The Role of Different TET Proteins in Cytosine Demethylation Revealed by Mathematical Modeling.

In living cells, some reactions can be conducted by more than one enzyme and sometimes it is difficult to establish which enzyme is responsible. Such is the case with proteins from the TET family, capable of converting 5-methyl-2'-deoxycytidine (5-mdC) in DNA to 5-(hydroxymethyl)-2'-deoxycytidine (5-hmdC) and further to 5-formyl-2'-deoxycytidine (5-fdC) and 5-carboxy-2'-deoxycytidine (5-cadC). The estimation of the efficiency of particular TETs in particular oxidative reactions and different cell types is important but experimentally difficult. Here, we propose an approach with mathematical modeling in which methylation and known deoxycytidine modification pathways are presented by 343 possible model versions with assumed different combinations of TET1, 2, and 3 activities in different pathways. Model parameters were calculated on the basis of 5-mdC, 5-hmdC, 5-fdC, 5-cadC, and 5-hmdU levels experimentally assessed in five human cultured cell lines and previously published. Selection of the model versions that give in simulations the best average fit to experimental data suggested that not all TET proteins participate in all modification reactions and that TET3 activity may be especially important in the reaction of 5-fdC removal.

Radiomic signature accurately predicts the risk of metastatic dissemination in late-stage non-small cell lung cancer.

Background: Non-small cell lung cancer (NSCLC) is the most common type of lung cancer, and the median overall survival (OS) is approximately 2-3 years among patients with stage III disease. Furthermore, it is one of the deadliest types of cancer globally due to non-specific symptoms and the lack of a biomarker for early detection. The most important decision that clinicians need to make after a lung cancer diagnosis is the selection of a treatment schedule. This decision is based on, among others factors, the risk of developing metastasis. Methods: A cohort of 115 NSCLC patients treated using chemotherapy and radiotherapy (RT) with curative intent was retrospectively collated and included patients for whom positron emission tomography/computed tomography (PET/CT) images, acquired before RT, were available. The PET/CT images were used to compute radiomic features extracted from a region of interest (ROI), the primary tumor. Radiomic and clinical features were then classified to stratify the patients into short and long time to metastasis, and regression analysis was used to predict the risk of metastasis. Results: Classification based on binarized metastasis-free survival (MFS) was applied with moderate success. Indeed, an accuracy of 0.73 was obtained for the selection of features based on the Wilcoxon test and logistic regression model. However, the Cox regression model for metastasis risk prediction performed very well, with a concordance index (C-index) score equal to 0.84. Conclusions: It is possible to accurately predict the risk of metastasis in NSCLC patients based on radiomic features. The results demonstrate the potential use of features extracted from cancer imaging in predicting the risk of metastasis.

Data Integration-Possibilities of Molecular and Clinical Data Fusion on the Example of Thyroid Cancer Diagnostics.

(1) Background: The data from independent gene expression sources may be integrated for the purpose of molecular diagnostics of cancer. So far, multiple approaches were described. Here, we investigated the impacts of different data fusion strategies on classification accuracy and feature selection stability, which allow the costs of diagnostic tests to be reduced. (2) Methods: We used molecular features (gene expression) combined with a feature extracted from the independent clinical data describing a patient's sample. We considered the dependencies between selected features in two data fusion strategies (early fusion and late fusion) compared to classification models based on molecular features only. We compared the best accuracy classification models in terms of the number of features, which is connected to the potential cost reduction of the diagnostic classifier. (3) Results: We show that for thyroid cancer, the extracted clinical feature is correlated with (but not redundant to) the molecular data. The usage of data fusion allows a model to be obtained with similar or even higher classification quality (with a statistically significant accuracy improvement, a p-value below 0.05) and with a reduction in molecular dimensionality of the feature space from 15 to 3-8 (depending on the feature selection method). (4) Conclusions: Both strategies give comparable quality results, but the early fusion method provides better feature selection stability.

Impact of government policies on the COVID-19 pandemic unraveled by mathematical modelling.

