The development of a custom RNA-sequencing panel for the identification of predictive and diagnostic biomarkers in glioma.

PURPOSE: Various molecular profiles are needed to classify malignant brain tumors, including gliomas, based on the latest classification criteria of the World Health Organization, and their poor prognosis necessitates new therapeutic targets. The Todai OncoPanel 2 RNA Panel (TOP2-RNA) is a custom-target RNA-sequencing (RNA-seq) using the junction capture method to maximize the sensitivity of detecting 455 fusion gene transcripts and analyze the expression profiles of 1,390 genes. This study aimed to classify gliomas and identify their molecular targets using TOP2-RNA. METHODS: A total of 124 frozen samples of malignant gliomas were subjected to TOP2-RNA for classification based on their molecular profiles and the identification of molecular targets. RESULTS: Among 55 glioblastoma cases, gene fusions were detected in 11 cases (20%), including novel MET fusions. Seven tyrosine kinase genes were found to be overexpressed in 15 cases (27.3%). In contrast to isocitrate dehydrogenase (IDH) wild-type glioblastoma, IDH-mutant tumors, including astrocytomas and oligodendrogliomas, barely harbor fusion genes or gene overexpression. Of the 34 overexpressed tyrosine kinase genes, MDM2 and CDK4 in glioblastoma, 22 copy number amplifications (64.7%) were observed. When comparing astrocytomas and oligodendrogliomas in gene set enrichment analysis, the gene sets related to 1p36 and 19q were highly enriched in astrocytomas, suggesting that regional genomic DNA copy number alterations can be evaluated by gene expression analysis. CONCLUSIONS: TOP2-RNA is a highly sensitive assay for detecting fusion genes, exon skipping, and aberrant gene expression. Alterations in targetable driver genes were identified in more than 50% of glioblastoma. Molecular profiling by TOP2-RNA provides ample predictive, prognostic, and diagnostic biomarkers that may not be identified by conventional assays and, therefore, is expected to increase treatment options for individual patients with glioma.

Multiomic molecular characterization of the response to combination immunotherapy in MSS/pMMR metastatic colorectal cancer.

BACKGROUND: Immune checkpoint inhibitor (ICI) combinations represent an emerging treatment strategies in cancer. However, their efficacy in microsatellite stable (MSS) or mismatch repair-proficient (pMMR) colorectal cancer (CRC) is variable. Here, a multiomic characterization was performed to identify predictive biomarkers associated with patient response to ICI combinations in MSS/pMMR CRC for the further development of ICI combinations. METHODS: Whole-exome sequencing, RNA sequencing, and multiplex fluorescence immunohistochemistry of tumors from patients with MSS/pMMR CRC, who received regorafenib plus nivolumab (REGONIVO) or TAS-116 plus nivolumab (TASNIVO) in clinical trials were conducted. Twenty-two and 23 patients without prior ICI from the REGONIVO and TASNIVO trials were included in this study. A biomarker analysis was performed using samples from each of these studies. RESULTS: The epithelial-mesenchymal transition pathway and genes related to cancer-associated fibroblasts were upregulated in the REGONIVO responder group, and the G2M checkpoint pathway was upregulated in the TASNIVO responder group. The MYC pathway was upregulated in the REGONIVO non-responder group. Consensus molecular subtype 4 was significantly associated with response (p=0.035) and longer progression-free survival (p=0.006) in the REGONIVO trial. CD8+ T cells, regulatory T cells, and M2 macrophages density was significantly higher in the REGONIVO trial responders than in non-responders. Mutations in the POLE gene and patient response were significantly associated in the TASNIVO trial; however, the frequencies of other mutations or tumor mutational burden were not significantly different between responders and non-responders in either trial. CONCLUSIONS: We identified molecular features associated with the response to the REGONIVO and TASNIVO, particularly those related to tumor microenvironmental factors. These findings are likely to contribute to the development of biomarkers to predict treatment efficacy for MSS/pMMR CRC and future immunotherapy combinations for treatment.

Evaluation of deep learning reconstruction on diffusion-weighted imaging quality and apparent diffusion coefficient using an ice-water phantom.

