Achieved AUC Values Research Articles

Advances in radiological studies of gastrointestinal stromal tumors

The commonly used radiological examinations for gastrointestinal stromal tumor (GIST) include ultrasound, CT, magnetic resonance imaging and PET, which play important roles in the individualized treatment of GIST through the assistance of the detection, location, diagnose and differential diagnosis, biological behavior and risk evaluation, operation planning, therapeutic effects prediction and follow-up. Standardized predisposal, multi-planar reconstruction and multi-phase enhancement are of great significance in the detection of small GIST lesions. Differences in drainage vessels and enhancement characteristics of tumors are helpful in differentiating small intestinal GIST from pancreatic cancer, neuroendocrine tumors and other abdominal non-GIST tumors. Radiology can provide multiple indicators to qualitatively or quantitatively reflect the risk of GIST. Recent study found that the risk classification of GIST has correlation with necrosis, but not cystic degeneration. The risk stratification of GIST by dual-energy CT showed that CT value and standardized iodine concentration in arterial-venous and delayed phases of high-risk tumors were significantly higher than those of non-high-risk tumors. The correlation between tumor enhancement pattern and risk stratificationwas still controversial. Recently, the rise of radiomics and deep learning techniques provide new direction for risk assessment of GIST. Some studies used various models and parameters to predict the risk classifications of GIST and achieved AUC values above 0.85, and some results were also superior to conventional indicators and radiologists' subjective judgment in the prediction of exon 11 mutations, mitotic counts and survivals. Recent developments in PET and diffusion-weighted MRI provide more precise functional quantitative indicators for the prediction and early evaluation of the response of GIST to targeted therapy. These studies are constantly expanding our field of exploration, and prompting us to stride toward the goal of individualized treatment of GIST.

Intestinal microbiota dysbiosis in children with recurrent respiratory tract infections

BackgroundThe impact of the gut microbiota on recurrent respiratory tract infection (RRTI) remains to be fully elucidated. MethodsTo characterize the gut microbiota in patients with RRTI, fecal samples from 26 patients with RRTI and 23 healthy volunteers were profiled using the Illumina MiSeq platform. Beta diversity (Principal Component Analysis (PCA), Principal Co-ordinates Analysis (PCoA), Non-metric multidimensional scaling (NMDS)) analysis showed that the bacterial community structure segregated differently between the RRTI and control groups. ResultsResults from alpha diversity analysis revealed lower microbiota diversity in samples from RRTI patients than in normal controls. Taxonomic analysis illustrated that the abundance of six phyla (Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, Verrucomicrobia, Tenericutes) and four genera (Enterococcus, Faecalibacterium, Bifidobacterium, Eubacterium were significantly different between these two groups. In addition, Enterococcus (P < 0.001) was more enriched in the RRTI group, whereas the abundances of Eubacterium (P < 0.001), Faecalibacterium (0.01 < P < 0.05) and Bifidobacterium (0.01 < P < 0.05) were reduced in the RRTI group compared to those in the normal control group. The performance of the model was assessed using ROC analysis, and Enterococcus, Eubacterium and Bifidobacterium achieved AUC values of 0.860, 0.820, and 0.689, respectively. ConclusionsThese results provide fundamental evidence in support of intestinal microbiota dysbiosis in children with RRTI.

A CNN-BASED RETINAL IMAGE QUALITY ASSESSMENT SYSTEM FOR TELEOPHTHALMOLOGY

Oculopathy is a widespread disease among people of all ages around the world. Teleophthalmology can facilitate the ophthalmological diagnosis for less developed countries that lack medical resources. In teleophthalmology, the assessment of retinal image quality is of great importance. In this paper, we propose a no-reference retinal image assessment system based on DenseNet, a convolutional neural network architecture. This system classified fundus images into good and bad quality or five categories: adequate, just noticeable blur, inappropriate illumination, incomplete optic disc, and opacity. The proposed system was evaluated on different datasets and compared to the applications based on other two networks: VGG-16 and GoogLenet. For binary classification, the good-and-bad binary classifier achieves an AUC of 1.000, and the degradation-specified classifiers that distinguish one specified degradation versus the rest achieve AUC values of 0.972, 0.990, 0.982, 0.982 for four categories, respectively. The multi-classification based on DenseNet achieves an overall accuracy of 0.927, which is significantly higher than 0.549 and 0.757 obtained using VGG-16 and GoogLeNet, respectively. The experimental results indicate that the proposed approach produces outstanding performance in retinal image quality assessment and is worth applying in ophthalmological telemedicine applications. In addition, the proposed approach is robust to the image noise. This study fills the gap of multi-classification in retinal image quality assessment.