Since the very beginning of the COVID-19 pandemic, control policies and restrictions have been the hope for containing the rapid spread of the virus. However, the psychological and economic toll they take on society entails the necessity to develop an optimal control strategy. Assessment of the effectiveness of these interventions aided with mathematical modelling remains a non-trivial issue in terms of numerical conditioning due to the high number of parameters to estimate from a highly noisy dataset and significant correlations between policy timings. We propose a solution to the problem of parameter non-estimability utilizing data from a set of European countries. Treating a subset of parameters as common for all countries and the rest as country-specific, we construct a set of individualized models incorporating 13 different pandemic control measures, and estimate their parameters without prior assumptions. We demonstrate high predictive abilities of these models on an independent validation set and rank the policies by their effectiveness in reducing transmission rates. We show that raising awareness through information campaigns, providing income support, closing schools and workplaces, cancelling public events, and maintaining an open testing policy have the highest potential to mitigate the pandemic.

Classification of Thyroid Tumors Based on Mass Spectrometry Imaging of Tissue Microarrays; a Single-Pixel Approach.

The primary diagnosis of thyroid tumors based on histopathological patterns can be ambiguous in some cases, so proper classification of thyroid diseases might be improved if molecular biomarkers support cytological and histological assessment. In this work, tissue microarrays representative for major types of thyroid malignancies-papillary thyroid cancer (classical and follicular variant), follicular thyroid cancer, anaplastic thyroid cancer, and medullary thyroid cancer-and benign thyroid follicular adenoma and normal thyroid were analyzed by mass spectrometry imaging (MSI), and then different computation approaches were implemented to test the suitability of the registered profiles of tryptic peptides for tumor classification. Molecular similarity among all seven types of thyroid specimens was estimated, and multicomponent classifiers were built for sample classification using individual MSI spectra that corresponded to small clusters of cells. Moreover, MSI components showing the most significant differences in abundance between the compared types of tissues detected and their putative identity were established by annotation with fragments of proteins identified by liquid chromatography-tandem mass spectrometry in corresponding tissue lysates. In general, high accuracy of sample classification was associated with low inter-tissue similarity index and a high number of components with significant differences in abundance between the tissues. Particularly, high molecular similarity was noted between three types of tumors with follicular morphology (adenoma, follicular cancer, and follicular variant of papillary cancer), whose differentiation represented the major classification problem in our dataset. However, low level of the intra-tissue heterogeneity increased the accuracy of classification despite high inter-tissue similarity (which was exemplified by normal thyroid and benign adenoma). We compared classifiers based on all detected MSI components (n = 1536) and the subset of the most abundant components (n = 147). Despite relatively higher contribution of components with significantly different abundance and lower overall inter-tissue similarity in the latter case, the precision of classification was generally higher using all MSI components. Moreover, the classification model based on individual spectra (a single-pixel approach) outperformed the model based on mean spectra of tissue cores. Our result confirmed the high feasibility of MSI-based approaches to multi-class detection of cancer types and proved the good performance of sample classification based on individual spectra (molecular image pixels) that overcame problems related to small amounts of heterogeneous material, which limit the applicability of classical proteomics.

Differences in Gene Expression Profile of Primary Tumors in Metastatic and Non-Metastatic Papillary Thyroid Carcinoma-Do They Exist?

Molecular mechanisms of distant metastases (M1) in papillary thyroid cancer (PTC) are poorly understood. We attempted to analyze the gene expression profile in PTC primary tumors to seek the genes associated with M1 status and characterize their molecular function. One hundred and twenty-three patients, including 36 M1 cases, were subjected to transcriptome oligonucleotide microarray analyses: (set A-U133, set B-HG 1.0 ST) at transcript and gene group level (limma, gene set enrichment analysis (GSEA)). An additional independent set of 63 PTCs, including 9 M1 cases, was used to validate results by qPCR. The analysis on dataset A detected eleven transcripts showing significant differences in expression between metastatic and non-metastatic PTC. These genes were validated on microarray dataset B. The differential expression was positively confirmed for only two genes: IGFBP3, (most significant) and ECM1. However, when analyzed on an independent dataset by qPCR, the IGFBP3 gene showed no differences in expression. Gene group analysis showed differences mainly among immune-related transcripts, indicating the potential influence of tumor immune infiltration or signal within the primary tumor. The differences in gene expression profile between metastatic and non-metastatic PTC, if they exist, are subtle and potentially detectable only in large datasets.