This study assessed the influence of deep learning reconstruction (DLR) on the quality of diffusion-weighted images (DWI) and apparent diffusion coefficient (ADC) using an ice-water phantom. An ice-water phantom with known diffusion properties (true ADC = 1.1 × 10-3 mm2/s at 0 °C) was imaged at various b-values (0, 1000, 2000, and 4000 s/mm2) using a 3 T magnetic resonance imaging scanner with slice thicknesses of 1.5 and 3.0 mm. All DWIs were reconstructed with or without DLR. ADC maps were generated using combinations of b-values 0 and 1000, 0 and 2000, and 0 and 4000 s/mm2. Based on the quantitative imaging biomarker alliance profile, the signal-to-noise ratio (SNRs) in DWIs was calculated, and the accuracy, precision, and within-subject parameter variance (wCV) of the ADCs were evaluated. DLR improved the SNR in DWIs with b-values ranging from 0 to 2000s/mm2; however, its effectiveness was diminished at 4000 s/mm2. There was no noticeable difference in the ADCs of images generated with or without implementing DLR. For a slice thickness of 1.5 mm and combined b-values of 0 and 4000 s/mm2, the ADC values were 0.97 × 10-3and 0.98 × 10-3mm2/s with and without DLR, respectively, both being lower than the true ADC value. Furthermore, DLR enhanced the precision and wCV of the ADC measurements. DLR can enhance the SNR, repeatability, and precision of ADC measurements; however, it does not improve their accuracies.

Promoter swapping of truncated PDGFRB drives Ph-like acute lymphoblastic leukemia.

Philadelphia chromosome (Ph)-like acute lymphoblastic leukemia (ALL) is a subset of ALL that demonstrated a high treatment failure rate. One of the hallmarks of Ph-like ALL is PDGFRB gene fusion, with fusion partner proteins often harboring dimerization domains and enhancing the kinase activity of PDGFRB. We determined a novel oncogenic PDGFRB fusion gene, NRIP1::PDGFRB, from a pediatric patient with ALL, encoding a protein with the carboxy-terminal kinase domain of PDGFRB, without the partner peptide. We confirmed the oncogenic potential of NRIP1::PDGFRB in vitro and the efficacy of all ABL1-specific inhibitor generations, including imatinib, dasatinib, nilotinib, and ponatinib, in suppressing this potential. PDGFRB activation mechanism may include juxtamembrane domain truncation in the predicted peptide. In conclusion, we determined a novel fusion gene pattern in Ph-like ALL.

Extensive analysis of 59 sarcoma-related fusion genes identified pazopanib as a potential inhibitor to COL1A1-PDGFB fusion gene.

Sarcomas are malignant mesenchymal tumors that are extremely rare and divergent. Fusion genes are involved in approximately 30% of sarcomas as driver oncogenes; however, their detailed functions are not fully understood. In this study, we determined the functional significance of 59 sarcoma-related fusion genes. The transforming potential and drug sensitivities of these fusion genes were evaluated using a focus formation assay (FFA) and the mixed-all-nominated-in-one (MANO) method, respectively. The transcriptome was also examined using RNA sequencing of 3T3 cells transduced with each fusion gene. Approximately half (28/59, 47%) of the fusion genes exhibited transformation in the FFA assay, which was classified into five types based on the resulting phenotype. The sensitivity to 12 drugs including multityrosine kinase inhibitors was assessed using the MANO method and pazopanib was found to be more effective against cells expressing the COL1A1-PDGFB fusion gene compared with the others. The downstream MAPK/AKT pathway was suppressed at the protein level following pazopanib treatment. The fusion genes were classified into four subgroups by cluster analysis of the gene expression data and gene set enrichment analysis. In summary, the oncogenicity and drug sensitivity of 59 fusion genes were simultaneously evaluated using a high-throughput strategy. Pazopanib was selected as a candidate drug for sarcomas harboring the COL1A1-PDGFB fusion gene. This assessment could be useful as a screening platform and provides a database to evaluate customized therapy for fusion gene-associated sarcomas.

Impact of arm position on vertebral bone marrow proton density fat fraction in chemical-shift-encoded magnetic resonance imaging: a preliminary study.