A Robust 8-Gene Prognostic Signature for Early-Stage Non-small Cell Lung Cancer.

Background: The current staging system is imprecise for prognostic prediction of early-stage non–small cell lung cancer (NSCLC). This study aimed to develop a robust prognostic signature for early-stage NSCLC, allowing classification of patients with a high risk of poor outcome and specific treatment decision.Method: In the present study, a comprehensive genome-wide profiling analysis was conducted using a retrospective pool of early-stage NSCLC patient data from the previous datasets of Gene Expression Omnibus (GEO) including GSE31210, GSE37745, and GSE50081 and The Cancer Genome Atlas (TCGA). Cox proportional hazards models were implemented to determine the association between gene expression levels and overall patient survival in each dataset. The common genes among all datasets were selected as candidate prognostic genes. A risk score model was developed and validated using four independent datasets and the entire cohort. The Kaplan-Meier with log-rank test was used to assess survival difference.Results: A univariate Cox proportional hazards regression analysis for each dataset showed that a total of 2280 genes in GSE31210, 762 genes in GSE37745, 871 genes in GSE50081, and 666 genes in TCGA were identified as candidate protective genes, while overall 2131 genes in GSE31210, 913 in GSE37745, 1107 in GSE50081, and 997 in TCGA were identified as candidate risky genes. There were 8 common genes associated with overall survival, including 7 mRNA and 1 lncRNA. By using the Step-wise multivariate Cox analysis, an 8-gene prognostic signature (CDCP1, HMMR, TPX2, CIRBP, HLF, KBTBD7, SEC24B-AS1, and SH2B1) for early-stage NSCLC was developed. Patients in the high-risk group had shorter overall survival than those in the low-risk group. Multivariate regression and stratified analysis suggested that the prognostic power of the 8-gene signature was independent of other clinical factors. Furthermore, the 8-gene signature achieved AUC values of 0.726, 0.701, 0.725 and 0.650 in GSE31210, GSE37745, GSE50081 and TCGA, respectively. Moreover, the combination of the 8-gene signature and the stage resulted to a better patient classification for survival prediction and treatment decision.Conclusion: This study developed a robust gene signature with great value for prognostic prediction in early-stage NSCLC, which may contribute to patient classification and personalized treatment decisions.

Efficient Implementation of Penalized Regression for Genetic Risk Prediction.

Polygenic Risk Scores (PRS) combine genotype information across many single-nucleotide polymorphisms (SNPs) to give a score reflecting the genetic risk of developing a disease. PRS might have a major impact on public health, possibly allowing for screening campaigns to identify high-genetic risk individuals for a given disease. The "Clumping+Thresholding" (C+T) approach is the most common method to derive PRS. C+T uses only univariate genome-wide association studies (GWAS) summary statistics, which makes it fast and easy to use. However, previous work showed that jointly estimating SNP effects for computing PRS has the potential to significantly improve the predictive performance of PRS as compared to C+T. In this paper, we present an efficient method for the joint estimation of SNP effects using individual-level data, allowing for practical application of penalized logistic regression (PLR) on modern datasets including hundreds of thousands of individuals. Moreover, our implementation of PLR directly includes automatic choices for hyper-parameters. We also provide an implementation of penalized linear regression for quantitative traits. We compare the performance of PLR, C+T and a derivation of random forests using both real and simulated data. Overall, we find that PLR achieves equal or higher predictive performance than C+T in most scenarios considered, while being scalable to biobank data. In particular, we find that improvement in predictive performance is more pronounced when there are few effects located in nearby genomic regions with correlated SNPs; for instance, in simulations, AUC values increase from 83% with the best prediction of C+T to 92.5% with PLR. We confirm these results in a data analysis of a case-control study for celiac disease where PLR and the standard C+T method achieve AUC values of 89% and of 82.5%. Applying penalized linear regression to 350,000 individuals of the UK Biobank, we predict height with a larger correlation than with the best prediction of C+T (∼65% instead of ∼55%), further demonstrating its scalability and strong predictive power, even for highly polygenic traits. Moreover, using 150,000 individuals of the UK Biobank, we are able to predict breast cancer better than C+T, fitting PLR in a few minutes only. In conclusion, this paper demonstrates the feasibility and relevance of using penalized regression for PRS computation when large individual-level datasets are available, thanks to the efficient implementation available in our R package bigstatsr.

