A three-gene panel that distinguishes benign from malignant thyroid nodules

A multi-task model for reliable classification of thyroid nodules in ultrasound images

Thyroid nodules are common, and patients with potential malignant lesions are usually diagnosed using ultrasound imaging to determine further treatment options. This study aims to propose a computer-aided diagnosis method for benign and malignant classification of thyroid nodules in ultrasound images. We propose a novel multi-task framework that combines the advantages of dense connectivity, Squeeze-and-Excitation (SE) connectivity, and Atrous Spatial Pyramid Pooling (ASPP) layer to enhance feature extraction. The Dense connectivity is used to optimize feature reuse, the SE connectivity to optimize feature weights, the ASPP layer to fuse feature information, and a multi-task learning framework to adjust the attention of the network. We evaluate our model using a 10-fold cross-validation approach based on our established Thyroid dataset. We assess the performance of our method using six average metrics: accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and AUC, which are 93.49, 95.54, 91.52, 91.63, 95.47, and 96.84%, respectively. Our proposed method outperforms other classification networks in all metrics, achieving optimal performance. We propose a multi-task model, DSMA-Net, for distinguishing thyroid nodules in ultrasound images. This method can further enhance the diagnostic ability of doctors for suspected cancer patients and holds promise for clinical applications.

Human understandable thyroid ultrasound imaging AI report system - A bridge between AI and clinicians

Artificial intelligence (AI) enables accurate diagnosis of thyroid cancer; however, the lack of explanation limits its application. In this study, we collected 10,021 ultrasound images from 8,079 patients across four independent institutions to develop and validate a human understandable AI report system named TiNet for thyroid cancer prediction. TiNet can extract thyroid nodule features such as texture, margin, echogenicity, shape, and location using a deep learning method conforming to the clinical diagnosis standard. Moreover, it offers excellent prediction performance (AUC 0.88) and provides quantitative explanations for the predictions. We conducted a reverse cognitive test in which clinicians matched the correct ultrasound images according to TiNet and clinical reports. The results indicated that TiNet reports (87.1% accuracy) were significantly easier to understand than clinical reports (81.6% accuracy; p < 0.001). TiNet can serve as a bridge between AI-based diagnosis and clinicians, enhancing human-AI cooperative medical decision-making.

A meta-learning approach to improving radiation response prediction in cancers

PURPOSE

Predicting the efficacy of radiotherapy in individual patients has drawn widespread attention, but the limited sample size remains a bottleneck for utilizing high-dimensional multi-omics data to guide personalized radiotherapy. We hypothesize the recently developed meta-learning framework could address this limitation.

METHODS AND MATERIALS

By combining gene expression, DNA methylation, and clinical data of 806 patients who had received radiotherapy from The Cancer Genome Atlas (TCGA), we applied the Model-Agnostic Meta-Learning (MAML) framework to tasks consisting of pan-cancer data, to obtain the best initial parameters of a neural network for a specific cancer with smaller number of samples. The performance of meta-learning framework was compared with four traditional machine learning methods based on two training schemes, and tested on Cancer Cell Line Encyclopedia (CCLE) and Chinese Glioma Genome Atlas (CGGA) datasets. Moreover, biological significance of the models was investigated by survival analysis and feature interpretation.

RESULTS

The mean AUC (Area under the ROC Curve) [95% confidence interval] of our models across nine cancer types was 0.702 [0.691-0.713], which improved by 0.166 on average over other the four machine learning methods on two training schemes. Our models performed significantly better (p < 0.05) in seven cancer types and performed comparable to the other predictors in the rest of two cancer types. The more pan-cancer samples were used to transfer meta-knowledge, the greater the performance improved (p < 0.05). The predicted response scores that our models generated were negatively correlated with cell radiosensitivity index in four cancer types (p < 0.05), while not statistically significant in the other three cancer types. Moreover, the predicted response scores were shown to be prognostic factors in seven cancer types and eight potential radiosensitivity-related genes were identified.