A mathematical model as a tool to identify microRNAs with highest impact on transcriptome changes.

BACKGROUND: Rapid changes in the expression of many messenger RNA (mRNA) species follow exposure of cells to ionizing radiation. One of the hypothetical mechanisms of this response may include microRNA (miRNA) regulation, since the amounts of miRNAs in cells also vary upon irradiation. To address this possibility, we designed experiments using cancer-derived cell lines transfected with luciferase reporter gene containing sequences targeted by different miRNA species in its 3'- untranslated region. We focus on the early time-course response (1 h past irradiation) to eliminate secondary mRNA expression waves. RESULTS: Experiments revealed that the irradiation-induced changes in the mRNA expression depend on the miRNAs which interact with mRNA. To identify the strongest interactions, we propose a mathematical model which predicts the mRNA fold expression changes, caused by perturbation of microRNA-mRNA interactions. Model was applied to experimental data including various cell lines, irradiation doses and observation times, both ours and literature-based. Comparison of modelled and experimental mRNA expression levels given miRNA level changes allows estimating how many and which miRNAs play a significant role in transcriptome response to stress conditions in different cell types. As an example, in the human melanoma cell line the comparison suggests that, globally, a major part of the irradiation-induced changes of mRNA expression can be explained by perturbed miRNA-mRNA interactions. A subset of about 30 out of a few hundred miRNAs expressed in these cells appears to account for the changes. These miRNAs play crucial roles in regulatory mechanisms observed after irradiation. In addition, these miRNAs have a higher average content of GC and a higher number of targeted transcripts, and many have been reported to play a role in the development of cancer. CONCLUSIONS: Our proposed mathematical modeling approach may be used to identify miRNAs which participate in responses of cells to ionizing radiation, and other stress factors such as extremes of temperature, exposure to toxins, and drugs.

Modulation by neighboring cells of the responses and fate of melanoma cells irradiated with UVA.

UVA radiation, which accounts for about 95% of the solar spectrum, contributes to and may be the etiological factor of skin cancers of which malignant melanoma is the most aggressive. UVA causes oxidative stress in various types of cells in the skin, keratinocyte, melanocytes, and fibroblasts, which is responsible for its cytotoxic effect. Here we used a transwell system to explore how the responses of melanoma cells to a low dose of UVA (20kJ/m2, ~10% of the minimal erythema dose) are influenced by neighboring co-cultured melanoma cells or fibroblasts. This dose had a low toxicity for melanoma cells, but after irradiation, co-culture with non-irradiated melanoma cells caused a strong decline in their viability and an increased frequency of apoptosis, whereas co-culture with fibroblast exerted a protective effect on irradiated melanoma cells. At the same time, the presence of non-irradiated cells, especially fibroblasts, decreased the level of UVA-induced reactive oxygen and nitrogen species. Interleukins efficiently produced by fibroblasts seem to be main players in these effects. Our studies reveal that coexistence of fibroblasts with melanoma cells may strongly modulate the direct action and may change bystander effects exerted by UVA light. Similar modulation of the effect of UVA on melanoma cells in vivo by bystander-like signaling from neighboring cells would have consequences for the development of malignant melanoma.

Estimation of initial functions for systems with delays from discrete measurements.

The work presents a gradient-based approach to estimation of initial functions of time delay elements appearing in models of dynamical systems. It is shown how to generate the gradient of the estimation objective function in the initial function space using adjoint sensitivity analysis. It is assumed that the system is continuous-time and described by ordinary differential equations with delays but the estimation is done based on discrete-time measurements of the signals appearing in the system. Results of gradient-based estimation of initial functions for exemplary models are presented and discussed.