Ontology based text mining of gene-phenotype associations: application to candidate gene prediction.

Gene–phenotype associations play an important role in understanding the disease mechanisms which is a requirement for treatment development. A portion of gene–phenotype associations are observed mainly experimentally and made publicly available through several standard resources such as MGI. However, there is still a vast amount of gene–phenotype associations buried in the biomedical literature. Given the large amount of literature data, we need automated text mining tools to alleviate the burden in manual curation of gene–phenotype associations and to develop comprehensive resources. In this study, we present an ontology-based approach in combination with statistical methods to text mine gene–phenotype associations from the literature. Our method achieved AUC values of 0.90 and 0.75 in recovering known gene–phenotype associations from HPO and MGI respectively. We posit that candidate genes and their relevant diseases should be expressed with similar phenotypes in publications. Thus, we demonstrate the utility of our approach by predicting disease candidate genes based on the semantic similarities of phenotypes associated with genes and diseases. To the best of our knowledge, this is the first study using an ontology based approach to extract gene–phenotype associations from the literature. We evaluated our disease candidate prediction model on the gene–disease associations from MGI. Our model achieved AUC values of 0.90 and 0.87 on OMIM (human) and MGI (mouse) datasets of gene–disease associations respectively. Our manual analysis on the text mined data revealed that our method can accurately extract gene–phenotype associations which are not currently covered by the existing public gene–phenotype resources. Overall, results indicate that our method can precisely extract known as well as new gene–phenotype associations from literature. All the data and methods are available at https://github.com/bio-ontology-research-group/genepheno.

A Machine-Learning Approach for Detection and Quantification of QRS Fragmentation.

Fragmented QRS (fQRS) is an accessible biomarker and indication of myocardial scarring that can be detected from the electrocardiogram (ECG). Nowadays, fQRS scoring is done on a visual basis, which is time consuming and leads to subjective results. This study proposes an automated method to detect and quantify fQRS in a continuous way using features extracted from variational mode decomposition (VMD) and phase-rectified signal averaging (PRSA). In the proposed framework, QRS complexes in the ECG signals were first segmented using VMD. Then, ten VMD- and PRSA-based features were computed and fed into well-known classifiers such as support vector machine (SVM), K-nearest neighbors (KNN), Naive Bayesian (NB), and TreeBagger (TB) in order to compare their performance. The proposed method was evaluated with 12-lead ECG data of 616 patients from the University Hospitals Leuven. The presence of fQRS in each ECG lead was scored by five raters. Both detection and quantification of fQRS could be achieved in this way. The experimental results indicated that the proposed method achieved AUC values of 0.95, 0.94, 0.90, and 0.89 using SVM, KNN, NB, and TB classifiers, respectively, for detecting QRS fragmentation. Assessment of quantification performance was done by comparing the fQRS score with the total score, obtained by summing the scores from the individual raters. Results showed that the fQRS score clearly correlated with this estimate of fQRS certainty. The proposed method obtained good results in both fQRS detection and quantification, and is a novel way of assessing the certainty of QRS fragmentation in the ECG signal.

2018 update to the HIV-TRePS system: the development of new computational models to predict HIV treatment outcomes, with or without a genotype, with enhanced usability for low-income settings.

Optimizing antiretroviral drug combination on an individual basis can be challenging, particularly in settings with limited access to drugs and genotypic resistance testing. Here we describe our latest computational models to predict treatment responses, with or without a genotype, and compare their predictive accuracy with that of genotyping. Random forest models were trained to predict the probability of virological response to a new therapy introduced following virological failure using up to 50 000 treatment change episodes (TCEs) without a genotype and 18 000 TCEs including genotypes. Independent data sets were used to evaluate the models. This study tested the effects on model accuracy of relaxing the baseline data timing windows, the use of a new filter to exclude probable non-adherent cases and the addition of maraviroc, tipranavir and elvitegravir to the system. The no-genotype models achieved area under the receiver operator characteristic curve (AUC) values of 0.82 and 0.81 using the standard and relaxed baseline data windows, respectively. The genotype models achieved AUC values of 0.86 with the new non-adherence filter and 0.84 without. Both sets of models were significantly more accurate than genotyping with rules-based interpretation, which achieved AUC values of only 0.55-0.63, and were marginally more accurate than previous models. The models were able to identify alternative regimens that were predicted to be effective for the vast majority of cases in which the new regimen prescribed in the clinic failed. These latest global models predict treatment responses accurately even without a genotype and have the potential to help optimize therapy, particularly in resource-limited settings.