CONCLUSIONS

For the first time, we established the meta-learning approach to improving individual radiation response prediction by transferring common knowledge from pan-cancer data with MAML framework. The results demonstrated the superiority, generalizability, and biological significance of our approach.

Inferring Time-Lagged Causality Using the Derivative of Single-Cell Expression

Many computational methods have been developed to infer causality among genes using cross-sectional gene expression data, such as single-cell RNA sequencing (scRNA-seq) data. However, due to the limitations of scRNA-seq technologies, time-lagged causal relationships may be missed by existing methods. In this work, we propose a method, called causal inference with time-lagged information (CITL), to infer time-lagged causal relationships from scRNA-seq data by assessing the conditional independence between the changing and current expression levels of genes. CITL estimates the changing expression levels of genes by “RNA velocity”. We demonstrate the accuracy and stability of CITL for inferring time-lagged causality on simulation data against other leading approaches. We have applied CITL to real scRNA data and inferred 878 pairs of time-lagged causal relationships. Furthermore, we showed that the number of regulatory relationships identified by CITL was significantly more than that expected by chance. We provide an R package and a command-line tool of CITL for different usage scenarios.

Identification of a 3-Gene Prognostic Index for Papillary Thyroid Carcinoma

The accurate determination of the risk of cancer recurrence is a critical unmet need in managing thyroid cancer (TC). Although numerous studies have successfully demonstrated the use of high throughput molecular diagnostics in TC prediction, it has not been successfully applied in routine clinical use, particularly in Chinese patients. In our study, we objective to screen for characteristic genes specific to PTC and establish an accurate model for diagnosis and prognostic evaluation of PTC. We screen the differentially expressed genes by Python 3.6 in The Cancer Genome Atlas (TCGA) database. We discovered a three-gene signature Gap junction protein beta 4 (GJB4), Ripply transcriptional repressor 3 (RIPPLY3), and Adrenoceptor alpha 1B (ADRA1B) that had a statistically significant difference. Then we used Gene Expression Omnibus (GEO) database to establish a diagnostic and prognostic model to verify the three-gene signature. For experimental validation, immunohistochemistry in tissue microarrays showed that thyroid samples’ proteins expressed by this three-gene are differentially expressed. Our protocol discovered a robust three-gene signature that can distinguish prognosis, which will have daily clinical application.

The Role of Post-Translational Modifications of Chemokines by CD26 in Cancer

Chemokines are a large family of small chemotactic cytokines that fulfill a central function in cancer. Both tumor-promoting and -impeding roles have been ascribed to chemokines, which they exert in a direct or indirect manner. An important post-translational modification that regulates chemokine activity is the NH₂-terminal truncation by peptidases. CD26 is a dipeptidyl peptidase (DPPIV), which typically clips a NH₂-terminal dipeptide from the chemokine. With a certain degree of selectivity in terms of chemokine substrate, CD26 only recognizes chemokines with a penultimate proline or alanine. Chemokines can be protected against CD26 recognition by specific amino acid residues within the chemokine structure, by oligomerization or by binding to cellular glycosaminoglycans (GAGs). Upon truncation, the binding affinity for receptors and GAGs is altered, which influences chemokine function. The consequences of CD26-mediated clipping vary, as unchanged, enhanced, and reduced activities are reported. In tumors, CD26 most likely has the most profound effect on CXCL12 and the interferon (IFN)-inducible CXCR3 ligands, which are converted into receptor antagonists upon truncation. Depending on the tumor type, expression of CD26 is upregulated or downregulated and often results in the preferential generation of the chemokine isoform most favorable for tumor progression. Considering the tight relationship between chemokine sequence and chemokine binding specificity, molecules with the appropriate characteristics can be chemically engineered to provide innovative therapeutic strategies in a cancer setting.