Arm positions employed during magnetic resonance imaging (MRI) can affect magnetic field distribution, which may result in variability in proton density fat fraction (PDFF) measurements. This study evaluated the effect of arm position on lumbar PDFF measured using chemical-shift-encoded MRI (CSE-MRI). Fifteen healthy volunteers from a single-center underwent lumbar CSE-MRI at two different arm positions (side and elevated) using a single 3T scanner. Scans were performed twice in each position. PDFFs of the L1-L5 vertebrae were independently measured by two readers, and reader measurements were compared by calculating intraclass correlation coefficients (ICC). We compared PDFF measurements from two arm positions and from two consecutive scans using the Wilcoxon test and Bland-Altman analysis. Measurements from the two readers were in high agreement [ICC =0.999; 95% confidence interval (CI), 0.998-0.999]. No significant difference was observed between PDFFs from the first and second scans of all vertebrae for each reader (all P>0.05); however, PDFF for the elevated arm position was significantly higher than that for the side arm position (37.9-44.8% vs. 37.0-43.8%; all P<0.05), except at the L2 level by reader 2. The mean differences in PDFF measurements from the first and second scans [0.1%; 95% limits of agreement (LoA), -1.8% to 1.9%] and from the side arm and elevated arm positions (0.8%; 95% LoA, -1.6% to 3.2%) were small. In conclusion, these preliminary data suggest that different arm positions during CSE-MRI can slightly affect lumbar PDFF; however, the mean absolute differences were very small.

Identification of novel prognostic and predictive biomarkers in salivary duct carcinoma via comprehensive molecular profiling.

Molecular targets and predictive biomarkers for prognosis in salivary duct carcinoma (SDC) have not been fully identified. We conducted comprehensive molecular profiling to discover novel biomarkers for SDC. A total of 67 SDC samples were examined with DNA sequencing of 464 genes and transcriptome analysis in combination with the clinicopathological characteristics of the individuals. Prognostic biomarkers associated with response to combined androgen blockade (CAB) treatment were explored using mRNA expression data from 27 cases. Oncogenic mutations in receptor tyrosine kinase (RTK) genes or genes in the MAPK pathway were identified in 55 cases (82.1%). Alterations in the phosphatidylinositol 3-kinase (PI3K)/AKT signaling pathway were identified in 38 cases (56.7%). Interestingly, patient prognosis could be predicted using mRNA expression profiles, but not genetic mutation profiles. The risk score generated from the expression data of a four-gene set that includes the ADAMTS1, DSC1, RNF39, and IGLL5 genes was a significant prognostic marker for overall survival in the cohort (HR = 5.99, 95% confidence interval (CI) = 2.73-13.1, p = 7.8 × 10-6). Another risk score constructed from the expression of CD3E and LDB3 was a strong prognostic marker for progression-free survival for CAB treatment (p = 0.03). Mutations in RTK genes, MAPK pathway genes, and PI3K/AKT pathway genes likely represent key mutations in SDC tumorigenesis. The gene expression profiles identified in this study may be useful for stratifying patients who are good candidates for CAB treatment and may benefit from additional systemic therapies.

Variability in contrast and apparent diffusion coefficient of kiwifruit used as prostate MRI phantom: 1-week validation.

Kiwifruit has been proposed as a phantom for prostate magnetic resonance imaging (MRI) to adjust multiparametric MRI factors. This study evaluated the variability in contrasts and apparent diffusion coefficients (ADCs) via repeated scans of kiwifruits for 1 week. All scans were performed using a 3T MRI system. Six kiwifruits were prepared as phantoms. Each kiwifruit was scanned consecutively for 1 week, and T2-weighted and diffusion-weighted images were acquired. Contrasts between the peripheral placenta (PP) and central placenta (CP), and between the outer pericarp (OP) and CP were calculated. The ADCs of the PP, OP, and CP were also determined. The Friedman test with Scheffé's post hoc comparison was used to assess contrast and ADC variability over 1 week; no significant differences between the contrasts or ADCs were found. Thus, each kiwifruit phantom exhibited low variability when used as a prostate MRI phantom.

Estimation of patient's angle from skull radiographs using deep learning.

BACKGROUND: Skull radiography, an assessment method for initial diagnosis and post-operative follow-up, requires substantial retaking of various types of radiographs. During retaking, a radiologic technologist estimates a patient's rotation angle from the radiograph by comprehending the relationship between the radiograph and the patient's angle for adequate assessment, which requires extensive experience. OBJECTIVE: To develop and test a new deep learning model or method to automatically estimate patient's angle from radiographs. METHODS: The patient's position is assessed using deep learning to estimate their angle from skull radiographs. Skull radiographs are simulated using two-dimensional projections from head computed tomography images and used as input data to estimate the patient's angle, using deep learning under supervised training. A residual neural network model is used where the rectified linear unit is changed to a parametric rectified linear unit, and dropout is added. The patient's angle is estimated in the lateral and superior-inferior directions. RESULTS: Applying this new deep learning model, the estimation errors are 0.56±0.36° and 0.72±0.52° in the lateral and superior-inferior angles, respectively. CONCLUSIONS: These findings suggest that a patient's angle can be accurately estimated from a radiograph using a deep learning model leading to reduce retaking time, and then used to facilitate skull radiography.

