Combined microRNA and mRNA microfluidic TaqMan array cards for the diagnosis of malignancy of multiple types of pancreatico-biliary tumors in fine-needle aspiration material

Abstract

Pancreatic ductal adenocarcinoma (PDAC) continues to carry the lowest survival rates among all solid tumors. A marked resistance against available therapies, late clinical presentation and insufficient means for early diagnosis contribute to the dismal prognosis. Novel biomarkers are thus required to aid treatment decisions and improve patient outcomes. We describe here a multi-omics molecular platform that allows for the first time to simultaneously analyze miRNA and mRNA expression patterns from minimal amounts of biopsy material on a single microfluidic TaqMan Array card. Expression profiles were generated from 113 prospectively collected fine needle aspiration biopsies (FNAB) from patients undergoing surgery for suspect masses in the pancreas. Molecular classifiers were constructed using support vector machines, and rigorously evaluated for diagnostic performance using 10×10fold cross validation. The final combined miRNA/mRNA classifier demonstrated a sensitivity of 91.7%, a specificity of 94.5%, and an overall diagnostic accuracy of 93.0% for the differentiation between PDAC and benign pancreatic masses, clearly outperfoming miRNA-only classifiers. The classification algorithm also performed very well in the diagnosis of other types of solid tumors (acinar cell carcinomas, ampullary cancer and distal bile duct carcinomas), but was less suited for the diagnostic analysis of cystic lesions. We thus demonstrate that simultaneous analysis of miRNA and mRNA biomarkers from FNAB samples using multi-omics TaqMan Array cards is suitable to differentiate suspect solid pancreatic masses with high precision.

Keywords: fine needle aspiration biopsy; microfluidic TaqMan arrays; molecular diagnostics; pancreatic cancer; pancreatico-biliary tumors.

Conflict of interest statement

CONFLICTS OF INTEREST The authors declare no conflicts of interest.

Figures

Figure 1. Study design

Figure 2. Schematic representation of 10×10 cross validation procedure

The original dataset is split into ten subsets of approximately equal size. Nine of these subsets are used for training the classification model (SVM). The tenth subset is used as an independent subset for evaluating the performance of the trained classifier. This procedure is repeated for each subset and an overall accuracy is calculated as the mean accuracy over ten permutations of the original dataset.

Figure 3. Exhaustive evaluation of all miRNA marker combinations

All possible combinations of miRNA markers, from single miRNAs to the combination of all 9 markers (29-1 = 511 possible combinations), were tested for their diagnostic performances on the set of PDAC and CP samples. Each combination was evaluated by a linear SVM and sensitivity, specificity and diagnostic accuracy determined by 10×10-fold cross validation. Results are plotted in decreasing order according to accuracy. Note that none of the resulting classifiers exceeded a diagnostic accuracy of 75%, with the best-performing combination achieving a sensitivity of 58.5% and a specificity of 88.3%. “CP” = chronic pancreatitis; “PDAC” = pancreatic ductal adenocarcinoma; “Acc” = accuracy; “Sens” = sensitivity; “Spec” = specificity.

Figure 4. Unsupervised clustering of expression profiles from PDAC and CP biopsies

Individual samples are shown in columns; genes are shown in rows. Both the samples and the genes were hierarchically clustered (complete linkage clustering) using uncentered Pearson correlation as the similarity measure. Red cells indicate high expression, black intermediate expression, and green low expression of a gene in the respective group. Clusters of CP (blue bars) and PDAC samples (yellow bars) are readily apparent. “CP” = chronic pancreatitis; “PDAC” = pancreatic ductal adenocarcinoma.

Figure 5. Training of mixed miRNA / mRNA classifiers

Combined profiles of miRNA and mRNA markers (Panel i) were analyzed in a sequence of marker selection and classification experiment. First, a subset of candidate markers was selected via the univariate TNoMcw score (Panel ii). The reduced marker profiles were then used to train a multivariate SVM (Panel iii). The final decision can be projected to univariate space (Panel iv).

Figure 6. Evaluation of the combined miRNA/mRNA marker profiles

Accuracy, sensitivity and specificity gained in the 10×10 cross validation experiments with the combined miRNA/mRNA marker profiles. Classification models vary in the number of markers that were selected during their training phases. Results are given in increasing order from 1 to 88 selected biomarkers. Accuracies ≥ 90% are marked in yellow. “CP” = chronic pancreatitis; “PDAC” = pancreatic ductal adenocarcinoma.

Figure 7. Performance of the final classifier in identifying malignancy on biopsy samples from pancreatobiliary tumors other than PDAC

Retraining of the optimal model (77 markers) on all available samples. Prediction of PDAC and CP samples and categorization of all remaining tissue-types is given. “CP” = chronic pancreatitis; “PDAC” = pancreatic ductal adenocarcinoma; “CA” = carcinoma; “IPMN” = intraductal papillary mucinous neoplasm; “SPT” = solid pseudopapillary neoplasia.