TSPAN1, TMPRSS4, SDR16C5, and CTSE as Novel Panel for Pancreatic Cancer: A Bioinformatics Analysis and Experiments Validation

Pancreatic cancer is a lethal malignancy with a poor prognosis. This study aims to identify pancreatic cancer-related genes and develop a robust diagnostic model to detect this disease. Weighted gene co-expression network analysis (WGCNA) was used to determine potential hub genes for pancreatic cancer. Their mRNA and protein expression levels were validated through reverse transcription PCR (RT-PCR) and immunohistochemical (IHC). Diagnostic models were developed by eight machine learning algorithms and ten-fold cross-validation. Four hub genes (TSPAN1, TMPRSS4, SDR16C5, and CTSE) were identified based on bioinformatics. RT-PCR showed that the four hub genes were expressed at medium to high levels, IHC revealed that their protein expression levels were higher in pancreatic cancer tissues. For the panel of these four genes, eight models performed with 0.87-0.92 area under the curve value (AUC), 0.91-0.94 sensitivity, and 0.84-0.86 specificity in the validation cohort. In the external validation set, these models also showed good performance (0.86-0.98 AUC, 0.84-1.00 sensitivity, and 0.86-1.00 specificity). In conclusion, this study has identified four hub genes that might be closely related to pancreatic cancer: TSPAN1, TMPRSS4, SDR16C5, and CTSE. Four-gene panels might provide a theoretical basis for the diagnosis of pancreatic cancer.

Diagnostic significance of apical membranous and cytoplasmic dot-like CD26 expression in encapsulated follicular variant of papillary thyroid carcinoma: a useful marker for capsular invasion

Non-invasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) and invasive encapsulated follicular variant of papillary thyroid carcinoma (EFV-PTC) are indistinguishable preoperatively. CD26 expression in follicular tumor-uncertain malignant potential (FT-UMP) is reported to be clearly higher than in that without capsular invasion. To verify the diagnostic significance of CD26 immunostaining in EFV-PTC, we examined the expression pattern of CD26 in non-invasive EFV-PTC (NIFTP) and invasive EFV-PTC. We performed immunohistochemical analysis using CD26 antibody for 37 NIFTPs and 54 EFV-PTCs (34 minimally invasive EFV-PTCs and 20 widely invasive EFV-PTCs). Most NIFTP samples showed an apical membranous pattern or a cytoplasmic diffuse pattern of expression. Invasive EFV-PTCs more frequently showed a cytoplasmic dot-like pattern, and the labeling indices of tumor cells with cytoplasmic dot-like patterns were significantly higher than those in NIFTPs. The sizes of dots seen in NIFTPs (mean: 1.12 μm) were significantly smaller than in invasive EFV-PTCs (1.33 μm), minimally invasive EFV-PTC (1.27 μm), and widely invasive EFV-PTC (1.38 μm). We, therefore, conclude that cytoplasmic diffuse and/or cytoplasmic dot-like CD26 expression, particularly the larger CD26-positive dots, could be useful markers for capsular invasion in EFV-PTC. CD26 immunostaining, using cell blocks or cytological specimens, may preoperatively distinguish between NIFTP and invasive EFV-PTC.

BRAF mutation analysis by ARMS-PCR refines thyroid nodule management

BACKGROUND

Fine-needle aspiration (FNA) of thyroid nodules leads to nearly 25% indeterminate nodules, while BRAF^V600E mutation helps to predicting papillary thyroid carcinoma (PTC). However, the clinical validity and utility of the BRAF^V600E mutation detected using preoperative FNA samples in a large cohort were rarely reported.

AIM

To explore the clinical significance of the BRAF^V600E mutation on preoperative diagnosis and decision-making in a large FNA cohort in China.

DESIGN

This was a prospective study of BRAF^V600E mutation analysis using an amplification refractory mutation system-polymerase chain reaction (ARMS-PCR) in FNA samples.

PATIENTS

The study involved 2640 samples from 2307 patients undergoing FNA procedures in a Chinese medical centre.