Adjoint sensitivity analysis of a tumor growth model and its application to spatiotemporal radiotherapy optimization.

We investigate a spatial model of growth of a tumor and its sensitivity to radiotherapy. It is assumed that the radiation dose may vary in time and space, like in intensity modulated radiotherapy (IMRT). The change of the final state of the tumor depends on local differences in the radiation dose and varies with the time and the place of these local changes. This leads to the concept of a tumor's spatiotemporal sensitivity to radiation, which is a function of time and space. We show how adjoint sensitivity analysis may be applied to calculate the spatiotemporal sensitivity of the finite difference scheme resulting from the partial differential equation describing the tumor growth. We demonstrate results of this approach to the tumor proliferation, invasion and response to radiotherapy (PIRT) model and we compare the accuracy and the computational effort of the method to the simple forward finite difference sensitivity analysis. Furthermore, we use the spatiotemporal sensitivity during the gradient-based optimization of the spatiotemporal radiation protocol and present results for different parameters of the model.

The different radiation response and radiation-induced bystander effects in colorectal carcinoma cells differing in p53 status.

Radiation-induced bystander effect, appearing as different biological changes in cells that are not directly exposed to ionizing radiation but are under the influence of molecular signals secreted by irradiated neighbors, have recently attracted considerable interest due to their possible implication for radiotherapy. However, various cells present diverse radiosensitivity and bystander responses that depend, inter alia, on genetic status including TP53, the gene controlling the cell cycle, DNA repair and apoptosis. Here we compared the ionizing radiation and bystander responses of human colorectal carcinoma HCT116 cells with wild type or knockout TP53 using a transwell co-culture system. The viability of exposed to X-rays (0-8 Gy) and bystander cells of both lines showed a roughly comparable decline with increasing dose. The frequency of micronuclei was also comparable at lower doses but at higher increased considerably, especially in bystander TP53-/- cells. Moreover, the TP53-/- cells showed a significantly elevated frequency of apoptosis, while TP53+/+ counterparts expressed high level of senescence. The cross-matched experiments where irradiated cells of one line were co-cultured with non-irradiated cells of opposite line show that both cell lines were also able to induce bystander effects in their counterparts, however different endpoints revealed with different strength. Potential mediators of bystander effects, IL-6 and IL-8, were also generated differently in both lines. The knockout cells secreted IL-6 at lower doses whereas wild type cells only at higher doses. Secretion of IL-8 by TP53-/- control cells was many times lower than that by TP53+/+ but increased significantly after irradiation. Transcription of the NFκBIA was induced in irradiated TP53+/+ mainly, but in bystanders a higher level was observed in TP53-/- cells, suggesting that TP53 is required for induction of NFκB pathway after irradiation but another mechanism of activation must operate in bystander cells.

Analysis options for high-throughput sequencing in miRNA expression profiling.

BACKGROUND: Recently high-throughput sequencing (HTS) using next generation sequencing techniques became useful in digital gene expression profiling.Our study introduces analysis options for HTS data based on mapping to miRBase or counting and grouping of identical sequence reads. Those approaches allow a hypothesis free detection of miRNA differential expression. METHODS: We compare our results to microarray and qPCR data from one set of RNA samples. We use Illumina platforms for microarray analysis and miRNA sequencing of 20 samples from benign follicular thyroid adenoma and malignant follicular thyroid carcinoma. Furthermore, we use three strategies for HTS data analysis to evaluate miRNA biomarkers for malignant versus benign follicular thyroid tumors. RESULTS: High correlation of qPCR and HTS data was observed for the proposed analysis methods. However, qPCR is limited in the differential detection of miRNA isoforms. Moreover, we illustrate a much broader dynamic range of HTS compared to microarrays for small RNA studies. Finally, our data confirm hsa-miR-197-3p, hsa-miR-221-3p, hsa-miR-222-3p and both hsa-miR-144-3p and hsa-miR-144-5p as potential follicular thyroid cancer biomarkers. CONCLUSIONS: Compared to microarrays HTS provides a global profile of miRNA expression with higher specificity and in more detail. Summarizing of HTS reads as isoform groups (analysis pipeline B) or according to functional criteria (seed analysis pipeline C), which better correlates to results of qPCR are promising new options for HTS analysis. Finally, data opens future miRNA research perspectives for HTS and indicates that qPCR might be limited in validating HTS data in detail.