Adapting a low-count acquisition of the bone scintigraphy using deep denoising super-resolution convolutional neural network.

PURPOSE: Deep-layer learning processing may improve contrast imaging with greater precision in low-count acquisition. However, no data on noise reduction using super-resolution processing for deep-layer learning have been reported in nuclear medicine imaging. OBJECTIVES: This study was designed to evaluate the adaptability of deep denoising super-resolution convolutional neural networks (DDSRCNN) in nuclear medicine by comparing them with denoising convolutional natural networks (DnCNN), Gaussian processing, and nonlinear diffusion (NLD) processing. METHODS: In this study, 156 patients were included. Data were collected using a matrix size of 256 × 256 with a pixel size of 2.46 mm at 0.898 folds, 15% energy window at the center of the photopeak energy (140 keV), and total count of 1000 kilocounts (kct). Following the training and validation of two learning models, we created 100 images for each 20-test datum. The peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) between each image and the reference image were calculated. RESULTS: DDSRCNN showed the highest PSNR values for all total counts. Regarding SSIM, DDSRCNN had significantly higher values than the original and Gaussian. In DnCNN, false accumulation was observed as the total counts increased. Regarding PSNR and SSIM transition, the model using 100-500-kct training data was significantly higher than that using 100-kct training data. CONCLUSIONS: Edge-preserving noise reduction processing was possible, and adaptability to low-count acquisition was demonstrated using DDSRCNN. Using training data with different noise levels, DDSRCNN could learn the noise components with high accuracy and contrast improvement.

Exploration of predictive biomarkers for postoperative recurrence of stage II/III colorectal cancer using genomic sequencing.

Postoperative recurrence of colorectal cancer (CRC) eventually leads to therapeutic failure; therefore, treatment strategies based on accurate prediction of recurrence are urgently required. To identify biomarkers that can predict treatment outcomes, we compared the mutational profiles of surgically resected specimens from patients with recurrent cancer with those from patients with non-recurrent cancer. Target sequencing, whole-exome sequencing (WES), or whole-genome sequencing (WGS) was performed on 89 and 58 tumors from recurrent and non-recurrent cases, respectively. WGS revealed the driver mutations that were not detected with target sequencing or WES, including the structural variations affecting ZFP36L2. Loss of function of ZFP36L2 was frequently observed in primary tumors from recurrent cases. Furthermore, the recurrence-free survival of patients with loss of function of ZFP36L2 was significantly shorter relative to patients with no loss of ZFP36L2 function. In summary, the study demonstrated that detailed genomic analysis could help improve precision medicine for CRC.

HLA Class I Analysis Provides Insight Into the Genetic and Epigenetic Background of Immune Evasion in Colorectal Cancer With High Microsatellite Instability.

BACKGROUND & AIMS: A detailed understanding of antitumor immunity is essential for optimal cancer immune therapy. Although defective mutations in the B2M and HLA-ABC genes, which encode molecules essential for antigen presentation, have been reported in several studies, the effects of these defects on tumor immunity have not been quantitatively evaluated. METHODS: Mutations in HLA-ABC genes were analyzed in 114 microsatellite instability-high colorectal cancers using a long-read sequencer. The data were further analyzed in combination with whole-exome sequencing, transcriptome sequencing, DNA methylation array, and immunohistochemistry data. RESULTS: We detected 101 truncating mutations in 57 tumors (50%) and loss of 61 alleles in 21 tumors (18%). Based on the integrated analysis that enabled the immunologic subclassification of microsatellite instability-high colorectal cancers, we identified a subtype of tumors in which lymphocyte infiltration was reduced, partly due to reduced expression of HLA-ABC genes in the absence of apparent genetic alterations. Survival time of patients with such tumors was shorter than in patients with other tumor types. Paradoxically, tumor mutation burden was highest in the subtype, suggesting that the immunogenic effect of accumulating mutations was counterbalanced by mutations that weakened immunoreactivity. Various genetic and epigenetic alterations, including frameshift mutations in RFX5 and promoter methylation of PSMB8 and HLA-A, converged on reduced expression of HLA-ABC genes. CONCLUSIONS: Our detailed immunogenomic analysis provides information that will facilitate the improvement and development of cancer immunotherapy.