Liposomal Irinotecan Accumulates in Metastatic Lesions, Crosses the Blood-Tumor Barrier (BTB), and Prolongs Survival in an Experimental Model of Brain Metastases of Triple Negative Breast Cancer.

The blood-tumor barrier (BTB) limits irinotecan distribution in tumors of the central nervous system. However, given that the BTB has increased passive permeability we hypothesize that liposomal irinotecan would improve local exposure of irinotecan and its active metabolite SN-38 in brain metastases relative to conventional irinotecan due to enhanced-permeation and retention (EPR) effect. Female nude mice were intracardially or intracranially implanted with human brain seeking breast cancer cells (brain metastases of breast cancer model). Mice were administered vehicle, non-liposomal irinotecan (50 mg/kg), liposomal irinotecan (10 mg/kg and 50 mg/kg) intravenously starting on day 21. Drug accumulation, tumor burden, and survival were evaluated. Liposomal irinotecan showed prolonged plasma drug exposure with mean residence time (MRT) of 17.7 ± 3.8 h for SN-38, whereas MRT was 3.67 ± 1.2 for non-liposomal irinotecan. Further, liposomal irinotecan accumulated in metastatic lesions and demonstrated prolonged exposure of SN-38 compared to non-liposomal irinotecan. Liposomal irinotecan achieved AUC values of 6883 ± 4149 ng-h/g for SN-38, whereas non-liposomal irinotecan showed significantly lower AUC values of 982 ± 256 ng-h/g for SN-38. Median survival for liposomal irinotecan was 50 days, increased from 37 days (p<0.05) for vehicle. Liposomal irinotecan accumulates in brain metastases, acts as depot for sustained release of irinotecan and SN-38, which results in prolonged survival in preclinical model of breast cancer brain metastasis.

Investigation of metabolites for estimating blood deposition time

Trace deposition timing reflects a novel concept in forensic molecular biology involving the use of rhythmic biomarkers for estimating the time within a 24-h day/night cycle a human biological sample was left at the crime scene, which in principle allows verifying a sample donor’s alibi. Previously, we introduced two circadian hormones for trace deposition timing and recently demonstrated that messenger RNA (mRNA) biomarkers significantly improve time prediction accuracy. Here, we investigate the suitability of metabolites measured using a targeted metabolomics approach, for trace deposition timing. Analysis of 171 plasma metabolites collected around the clock at 2-h intervals for 36 h from 12 male participants under controlled laboratory conditions identified 56 metabolites showing statistically significant oscillations, with peak times falling into three day/night time categories: morning/noon, afternoon/evening and night/early morning. Time prediction modelling identified 10 independently contributing metabolite biomarkers, which together achieved prediction accuracies expressed as AUC of 0.81, 0.86 and 0.90 for these three time categories respectively. Combining metabolites with previously established hormone and mRNA biomarkers in time prediction modelling resulted in an improved prediction accuracy reaching AUCs of 0.85, 0.89 and 0.96 respectively. The additional impact of metabolite biomarkers, however, was rather minor as the previously established model with melatonin, cortisol and three mRNA biomarkers achieved AUC values of 0.88, 0.88 and 0.95 for the same three time categories respectively. Nevertheless, the selected metabolites could become practically useful in scenarios where RNA marker information is unavailable such as due to RNA degradation. This is the first metabolomics study investigating circulating metabolites for trace deposition timing, and more work is needed to fully establish their usefulness for this forensic purpose.