RESULTS

A high mutation rate of 86.7% was found in the PTC population. For indeterminate thyroid nodules, the malignant rate of BRAF^V600E+ and BRAF^V600E- was 87.8% and 39.5% in the Bethesda System for Reporting Thyroid Cytopathology (BSRTC) III, and 88.2% and 31.8% in the BSRTC IV, respectively. A cost-effective diagnostic model combining both BSRTC grading and BRAF^V600E mutation status showed the highest sensitivity of 82.9% and specificity of 85.4%. Central lymph node metastasis (CLNM) was independent of the BRAF mutation status and accounted for 34.2% of the patients with PTC. C_T values of BRAF^V600E of patients with PTMC were significantly lower in young patients and patients with CLNM.

CONCLUSIONS

The combined analysis of cytological results and BRAF^V600E mutation is highly recommended in BRAF^V600E high-prevalence regions, including China. Prophylactic central neck dissection should be performed in selected patients without regard to the BRAF mutation status.

Development of novel monoclonal antibodies with specific binding affinity for denatured human CD26 in formalin-fixed paraffin-embedded and decalcified specimens

A 110-kDa type II transmembrane glycoprotein with dipeptidyl peptidase IV (DPPIV) activity in its extracellular region, CD26 has a multitude of biological functions and plays an important role in the regulation of inflammatory responses and tumor biology. Our work has focused on CD26 as a novel therapeutic target for various tumors and immune disorders, and we have recently developed a humanized anti-CD26 monoclonal antibody (mAb), YS110, which has promising safety profile and clinical activity in patients with malignant pleural mesothelioma. The development of an anti-human CD26 mAb that can clearly and reliably detect the denatured CD26 molecule in formalin-fixed paraffin-embedded (FFPE) tissues in the clinical setting is therefore of the utmost importance. To develop novel anti-CD26 mAbs capable of binding to denatured CD26, we immunized mice with urea-treated CD26 protein. Hybridoma supernatants were screened for specific reactivity with human CD26 by immunostaining through the use of a set of FFPE human CD26-positive or negative tumor cell lines. This screening method enables us to develop novel anti-human CD26 mAbs suitable for immunohistochemical staining of CD26 in FFPE non-tumor and tumor tissue sections with reliable clarity and intensity. Specifically, these mAbs display strong binding affinity to denatured human CD26 rather than undenatured human CD26, and are capable of detecting denatured human CD26 in decalcified specimens. These novel anti-CD26 mAbs are potentially useful for the analysis of CD26 expression in cancer patients with bony metastasis, and may help decide the appropriateness of YS110 therapy for future cancer patients.

A Novel Joint Gene Set Analysis Framework Improves Identification of Enriched Pathways in Cross Disease Transcriptomic Analysis

Motivation: Gene set enrichment analysis is a widely accepted expression analysis tool which aims at detecting coordinated expression change within a pre-defined gene sets rather than individual genes. The benefit of gene set analysis over individual differentially expressed (DE) gene analysis includes more reproducible and interpretable results and detecting small but consistent change among gene set which could not be detected by DE gene analysis. There have been many successful gene set analysis applications in human diseases. However, when the sample size of a disease study is small and no other public data sets of the same disease are available, it will lead to lack of power to detect pathways of importance to the disease. Results: We have developed a novel joint gene set analysis statistical framework which aims at improving the power of identifying enriched gene sets through integrating multiple similar disease data sets. Through comprehensive simulation studies, we demonstrated that our proposed frameworks obtained much better AUC scores than single data set analysis and another meta-analysis method in identification of enriched pathways. When applied to two real data sets, the proposed framework could retain the enriched gene sets identified by single data set analysis and exclusively obtained up to 200% more disease-related gene sets demonstrating the improved identification power through information shared between similar diseases. We expect that the proposed framework would enable researchers to better explore public data sets when the sample size of their study is limited.