On the sensitivity of feature ranked lists for large-scale biological data.

The problem of feature selection for large-scale genomic data, for example from DNA microarray experiments, is one of the fundamental and well-investigated problems in modern computational biology. From the computational point of view, a selected gene list should be characterized by good predictive power and should be understood and well explained from the biological point of view. Recently, another feature of selected gene lists is increasingly investigated, namely their stability which measures how the content and/or the gene order change when the data are perturbed. In this paper we propose a new approach to analysis of gene list stability, termed the sensitivity index, that does not require any data perturbation and allows the gene list that is most reliable in a biological sense to be chosen.

Repair of DNA strand breaks in a minichromosome in vivo: kinetics, modeling, and effects of inhibitors.

To obtain an overall picture of the repair of DNA single and double strand breaks in a defined region of chromatin in vivo, we studied their repair in a ~170 kb circular minichromosome whose length and topology are analogous to those of the closed loops in genomic chromatin. The rate of repair of single strand breaks in cells irradiated with γ photons was quantitated by determining the sensitivity of the minichromosome DNA to nuclease S1, and that of double strand breaks by assaying the reformation of supercoiled DNA using pulsed field electrophoresis. Reformation of supercoiled DNA, which requires that all single strand breaks have been repaired, was not slowed detectably by the inhibitors of poly(ADP-ribose) polymerase-1 NU1025 or 1,5-IQD. Repair of double strand breaks was slowed by 20-30% when homologous recombination was supressed by KU55933, caffeine, or siRNA-mediated depletion of Rad51 but was completely arrested by the inhibitors of nonhomologous end-joining wortmannin or NU7441, responses interpreted as reflecting competition between these repair pathways similar to that seen in genomic DNA. The reformation of supercoiled DNA was unaffected when topoisomerases I or II, whose participation in repair of strand breaks has been controversial, were inhibited by the catalytic inhibitors ICRF-193 or F11782. Modeling of the kinetics of repair provided rate constants and showed that repair of single strand breaks in minichromosome DNA proceeded independently of repair of double strand breaks. The simplicity of quantitating strand breaks in this minichromosome provides a usefull system for testing the efficiency of new inhibitors of their repair, and since the sequence and structural features of its DNA and its transcription pattern have been studied extensively it offers a good model for examining other aspects of DNA breakage and repair.

Stable feature selection and classification algorithms for multiclass microarray data.

BACKGROUND: Recent studies suggest that gene expression profiles are a promising alternative for clinical cancer classification. One major problem in applying DNA microarrays for classification is the dimension of obtained data sets. In this paper we propose a multiclass gene selection method based on Partial Least Squares (PLS) for selecting genes for classification. The new idea is to solve multiclass selection problem with the PLS method and decomposition to a set of two-class sub-problems: one versus rest (OvR) and one versus one (OvO). We use OvR and OvO two-class decomposition for other recently published gene selection method. Ranked gene lists are highly unstable in the sense that a small change of the data set often leads to big changes in the obtained ordered lists. In this paper, we take a look at the assessment of stability of the proposed methods. We use the linear support vector machines (SVM) technique in different variants: one versus one, one versus rest, multiclass SVM (MSVM) and the linear discriminant analysis (LDA) as a classifier. We use balanced bootstrap to estimate the prediction error and to test the variability of the obtained ordered lists. RESULTS: This paper focuses on effective identification of informative genes. As a result, a new strategy to find a small subset of significant genes is designed. Our results on real multiclass cancer data show that our method has a very high accuracy rate for different combinations of classification methods, giving concurrently very stable feature rankings. CONCLUSIONS: This paper shows that the proposed strategies can improve the performance of selected gene sets substantially. OvR and OvO techniques applied to existing gene selection methods improve results as well. The presented method allows to obtain a more reliable classifier with less classifier error. In the same time the method generates more stable ordered feature lists in comparison with existing methods.