Spatial prediction of wildfire probability in the Hyrcanian ecoregion using evidential belief function model and GIS

Accurate estimates of wildfire probability and production of distribution maps are the first important steps in wildfire management and risk assessment. In this study, geographical information system (GIS)-automated techniques were integrated with the quantitative data-driven evidential belief function (EBF) model to predict spatial pattern of wildfire probability in a part of the Hyrcanian ecoregion, northern Iran. The historical fire events were identified using earlier reports and MODIS hot spot product as well as by carrying out multiple field surveys. Using the GIS-based EBF model, the relationships among existing fire events and various predictor variables predisposing fire ignition were analyzed. Model results were used to produce a distribution map of wildfire probability. The derived probability map revealed that zones of moderate, high, and very high probability covered nearly 60% of the landscape. Further, the probability map clearly demonstrated that the probability of a fire was strongly dependent upon human infrastructure and associated activities. By comparing the probability map and the historical fire events, a satisfactory spatial agreement between the five probability levels and fire density was observed. The probability map was further validated by receiver operating characteristic using both success rate and prediction rate curves. The validation results confirmed the effectiveness of the GIS-based EBF model that achieved AUC values of 84.14 and 81.03% for success and prediction rates, respectively.

Prediction of Protein Phosphorylation Sites by Integrating Secondary Structure Information and Other One-Dimensional Structural Properties.

Studies on phosphorylation are important but challenging for both wet-bench experiments and computational studies, and accurate non-kinase-specific prediction tools are highly desirable for whole-genome annotation in a wide variety of species. Here, we describe a phosphorylation site prediction webserver, PhosphoSVM, that employs Support Vector Machine to combine protein secondary structure information and seven other one-dimensional structural properties, including Shannon entropy, relative entropy, predicted protein disorder information, predicted solvent accessible area, amino acid overlapping properties, averaged cumulative hydrophobicity, and subsequence k-nearest neighbor profiles. This method achieved AUC values of 0.8405/0.8183/0.7383 for serine (S), threonine (T), and tyrosine (Y) phosphorylation sites, respectively, in animals with a tenfold cross-validation. The model trained by the animal phosphorylation sites was also applied to a plant phosphorylation site dataset as an independent test. The AUC values for the independent test data set were 0.7761/0.6652/0.5958 for S/T/Y phosphorylation sites, respectively. This algorithm with the optimally trained model was implemented as a webserver. The webserver, trained model, and all datasets used in the current study are available at http://sysbio.unl.edu/PhosphoSVM .

Identifying Individual-Cancer-Related Genes by Rebalancing the Training Samples.

The identification of individual-cancer-related genes typically is an imbalanced classification issue. The number of known cancer-related genes is far less than the number of all unknown genes, which makes it very hard to detect novel predictions from such imbalanced training samples. A regular machine learning method can either only detect genes related to all cancers or add clinical knowledge to circumvent this issue. In this study, we introduce a training sample rebalancing strategy to overcome this issue by using a two-step logistic regression and a random resampling method. The two-step logistic regression is to select a set of genes that related to all cancers. While the random resampling method is performed to further classify those genes associated with individual cancers. The issue of imbalanced classification is circumvented by randomly adding positive instances related to other cancers at first, and then excluding those unrelated predictions according to the overall performance at the following step. Numerical experiments show that the proposed resampling method is able to identify cancer-related genes even when the number of known genes related to it is small. The final predictions for all individual cancers achieve AUC values around 0.93 by using the leave-one-out cross validation method, which is very promising, compared with existing methods.

DephosSite: a machine learning approach for discovering phosphotase-specific dephosphorylation sites.

Protein dephosphorylation, which is an inverse process of phosphorylation, plays a crucial role in a myriad of cellular processes, including mitotic cycle, proliferation, differentiation, and cell growth. Compared with tyrosine kinase substrate and phosphorylation site prediction, there is a paucity of studies focusing on computational methods of predicting protein tyrosine phosphatase substrates and dephosphorylation sites. In this work, we developed two elegant models for predicting the substrate dephosphorylation sites of three specific phosphatases, namely, PTP1B, SHP-1, and SHP-2. The first predictor is called MGPS-DEPHOS, which is modified from the GPS (Group-based Prediction System) algorithm with an interpretable capability. The second predictor is called CKSAAP-DEPHOS, which is built through the combination of support vector machine (SVM) and the composition of k-spaced amino acid pairs (CKSAAP) encoding scheme. Benchmarking experiments using jackknife cross validation and 30 repeats of 5-fold cross validation tests show that MGPS-DEPHOS and CKSAAP-DEPHOS achieved AUC values of 0.921, 0.914 and 0.912, for predicting dephosphorylation sites of the three phosphatases PTP1B, SHP-1, and SHP-2, respectively. Both methods outperformed the previously developed kNN-DEPHOS algorithm. In addition, a web server implementing our algorithms is publicly available at http://genomics.fzu.edu.cn/dephossite/ for the research community.

