Detection of heteroresistance of remains challenging using current genotypic drug susceptibility testing methods. Here, we described a melting curve analysis-based approach, termed DeepMelt, that can detect less-abundant mutants through selective clamping of the wild type in mixed populations. The singleplex DeepMelt assay detected 0.01% S315T in 10 genomes/μl. The multiplex DeepMelt TB/INH detected 1% of mutant species in the four loci associated with isoniazid resistance in 10 genomes/μl. The DeepMelt TB/INH assay was tested on a panel of DNA extracted from 602 precharacterized clinical isolates. Using the 1% proportion method as the gold standard, the sensitivity was found to be increased from 93.6% (176/188, 95% confidence interval [CI] = 89.2 to 96.3%) to 95.7% (180/188, 95% CI = 91.8 to 97.8%) compared to the MeltPro TB/INH assay. Further evaluation of 109 smear-positive sputum specimens increased the sensitivity from 83.3% (20/24, 95% CI = 64.2 to 93.3%) to 91.7% (22/24, 95% CI = 74.2 to 97.7%). In both cases, the specificity remained nearly unchanged. All heteroresistant samples newly identified by the DeepMelt TB/INH assay were confirmed by DNA sequencing and even partially by digital PCR. The DeepMelt assay may fill the gap between current genotypic and phenotypic drug susceptibility testing for detecting drug-resistant tuberculosis patients.

Recent studies in biometric-based identification tasks have shown that deep learning methods can achieve better performance. These methods generally extract the global features as descriptor to represent the original image. Nonetheless, it does not perform well for biometric identification under fine-grained tasks. The main reason is that the single image descriptor contains insufficient information to represent image. In this paper, we present a dual global descriptor model, which combines multiple global descriptors to exploit multi level image features. Moreover, we utilize a contrastive loss to enlarge the distance between image representations of confusing classes. The proposed framework achieves the top2 on the

Background. Non-small-cell lung cancer (NSCLC) is the most common malignant tumor among males and females worldwide. Hypoxia is a typical feature of the tumor microenvironment, and it affects cancer development. Circular RNAs (circRNAs) have been reported to sponge miRNAs to regulate target gene expression and play an essential role in tumorigenesis and progression. This study is aimed at identifying whether circRNAs could be used as the diagnostic biomarkers for NSCLC. Methods. The heterogeneity of samples in this study was assessed by principal component analysis (PCA). Furthermore, the Gene Expression Omnibus (GEO) database was normalized by the affy R package. We further screened the differentially expressed genes (DEGs) and differentially expressed circular RNAs (DEcircRNAs) using the DEseq2 R package. Moreover, we analyzed the Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment of DEGs using the cluster profile R package. Besides, the Gene Set Enrichment Analysis (GSEA) was used to identify the biological function of DEGs. The interaction between DEGs and the competing endogenous RNAs (ceRNA) network was detected using STRING and visualized using Cytoscape. Starbase predicted the miRNAs of target hub genes, and miRanda predicted the target miRNAs of circRNAs. The RNA-seq profiler and clinical information were downloaded from The Cancer Genome Atlas (TCGA) database. Then, the variables were assessed by the univariate and multivariate Cox proportional hazard regression models. Significant variables in the univariate Cox proportional hazard regression model were included in the multivariate Cox proportional hazard regression model to analyze the association between the variables of clinical features. Furthermore, the overall survival of variables was determined by the Kaplan-Meier survival curve, and the time-dependent receiver operating characteristic (ROC) curve analysis was used to calculate and validate the risk score in NSCLC patients. Moreover, predictive nomograms were constructed and used to predict the prognostic features between the high-risk and low-risk score groups. Results. We screened a total of 2039 DEGs, including 1293 upregulated DEGs and 746 downregulated DEGs in hypoxia-treated A549 cells. A549 cells treated with hypoxia had a total of 70 DEcircRNAs, including 21 upregulated and 49 downregulated DEcircRNAs, compared to A549 cells treated with normoxia. The upregulated genes were significantly enriched in 284 GO terms and 42 KEGG pathways, while the downregulated genes were significantly enriched in 184 GO terms and 25 KEGG pathways. Moreover, the function analysis by GSEA showed enrichment in the enzyme-linked receptor protein signaling pathway, hypoxia-inducible factor- (HIF-) 1 signaling pathway, and G protein-coupled receptor (GPCR) downstream signaling. Furthermore, six hub modules and 10 hub genes, CDC45, EXO1, PLK1, RFC4, CCNB1, CDC6, MCM10, DLGAP5, AURKA, and POLE2, were identified. The ceRNA network was constructed, and it consisted of 4 circRNAs, 14 miRNAs, and 38 mRNAs. The ROC curve was constructed and calculated. The area under the curve (AUC) value was 0.62, and the optimal threshold was 0.28. Based on the optimal threshold, the patients were divided into the high-risk score and low-risk score groups. The survival rate in the high-risk score group was lower than that in the low-risk score group. The expression of SERPINE1, STC2, and LPCAT1; clinical stage; and age of the patient were significantly correlated with the high-risk score. Moreover, nomograms were established based on the risk factors in multivariate analysis, and the median survival time, 3-year survival probability, and 5-year survival were possibly predicted according to nomograms. Conclusion. The ceRNA network associated with NSCLC was identified, and the hub genes, circRNAs, might act as the potential biomarkers for NSCLC.

No abstract