Two novel high-risk adult B-cell acute lymphoblastic leukemia subtypes with high expression of CDX2 and IDH1/2 mutations.

The genetic basis of leukemogenesis in adults with B-cell acute lymphoblastic leukemia (B-ALL) is largely unclear, and its clinical outcome remains unsatisfactory. This study aimed to advance the understanding of biological characteristics, improve disease stratification, and identify molecular targets of adult B-ALL. Adolescents and young adults (AYA) (15 to 39 years old, n = 193) and adults (40 to 64 years old, n = 161) with Philadelphia chromosome-negative (Ph-) B-ALL were included in this study. Integrated transcriptomic and genetic analyses were used to classify the cohort into defined subtypes. Of the 323 cases included in the RNA sequencing analysis, 278 (86.1%) were classified into 18 subtypes. The ZNF384 subtype (22.6%) was the most prevalent, with 2 novel subtypes (CDX2-high and IDH1/2-mut) identified among cases not assigned to the established subtypes. The CDX2-high subtype (3.4%) was characterized by high expression of CDX2 and recurrent gain of chromosome 1q. The IDH1/2-mut subtype (1.9%) was defined by IDH1 R132C or IDH2 R140Q mutations with specific transcriptional and high-methylation profiles. Both subtypes showed poor prognosis and were considered inferior prognostic factors independent of clinical parameters. Comparison with a previously reported pediatric B-ALL cohort (n = 1003) showed that the frequencies of these subtypes were significantly higher in AYA/adults than in children. We delineated the genetic and transcriptomic landscape of adult B-ALL and identified 2 novel subtypes that predict poor disease outcomes. Our findings highlight the age-dependent distribution of subtypes, which partially accounts for the prognostic differences between adult and pediatric B-ALL.

Denoising Using Noise2Void for Low-Field Magnetic Resonance Imaging: A Phantom Study.

To reduce noise for low-field magnetic resonance imaging (MRI) using Noise2Void (N2V) and to demonstrate the N2V validity. N2V is one of the denoising convolutional neural network methods that allows the training of a model without a noiseless clean image. In this study, a kiwi fruit was scanned using a 0.35 Tesla MRI system, and the image qualities at pre- and postdenoising were evaluated. Structural similarity (SSIM), signal-to-noise ratio (SNR), and contrast ratio (CR) were measured, and visual assessment of noise and sharpness was observed. Both SSIM and SNR were significantly improved using N2V (P < 0.05). CR was unchanged between pre- and postdenoising images. The results of visual assessment for noise revealed higher scores in postdenoising images than that in predenoising images. The sharpness scores of postdenoising images were high when SNR was low. N2V provides effective noise reduction and is a useful denoising technique in low-field MRI.

Transcript-targeted analysis reveals isoform alterations and double-hop fusions in breast cancer.

Although transcriptome alteration is an essential driver of carcinogenesis, the effects of chromosomal structural alterations on the cancer transcriptome are not yet fully understood. Short-read transcript sequencing has prevented researchers from directly exploring full-length transcripts, forcing them to focus on individual splice sites. Here, we develop a pipeline for Multi-Sample long-read Transcriptome Assembly (MuSTA), which enables construction of a transcriptome from long-read sequence data. Using the constructed transcriptome as a reference, we analyze RNA extracted from 22 clinical breast cancer specimens. We identify a comprehensive set of subtype-specific and differentially used isoforms, which extended our knowledge of isoform regulation to unannotated isoforms including a short form TNS3. We also find that the exon-intron structure of fusion transcripts depends on their genomic context, and we identify double-hop fusion transcripts that are transcribed from complex structural rearrangements. For example, a double-hop fusion results in aberrant expression of an endogenous retroviral gene, ERVFRD-1, which is normally expressed exclusively in placenta and is thought to protect fetus from maternal rejection; expression is elevated in several TCGA samples with ERVFRD-1 fusions. Our analyses provide direct evidence that full-length transcript sequencing of clinical samples can add to our understanding of cancer biology and genomics in general.

Highly sensitive fusion detection using plasma cell-free RNA in non-small-cell lung cancers.