Network-based inference methods for drug repositioning.

Mining potential drug-disease associations can speed up drug repositioning for pharmaceutical companies. Previous computational strategies focused on prior biological information for association inference. However, such information may not be comprehensively available and may contain errors. Different from previous research, two inference methods, ProbS and HeatS, were introduced in this paper to predict direct drug-disease associations based only on the basic network topology measure. Bipartite network topology was used to prioritize the potentially indicated diseases for a drug. Experimental results showed that both methods can receive reliable prediction performance and achieve AUC values of 0.9192 and 0.9079, respectively. Case studies on real drugs indicated that some of the strongly predicted associations were confirmed by results in the Comparative Toxicogenomics Database (CTD). Finally, a comprehensive prediction of drug-disease associations enables us to suggest many new drug indications for further studies.

Automated identification of cell-type-specific genes in the mouse brain by image computing of expression patterns.

BackgroundDifferential gene expression patterns in cells of the mammalian brain result in the morphological, connectional, and functional diversity of cells. A wide variety of studies have shown that certain genes are expressed only in specific cell-types. Analysis of cell-type-specific gene expression patterns can provide insights into the relationship between genes, connectivity, brain regions, and cell-types. However, automated methods for identifying cell-type-specific genes are lacking to date.ResultsHere, we describe a set of computational methods for identifying cell-type-specific genes in the mouse brain by automated image computing of in situ hybridization (ISH) expression patterns. We applied invariant image feature descriptors to capture local gene expression information from cellular-resolution ISH images. We then built image-level representations by applying vector quantization on the image descriptors. We employed regularized learning methods for classifying genes specifically expressed in different brain cell-types. These methods can also rank image features based on their discriminative power. We used a data set of 2,872 genes from the Allen Brain Atlas in the experiments. Results showed that our methods are predictive of cell-type-specificity of genes. Our classifiers achieved AUC values of approximately 87% when the enrichment level is set to 20. In addition, we showed that the highly-ranked image features captured the relationship between cell-types.ConclusionsOverall, our results showed that automated image computing methods could potentially be used to identify cell-type-specific genes in the mouse brain.

PhosphoSVM: prediction of phosphorylation sites by integrating various protein sequence attributes with a support vector machine

Phosphorylation is one of the most essential post-translational modifications in eukaryotes. Studies on kinases and their substrates are important for understanding cellular signaling networks. Because of the cost in time and labor associated with large-scale wet-bench experiments, computational prediction of phosphorylation sites becomes important and many computational tools have been developed in the recent decades. The prediction tools can be grouped into two categories: kinase-specific and non-kinase-specific tools. With more kinases being discovered by the new sequencing technologies, accurate non-kinase-specific prediction tools are highly desirable for whole-genome annotation in a wider variety of species. In this manuscript, a support vector machine is used to combine eight different sequence level scoring functions to predict phosphorylation sites. The attributes used by this work, including Shannon entropy, relative entropy, predicted protein secondary structure, predicted protein disorder, solvent accessible area, overlapping properties, averaged cumulative hydrophobicity, and k-nearest neighbor, were able to obtain better results than the previously used attributes by other similar methods. This method achieved AUC values of 0.8405/0.8183/0.7383 for serine (S), threonine (T), and tyrosine (Y) phosphorylation sites, respectively, in animals with a tenfold cross-validation. The model trained by the animal phosphorylation sites was also applied to a plant phosphorylation site dataset as an independent test. The AUC values for the independent test dataset were 0.7761/0.6652/0.5958 for S/T/Y phosphorylation sites, which compared favorably with those of several existing methods. A web server based on our method was constructed for public use. The server, trained model, and all datasets used in the current study are available at http://sysbio.unl.edu/PhosphoSVM .