CD26/DPP4 - a potential biomarker and target for cancer therapy

CD26/dipeptidyl peptidase (DPP)4 is a membrane-bound protein found in many cell types of the body, and a soluble form is present in body fluids. There is longstanding evidence that various primary tumors and also metastases express CD26/DPP4 to a variable extent. By cleaving dipeptides from peptides with a proline or alanine in the penultimate position at the N-terminus, it regulates the activity of incretin hormones, chemokines and many other peptides. Due to these effects and interactions with other molecules, a tumor promoting or suppressing role can be attributed to CD26/DPP4. In this review, we discuss the existing evidence on the expression of soluble or membrane-bound CD26/DPP4 in malignant diseases, along with the most recent findings on CD26/DPP4 as a therapeutic target in specific malignancies. The expression and possible involvement of the related DPP8 and DPP9 in cancer are also reviewed. A higher expression of CD26/DPP4 is found in a wide variety of tumor entities, however more research on CD26/DPP4 in the tumor microenvironment is needed to fully explore its use as a tumor biomarker. Circulating soluble CD26/DPP4 has also been studied as a cancer biomarker, however, the observed decrease in most cancer patients does not seem to be cancer specific. Encouraging results from experimental work and a recently reported first phase clinical trial targeting CD26/DPP4 in mesothelioma, renal and urological tumors pave the way for follow-up clinical studies, also in other tumor entities, possibly leading to the development of more effective complementary therapies against cancer.

P-glycoprotein-mediated chemoresistance is reversed by carbonic anhydrase XII inhibitors

Carbonic anhydrase XII (CAXII) is a membrane enzyme that maintains pH homeostasis and sustains optimum P-glycoprotein (Pgp) efflux activity in cancer cells. Here, we investigated a panel of eight CAXII inhibitors (compounds 1–8), for their potential to reverse Pgp mediated tumor cell chemoresistance. Inhibitors (5 nM) were screened in human and murine cancer cells (colon, lung, breast, bone) with different expression levels of CAXII and Pgp. We identified three CAXII inhibitors (compounds 1, 2 and 4) that significantly (≥ 2 fold) increased the intracellular retention of the Pgp-substrate and chemotherapeutic doxorubicin, and restored its cytotoxic activity. The inhibitors lowered intracellular pH to indirectly impair Pgp activity. Ca12-knockout assays confirmed that the chemosensitizing property of the compounds was dependent on active CAXII. Furthermore, in a preclinical model of drug-resistant breast tumors compound 1 (1900 ng/kg) restored the efficacy of doxorubicin to the same extent as the direct Pgp inhibitor tariquidar. The expression of carbonic anhydrase IX had no effect on the intracellular doxorubicin accumulation. Our work provides strong evidence that CAXII inhibitors are effective chemosensitizer agents in CAXII-positive and Pgp-positive cancer cells. The use of CAXII inhibitors may represent a turning point in combinatorial chemotherapeutic schemes to treat multidrug-resistant tumors.

Dipeptidyl Peptidase IV as a Prognostic Marker and Therapeutic Target in Papillary Thyroid Carcinoma

CONTEXT

Dipeptidyl peptidase IV (DPP4) is overexpressed in thyroid cancer and certain malignancies. Furthermore, DPP4 has been identified as a discriminatory marker for thyroid cancer. However, it remains unclear whether DPP4 expression plays a prognostic role.

OBJECTIVE

The aim of this study was to investigate the expression and function of DPP4 in thyroid cancer and the mechanisms involved.

DESIGN

We determined the expression of DPP4 by immunohistochemistry in tissue microarrays of thyroid tumors. In vitro functional studies were performed after genetic and pharmacological inhibition of DPP4. Gene expression and pathway analyses were used to identify downstream targets. The therapeutic potential of DPP4 inhibition was evaluated in a mouse xenograft model.

RESULTS

High DPP4 expression was associated with extrathyroidal extension (P < 0.001), BRAF mutation (P < 0.001), and advanced tumor stage (P = 0.007) in papillary thyroid cancer. Patients in the high-DPP4 expression group were less likely to be classified as having no evidence of disease at final follow-up (P = 0.042). DPP4 silencing or treatment with DPP4 inhibitors significantly suppressed colony formation, cell migration, and invasion. Analysis of differentially expressed genes after DPP4 knockdown suggested that the transforming growth factor-β signaling pathway is involved. In vivo experiments revealed that sitagliptin treatment reduced tumor growth and xenograft transforming growth factor-β receptor I expression.