A multi-gene approach to differentiate papillary thyroid carcinoma from benign lesions: gene selection using support vector machines with bootstrapping.

Selection of novel molecular markers is an important goal of cancer genomics studies. The aim of our analysis was to apply the multivariate bioinformatical tools to rank the genes - potential markers of papillary thyroid cancer (PTC) according to their diagnostic usefulness. We also assessed the accuracy of benign/malignant classification, based on gene expression profiling, for PTC. We analyzed a 180-array dataset (90 HG-U95A and 90 HG-U133A oligonucleotide arrays), which included a collection of 57 PTCs, 61 benign thyroid tumors, and 62 apparently normal tissues. Gene selection was carried out by the support vector machines method with bootstrapping, which allowed us 1) ranking the genes that were most important for classification quality and appeared most frequently in the classifiers (bootstrap-based feature ranking, BBFR); 2) ranking the samples, and thus detecting cases that were most difficult to classify (bootstrap-based outlier detection). The accuracy of PTC diagnosis was 98.5% for a 20-gene classifier, its 95% confidence interval (CI) was 95.9-100%, with the lower limit of CI exceeding 95% already for five genes. Only 5 of 180 samples (2.8%) were misclassified in more than 10% of bootstrap iterations. We specified 43 genes which are most suitable as molecular markers of PTC, among them some well-known PTC markers (MET, fibronectin 1, dipeptidylpeptidase 4, or adenosine A1 receptor) and potential new ones (UDP-galactose-4-epimerase, cadherin 16, gap junction protein 3, sushi, nidogen, and EGF-like domains 1, inhibitor of DNA binding 3, RUNX1, leiomodin 1, F-box protein 9, and tripartite motif-containing 58). The highest ranking gene, metallophosphoesterase domain-containing protein 2, achieved 96.7% of the maximum BBFR score.

Adjoint systems for models of cell signaling pathways and their application to parameter fitting.

The paper concerns the problem of fitting mathematical models of cell signaling pathways. Such models frequently take the form of sets of nonlinear ordinary differential equations. While the model is continuous in time, the performance index used in the fitting procedure, involves measurements taken at discrete time moments. Adjoint sensitivity analysis is a tool, which can be used for finding the gradient of a performance index in the space of parameters of the model. In the paper a structural formulation of adjoint sensitivity analysis called the Generalized Backpropagation Through Time (GBPTT) is used. The method is especially suited for hybrid, continuous-discrete time systems. As an example we use the mathematical model of the NF-kB regulatory module, which plays a major role in the innate immune response in animals.

[Gene expression profile of medullary thyroid carcinoma--preliminary results]. / Profil eksresji genów w raku rdzeniastym tarczycy--doniesienie wstepne.

INTRODUCTION: Medullary thyroid carcinoma occurs both as a sporadic and a familial disease. Inherited MTC (iMTC) patients usually exhibit better prognosis than patients with sporadic form of MTC (sMTC), however, in both subtypes the outcome is unpredictable. No molecular markers contributing to the prognosis or predicting the type of therapy have been introduced to clinical practice until now. The aim of this study was to analyze gene expression pattern of MTC by high density oligonucleotide microarray. MATERIAL AND METHODS: 24 samples were studied: 12 MTC and 12 corresponding normal tissues, (Affymetrix HG-U 133A). Among MTC patients there were half inherited cases and half sporadic ones. RESULTS: First, the differences between MTC and thyroid tissue were analyzed by Singular Value Decomposition (SVD) which indicated three main modes determining the variability of gene expression profile: the first two were related to the tumor/normal tissue difference and the third one was related to the immune response. The characteristic expression pattern, beside of numerous changes within cancer- related genes, included many up-regulated genes specific for thyroid C cells. Further analysis of the second component revealed two subgroups of MTC, but the subdivision was not related to the iMTC/sMTC difference. Recursive Feature Replacement (RFR) confirmed the very similar expression profile in both forms of MTC. With subsequent ANOVA analysis some genes with differential expression could be specified, among them monoamine oxidase B (MAOB) and gamma-aminobutyric acid receptor (GABRR1) which were consistently up-regulated in sMTC. In contrary, some genes involved in regulation of cell proliferation: opioid growth factor receptor(OGFR) and synaptotagmin V (SYT 5) were up-regulated in iMTC. CONCLUSIONS: The obtained data indicate a very similar gene expression pattern in inherited and sporadic MTC. Minor differences in their molecular profile require further analysis.