ALK, ROS1, and RET kinase fusions are important predictive biomarkers of tyrosine kinase inhibitors (TKIs) in non-small-cell lung cancer (NSCLC). Analysis of cell-free DNA (cfDNA) provides a noninvasive method to identify gene changes in tumor cells. The present study sought to use cfRNA and cfDNA for identifying fusion genes. A reliable protocol was established to detect fusion genes using cfRNA and assessed the analytical validity and clinical usefulness in 30 samples from 20 cases of fusion-positive NSCLC. The results of cfRNA-based assays were compared with tissue biopsy and cfDNA-based liquid biopsy (Guardant360 plasma next-generation sequencing [NGS] assay). The overall sensitivity of the cfRNA-based assay was 26.7% (8/30) and that of cfDNA-based assay was 16.7% (3/18). When analysis was limited to the samples collected at chemo-naïve or progressive disease status and available for both assays, the sensitivity of the cfRNA-based assay was 77.8% (7/9) and that of cfDNA-based assay was 33.3% (3/9). Fusion gene identification in cfRNA was correlated with treatment response. These results suggest that the proposed cfRNA assay is a useful diagnostic test for patients with insufficient tissues to facilitate effective administration of first-line treatment and is a useful tool to monitor the progression of NSCLC for consideration of second-line treatments.

Comprehensive molecular profiling of pulmonary pleomorphic carcinoma.

Information regarding the molecular features of pulmonary pleomorphic carcinoma (PPC) is insufficient. Here, we performed next-generation sequencing to determine the genomic and transcriptomic profiles of PPC. We sequenced the DNAs and RNAs of 78 specimens from 52 patients with PPC. We analyzed 15 PPC cases to identify intratumoral differences in gene alterations, tumor mutation burden (TMB), RNA expression, and PD-L1 expression between epithelial and sarcomatoid components. The genomic alterations of six cases of primary tumors and corresponding metastatic tumors were analyzed. KRAS mutations (27%) were the most common driver mutations, followed by EGFR (8%), and MET (8%) mutations. Epithelial and sarcomatoid components shared activating driver mutations, and there were no significant differences in CD274 expression or TMB between the two components. However, PD-L1 was highly expressed in the sarcomatoid component of several cases compared with the epithelial component. Primary and metastatic tumors shared oncogenic mutations among genes such as KRAS and TP53, and additional alterations including NOTCH4 mutations were specifically identified in the metastatic regions. Our data suggest that therapies targeting activating driver mutations may be effective for patients with PPC and that immune checkpoint inhibitors of PPC may be recommended after careful assessment of PD-L1 expression in each epithelial and sarcomatoid component.

Rapid reprogramming of tumour cells into cancer stem cells on double-network hydrogels.

Cancer recurrence can arise owing to rare circulating cancer stem cells (CSCs) that are resistant to chemotherapies and radiotherapies. Here, we show that a double-network hydrogel can rapidly reprogramme differentiated cancer cells into CSCs. Spheroids expressing elevated levels of the stemness genes Sox2, Oct3/4 and Nanog formed within 24 h of seeding the gel with cells from any of six human cancer cell lines or with brain cancer cells resected from patients with glioblastoma. Human brain cancer cells cultured on the double-network hydrogel and intracranially injected in immunodeficient mice led to higher tumorigenicity than brain cancer cells cultured on single-network gels. We also show that the double-network gel induced the phosphorylation of tyrosine kinases, that gel-induced CSCs from primary brain cancer cells were eradicated by an inhibitor of the platelet-derived growth factor receptor, and that calcium channel receptors and the protein osteopontin were essential for the regulation of gel-mediated induction of stemness in brain cancer cells.

Comprehensive molecular and clinicopathological profiling of desmoid tumours.

Previous studies have not clearly identified a prognostic factor for desmoid tumours (DT). Whole-exome sequencing (WES) and/or RNA sequencing (RNA-seq) were performed in 64 cases of DT to investigate the molecular profiles in combination with the clinicopathological characteristics. CTNNB1 mutations with specific hotspots were identified in 56 cases (87.5%). A copy number loss in chromosome 6 (chr6) was identified in 14 cases (21.9%). Clustering based on the mRNA expression profiles was predictive of the patients' prognoses. The risk score generated by the expression of a three-gene set (IFI6, LGMN, and CKLF) was a strong prognostic marker for recurrence-free survival (RFS) in our cohort. In risk groups stratified by the expression of IFI6, the hazard ratio for recurrence-free survival in the high-risk group relative to the low-risk group was 12.12 (95% confidence interval: 1.56-94.2; p = 8.0 × 106). In conclusion, CTNNB1 mutations and a chr6 copy number loss are likely the causative mutations underlying the tumorigenesis of DT while the gene expression profiles may help to differentiate patients who would be good candidates for wait-and-see management and those who might benefit from additional systemic or radiation therapies.