CONCLUSIONS

Increased DPP4 expression is associated with cellular invasion and more aggressive disease in papillary thyroid cancer. Targeting DPP4 may be a therapeutic strategy for DPP4-expressing thyroid cancer.

Ultrasound-based differentiation of malignant and benign thyroid Nodules: An extreme learning machine approach

BACKGROUND AND OBJECTIVES

It is important to be able to accurately distinguish between benign and malignant thyroid nodules in order to make appropriate clinical decisions. The purpose of this study was to improve the effectiveness and efficiency for discriminating the malignant from benign thyroid cancers based on the Ultrasonography (US) features.

METHODS

There were 114 benign nodules in 106 patients (82 women and 24 men) and 89 malignant nodules in 81 patients (69 women and 12 men) included in this study. The potential of extreme learning machine (ELM) has been explored for the first time to discriminate malignant and benign thyroid nodules based on the sonographic features in ultrasound images. The influence of two key parameters (the number of hidden neurons and type of activation function) on the performance of ELM was investigated. The relationship between feature subsets obtained by the feature selection method and the classification performance of ELM was also examined. A real-life dataset was used to evaluate the effectiveness of the proposed method in terms of classification accuracy, sensitivity, specificity, and area under the ROC (receiver operating characteristic) curve (AUC).

RESULTS

The results demonstrate that there are significant differences between the malignant and benign thyroid nodules (p-value<0.01), the most discriminative features are echogenicity, calcification, margin, composition and shape. Compared with other methods, the proposed method not only has achieved very promising classification accuracy via 10-fold cross-validation (CV) scheme, but also greatly reduced the computational cost compared to other counterparts. The proposed ELM-based approach achieves 87.72% ACC, 0.8672 AUC, 78.89% sensitivity, and 94.55% specificity.

CONCLUSIONS

Based on the empirical analysis, the proposed ELM-based approach for thyroid cancer detection has promising potential in clinical use, and it can be of assistance as an optional tool for the clinicians.

A disease similarity matrix based on the uniqueness of shared genes

Background

Complex diseases involve many genes, and these genes are often associated with several different illnesses. Disease similarity measurement can be based on shared genotype or phenotype. Quantifying relationships between genes can reveal previously unknown connections and form a reference base for therapy development and drug repurposing.

Methods

Here we introduce a method to measure disease similarity that incorporates the uniqueness of shared genes. For each disease pair, we calculated the uniqueness score and constructed disease similarity matrices using OMIM and Disease Ontology annotation.

Results

Using the Disease Ontology-based matrix, we identified several interesting connections between cancer and other disease and conditions such as malaria, along with studies to support our findings. We also found several high scoring pairwise relationships for which there was little or no literature support, highlighting potentially interesting connections warranting additional study.

Conclusions

We developed a co-occurrence matrix based on gene uniqueness to examine the relationships between diseases from OMIM and DORIF data. Our similarity matrix can be used to identify potential disease relationships and to motivate further studies investigating the causal mechanisms in diseases.

High-dimensional omics data analysis using a variable screening protocol with prior knowledge integration (SKI)

BACKGROUND

High-throughput technology could generate thousands to millions biomarker measurements in one experiment. However, results from high throughput analysis are often barely reproducible due to small sample size. Different statistical methods have been proposed to tackle this "small n and large p" scenario, for example different datasets could be pooled or integrated together to provide an effective way to improve reproducibility. However, the raw data is either unavailable or hard to integrate due to different experimental conditions, thus there is an emerging need to develop a method for "knowledge integration" in high-throughput data analysis.