Meta- and reanalysis of gene expression profiles of hot and cold thyroid nodules and papillary thyroid carcinoma for gene groups.

CONTEXT: There are an increasing number of studies analyzing gene expression profiles in various benign and malignant thyroid tumors. This creates the opportunity to validate results obtained from one microarray study with those from other data sets. This process requires rigorous methods for accurate comparison. OBJECTIVE: The ability to compare data sets derived from different Affymetrix GeneChip generations and the influence of intra- and interindividual comparisons of gene expression data were evaluated to build multigene classifiers of benign thyroid nodules to verify a previously proposed papillary thyroid carcinoma (PTC) classifier and to look for molecular pathways essential for PTC oncogenesis. METHODS: Gene expression profile data sets from autonomously functioning and cold thyroid nodules and from PTC were analyzed by support vector machines. GenMAPP analysis was used for PTC data analysis to examine the expression patterns of biologically relevant gene sets. RESULTS: Only intraindividual reference samples allowed the identification of subtle changes in the expression patterns of relevant signaling cascades, such as the MAPK pathway in PTC. Using an artificial intelligence approach, the autonomously functioning and cold thyroid nodule multigene classifiers were derived and evaluated by cross-comparisons. CONCLUSION: We recommend defining classifiers within one generation of gene chips and subsequently checking them across different array generations. Using this approach, we have demonstrated the specificity of a previously reported PTC classifier on an independent collection of benign tumors. Moreover, we propose multigene classifiers for different types of benign thyroid nodules.

Gene expression profile of papillary thyroid cancer: sources of variability and diagnostic implications.

The study looked for an optimal set of genes differentiating between papillary thyroid cancer (PTC) and normal thyroid tissue and assessed the sources of variability in gene expression profiles. The analysis was done by oligonucleotide microarrays (GeneChip HG-U133A) in 50 tissue samples taken intraoperatively from 33 patients (23 PTC patients and 10 patients with other thyroid disease). In the initial group of 16 PTC and 16 normal samples, we assessed the sources of variability in the gene expression profile by singular value decomposition which specified three major patterns of variability. The first and the most distinct mode grouped transcripts differentiating between tumor and normal tissues. Two consecutive modes contained a large proportion of immunity-related genes. To generate a multigene classifier for tumor-normal difference, we used support vector machines-based technique (recursive feature replacement). It included the following 19 genes: DPP4, GJB3, ST14, SERPINA1, LRP4, MET, EVA1, SPUVE, LGALS3, HBB, MKRN2, MRC2, IGSF1, KIAA0830, RXRG, P4HA2, CDH3, IL13RA1, and MTMR4, and correctly discriminated 17 of 18 additional PTC/normal thyroid samples and all 16 samples published in a previous microarray study. Selected novel genes (LRP4, EVA1, TMPRSS4, QPCT, and SLC34A2) were confirmed by Q-PCR. Our results prove that the gene expression signal of PTC is easily detectable even when cancer cells do not prevail over tumor stroma. We indicate and separate the confounding variability related to the immune response. Finally, we propose a potent molecular classifier able to discriminate between PTC and nonmalignant thyroid in more than 90% of investigated samples.