated and recovered CTCs by a silicon-based nanostructured microfluidics system, the automated Cell RevealTM. We examined the genomics of CTCs as well as the primary tumor and germline control (peripheral blood mononuclear cells) by whole exome sequencing. Different signatures were then investigated by comparisons of identified mutation loci distinguishing those that may only arise in the primary tumor or CTCs. A novel model is proposed to test if the highly variable allele frequencies, between primary tumor and CTCs results, are due to allele dropout in plural CTCs or tumor heterogeneity. This proof-of-principle study provides a strategy to elucidate the possible cause of genomic discrepancy between the germline, primary tumor, and CTCs, which is helpful for further large-scale use of such technology to be integrated into clinical management protocols.

A Human-Algorithm Integration System for Hip Fracture Detection on Plain Radiography: System Development and Validation Study.

BACKGROUND: Hip fracture is the most common type of fracture in elderly individuals. Numerous deep learning (DL) algorithms for plain pelvic radiographs (PXRs) have been applied to improve the accuracy of hip fracture diagnosis. However, their efficacy is still undetermined. OBJECTIVE: The objective of this study is to develop and validate a human-algorithm integration (HAI) system to improve the accuracy of hip fracture diagnosis in a real clinical environment. METHODS: The HAI system with hip fracture detection ability was developed using a deep learning algorithm trained on trauma registry data and 3605 PXRs from August 2008 to December 2016. To compare their diagnostic performance before and after HAI system assistance using an independent testing dataset, 34 physicians were recruited. We analyzed the physicians' accuracy, sensitivity, specificity, and agreement with the algorithm; we also performed subgroup analyses according to physician specialty and experience. Furthermore, we applied the HAI system in the emergency departments of different hospitals to validate its value in the real world. RESULTS: With the support of the algorithm, which achieved 91% accuracy, the diagnostic performance of physicians was significantly improved in the independent testing dataset, as was revealed by the sensitivity (physician alone, median 95%; HAI, median 99%; P<.001), specificity (physician alone, median 90%; HAI, median 95%; P<.001), accuracy (physician alone, median 90%; HAI, median 96%; P<.001), and human-algorithm agreement [physician alone κ, median 0.69 (IQR 0.63-0.74); HAI κ, median 0.80 (IQR 0.76-0.82); P<.001. With the help of the HAI system, the primary physicians showed significant improvement in their diagnostic performance to levels comparable to those of consulting physicians, and both the experienced and less-experienced physicians benefited from the HAI system. After the HAI system had been applied in 3 departments for 5 months, 587 images were examined. The sensitivity, specificity, and accuracy of the HAI system for detecting hip fractures were 97%, 95.7%, and 96.08%, respectively. CONCLUSIONS: HAI currently impacts health care, and integrating this technology into emergency departments is feasible. The developed HAI system can enhance physicians' hip fracture diagnostic performance.

Artificial intelligence-based education assists medical students' interpretation of hip fracture.

BACKGROUND: With recent transformations in medical education, the integration of technology to improve medical students' abilities has become feasible. Artificial intelligence (AI) has impacted several aspects of healthcare. However, few studies have focused on medical education. We performed an AI-assisted education study and confirmed that AI can accelerate trainees' medical image learning. MATERIALS: We developed an AI-based medical image learning system to highlight hip fracture on a plain pelvic film. Thirty medical students were divided into a conventional (CL) group and an AI-assisted learning (AIL) group. In the CL group, the participants received a prelearning test and a postlearning test. In the AIL group, the participants received another test with AI-assisted education before the postlearning test. Then, we analyzed changes in diagnostic accuracy. RESULTS: The prelearning performance was comparable in both groups. In the CL group, postlearning accuracy (78.66 ± 14.53) was higher than prelearning accuracy (75.86 ± 11.36) with no significant difference (p = .264). The AIL group showed remarkable improvement. The WithAI score (88.87 ± 5.51) was significantly higher than the prelearning score (75.73 ± 10.58, p < 0.01). Moreover, the postlearning score (84.93 ± 14.53) was better than the prelearning score (p < 0.01). The increase in accuracy was significantly higher in the AIL group than in the CL group. CONCLUSION: The study demonstrated the viability of AI for augmenting medical education. Integrating AI into medical education requires dynamic collaboration from research, clinical, and educational perspectives.