RESULTS

In this study, we proposed an integrative prescreening approach, SKI, for high-throughput data analysis. A new rank is generated based on two initial ranks: (1) knowledge based rank; and (2) marginal correlation based rank. Our simulation shows the SKI outperforms other methods without knowledge-integration in terms of higher true positive rate given the same number of variables selected. We also applied our method in a drug response study and found its performance to be better than regular screening methods.

CONCLUSION

The proposed method provides an effective way to integrate knowledge for high-throughput analysis. It could easily implemented with our provided R package named SKI.

A panel of four genes accurately differentiates benign from malignant thyroid nodules

Background

Clinicians are confronted with an increasing number of patients with thyroid nodules. Reliable preoperative diagnosis of thyroid nodules remains a challenge because of inconclusive cytological examination of fine-needle aspiration biopsies. Although molecular analysis of thyroid tissue has shown promise as a diagnostic tool in recent years, it has not been successfully applied in routine clinical use, particularly in Chinese patients.

Methods

Whole-transcriptome sequencing of 19 primary papillary thyroid cancer (PTC) samples and matched adjacent normal thyroid tissue (NT) samples were performed. Bioinformatics analysis was carried out to identify candidate diagnostic genes. Then, RT-qPCR was performed to evaluate these candidate genes, and four genes were finally selected. Based on these four genes, diagnostic algorithm was developed (training set: 100 thyroid cancer (TC) and 65 benign thyroid lesions (BTL)) and validated (independent set: 123 TC and 81 BTL) using the support vector machine (SVM) approach.

Results

We discovered four genes, namely fibronectin 1 (FN1), gamma-aminobutyric acid type A receptor beta 2 subunit (GABRB2), neuronal guanine nucleotide exchange factor (NGEF) and high-mobility group AT-hook 2 (HMGA2). A SVM model with these four genes performed with 97.0 % sensitivity, 93.8 % specificity, 96.0 % positive predictive value (PPV), and 95.3 % negative predictive value (NPV) in training set. For additional independent validation, it also showed good performance (92.7 % sensitivity, 90.1 % specificity, 93.4 % PPV, and 89.0 % NPV).

Conclusions

Our diagnostic panel can accurately distinguish benign from malignant thyroid nodules using a simple and affordable method, which may have daily clinical application in the near future.

Electronic supplementary material

The online version of this article (doi:10.1186/s13046-016-0447-3) contains supplementary material, which is available to authorized users.

Classification of Benign and Malignant Thyroid Nodules Using a Combined Clinical Information and Gene Expression Signatures

BACKGROUND

A key challenge in thyroid carcinoma is preoperatively diagnosing malignant thyroid nodules. A novel diagnostic test that measures the expression of a 3-gene signature (DPP4, SCG5 and CA12) has demonstrated promise in thyroid carcinoma assessment. However, more reliable prediction methods combining clinical features with genomic signatures with high accuracy, good stability and low cost are needed.

METHODOLOGY/PRINCIPAL FINDINGS

25 clinical information were recorded in 771 patients. Feature selection and validation were conducted using random forest. Thyroid samples and clinical data were obtained from 142 patients at two different hospitals, and expression of the 3-gene signature was measured using quantitative PCR. The predictive abilities of three models (based on the selected clinical variables, the gene expression profile, and integrated gene expression and clinical information) were compared. Seven clinical characteristics were selected based on a training set (539 patients) and tested in three test sets, yielding predictive accuracies of 82.3% (n = 232), 81.4% (n = 70), and 81.9% (n = 72). The predictive sensitivity, specificity, and accuracy were 72.3%, 80.5% and 76.8% for the model based on the gene expression signature, 66.2%, 81.8% and 74.6% for the model based on the clinical data, and 83.1%, 84.4% and 83.8% for the combined model in a 10-fold cross-validation (n = 142).

CONCLUSIONS

These findings reveal that the integrated model, which combines clinical data with the 3-gene signature, is superior to models based on gene expression or clinical data alone. The integrated model appears to be a reliable tool for the preoperative diagnosis of thyroid tumors.