Similarity-based methods for potential human microRNA-disease association prediction

BackgroundThe identification of microRNA-disease associations is critical for understanding the molecular mechanisms of diseases. However, experimental determination of associations between microRNAs and diseases remains challenging. Meanwhile, target diseases need to be revealed for some new microRNAs without any known target disease association information as new microRNAs are discovered each year. Therefore, computational methods for microRNA-disease association prediction have gained a lot of research interest.MethodsHerein, based on the assumption that functionally related microRNAs tend to be associated with phenotypically similar diseases, three inference methods were presented for microRNA-disease association prediction, namely MBSI (microRNA-based similarity inference), PBSI (phenotype-based similarity inference) and NetCBI (network-consistency-based inference). Global network similarity measure was used in the three methods to predict new microRNA-disease associations.ResultsWe tested the three methods on 242 known microRNA-disease associations by leave-one-out cross-validation for prediction evaluation, and achieved AUC values of 74.83%, 54.02% and 80.66%, respectively. The best-performed method NetCBI was then chosen for novel microRNA-disease association prediction. Some associations strongly predicted by NetCBI were confirmed by the publicly accessible databases, which indicated the usefulness of this method. The newly predicted associations were publicly released to facilitate future studies. Moreover, NetCBI was especially applicable to predicting target diseases for microRNAs whose target association information was not available.ConclusionsThe encouraging results suggest that our method NetCBI can not only provide help in identifying novel microRNA-disease associations but also guide biological experiments for scientific research.

Busulfan pharmacokinetics following intravenous and oral dosing regimens in children receiving high-dose myeloablative chemotherapy for high-risk neuroblastoma as part of the HR-NBL-1/SIOPEN trial

IntroductionBusulfan is widely used in a neuroblastoma setting, with several studies reporting marked inter-patient variability in busulfan pharmacokinetics and pharmacodynamics. The current study reports on the pharmacokinetics of oral versus intravenous (IV) busulfan in high-risk neuroblastoma patients treated on the European HR-NBL-1/SIOPEN study. MethodsBusulfan was administered four times daily for 4days to children aged 0.7–13.1years, either orally (1.45–1.55mg/kg) or by the IV route (0.8–1.2mg/kg according to body weight strata). Blood samples were obtained prior to administration, 2, 4, and 6h after the start of administration on dose 1. Busulfan analysis was carried out by gas chromatography–mass spectrometry and data analysed using a NONMEM population pharmacokinetic approach. ResultsBusulfan plasma concentrations obtained from 38 patients receiving IV busulfan and 25 patients receiving oral busulfan, were fitted simultaneously using a one-compartment pharmacokinetic model. Lower variability in drug exposure was observed following IV administration, with a mean busulfan area under the plasma concentration versus time curve (AUC) of 1146±187μM.min (range 838–1622), as compared to 953±290μM.min (range 434–1427) following oral busulfan. A total of 87% of children treated with IV busulfan achieved AUC values within the target of 900–1500μM.min versus 56% of patients following oral busulfan. Busulfan AUC values were significantly higher in HR-NBL-1/SIOPEN trial patients who experienced hepatic toxicity or veno-occlusive disease (VOD) (1177±189μM.min versus 913±256μM.min; p=0.0086). Further stratification based on route of administration suggested that the incidence of hepatic toxicity was related to both high busulfan AUC and oral drug administration. ConclusionThe reduced pharmacokinetic variability and improved control of busulfan AUC observed following IV administration support its utility within the ongoing HR-NBL-1/SIOPEN trial.

Computerized three-class classification of MRI-based prognostic markers for breast cancer

The purpose of this study is to investigate whether computerized analysis using three-class Bayesian artificial neural network (BANN) feature selection and classification can characterize tumor grades (grade 1, grade 2 and grade 3) of breast lesions for prognostic classification on DCE-MRI. A database of 26 IDC grade 1 lesions, 86 IDC grade 2 lesions and 58 IDC grade 3 lesions was collected. The computer automatically segmented the lesions, and kinetic and morphological lesion features were automatically extracted. The discrimination tasks—grade 1 versus grade 3, grade 2 versus grade 3, and grade 1 versus grade 2 lesions—were investigated. Step-wise feature selection was conducted by three-class BANNs. Classification was performed with three-class BANNs using leave-one-lesion-out cross-validation to yield computer-estimated probabilities of being grade 3 lesion, grade 2 lesion and grade 1 lesion. Two-class ROC analysis was used to evaluate the performances. We achieved AUC values of 0.80 ± 0.05, 0.78 ± 0.05 and 0.62 ± 0.05 for grade 1 versus grade 3, grade 1 versus grade 2, and grade 2 versus grade 3, respectively. This study shows the potential for (1) applying three-class BANN feature selection and classification to CADx and (2) expanding the role of DCE-MRI CADx from diagnostic to prognostic classification in distinguishing tumor grades.