A Support Vector Machine Classification of Thyroid Bioptic Specimens Using MALDI-MSI Data

Applications of spatially resolved omics in the field of endocrine tumors

Endocrine tumors derive from endocrine cells with high heterogeneity in function, structure and embryology, and are characteristic of a marked diversity and tissue heterogeneity. There are still challenges in analyzing the molecular alternations within the heterogeneous microenvironment for endocrine tumors. Recently, several proteomic, lipidomic and metabolomic platforms have been applied to the analysis of endocrine tumors to explore the cellular and molecular mechanisms of tumor genesis, progression and metastasis. In this review, we provide a comprehensive overview of spatially resolved proteomics, lipidomics and metabolomics guided by mass spectrometry imaging and spatially resolved microproteomics directed by microextraction and tandem mass spectrometry. In this regard, we will discuss different mass spectrometry imaging techniques, including secondary ion mass spectrometry, matrix-assisted laser desorption/ionization and desorption electrospray ionization. Additionally, we will highlight microextraction approaches such as laser capture microdissection and liquid microjunction extraction. With these methods, proteins can be extracted precisely from specific regions of the endocrine tumor. Finally, we compare applications of proteomic, lipidomic and metabolomic platforms in the field of endocrine tumors and outline their potentials in elucidating cellular and molecular processes involved in endocrine tumors.

Rapid diagnosis of papillary thyroid carcinoma with machine learning and probe electrospray ionization mass spectrometry

Frozen section examination could provide pathological diagnosis for surgery of thyroid nodules, which is time-consuming, skill- and experience-dependent. This study developed a rapid classification method for thyroid nodules and machine learning. Total 69 tissues were collected including 43 nodules and 26 nodule-adjacent tissues. Intraoperative frozen section was first performed to give accurate diagnosis, and the rest frozen specimen were pretreated for probe electrospray ionization mass measurement. By multivariate analysis of mass scan data, a series compounds were found downregulated in the extraction solution of papillary thyroid carcinoma (PTC), but some were found upregulated by mass spectrometry imaging. m/z 758.5713 ([PC[34:2] + H]⁺ ), m/z 772.5845 ([PC[32:0] + K]⁺ ), and m/z 786.6037 ([PC[36:2] + H]⁺ ) were firstly identified as potential biomarkers for nodular goiter (NG). Machine learning was employed by means of support vector machine (SVM) and random forest (RF) algorithms. For classification of PTC from NG, SVM and RF algorithms exhibited the same performance and the concordance was 94.2% and 94.4% between prediction and pathological diagnosis with positive and negative mass dataset, respectively. For the classification of PTC from PTC adjacent tissues, SVM was better than RF and the concordance was 93.8% and 83.3% with positive and negative mass dataset, respectively. With the identified compounds as training features, the sensitivity and specificity are 87.5% and 88.9% for the test set. The developed method could also correctly predict the malignancy of one medullary thyroid carcinoma and one adenomatous goiter (benign). The diagnosis time is about 10 min for one specimen, and it is very promising for the intraoperative diagnosis of papillary thyroid carcinoma.

Fast Classification of Thyroid Nodules with Ultrasound Guided-Fine Needle Biopsy Samples and Machine Learning

A rapid classification method was developed for the malignant and benign thyroid nodules with ultrasound guided-fine needle aspiration biopsy (FNAB) samples. With probe electrospray ionization mass spectrometry, the mass-scan data of FNAB samples were used as datasets for machine learning. The patients were marked as malignant (98 patients), benign (110 patients) or undetermined (42 patients) by experienced doctors in terms of ultrasound, the B-Raf (BRAF) gene, and cytopathology inspections. Pairwise coupling was performed on 163 ions to generate 3630 ion ratios as new features for classifier training. With the new features, the performance of the multilayer perception (MLP) classifier is much better than that with the 163 ions as features directly. After training, the accuracy of the MLP classifier is as high as 92.0%. The accuracy of the single-blind test is 82.4%, which proved the good generalization ability of the MLP classifier. The overall concordance is 73.0% between prediction and six-month follow-up for patients in the undetermined group. Especially, the classifier showed high accuracy for the undetermined patients with suspicious for papillary carcinoma diagnosis (90.9%). In summary, the machine learning method based on FNAB samples has potential for real clinical applications.

Automated biomarker candidate discovery in imaging mass spectrometry data through spatially localized Shapley additive explanations

The search for molecular species that are differentially expressed between biological states is an important step towards discovering promising biomarker candidates. In imaging mass spectrometry (IMS), performing this search manually is often impractical due to the large size and high-dimensionality of IMS datasets. Instead, we propose an interpretable machine learning workflow that automatically identifies biomarker candidates by their mass-to-charge ratios, and that quantitatively estimates their relevance to recognizing a given biological class using Shapley additive explanations (SHAP). The task of biomarker candidate discovery is translated into a feature ranking problem: given a classification model that assigns pixels to different biological classes on the basis of their mass spectra, the molecular species that the model uses as features are ranked in descending order of relative predictive importance such that the top-ranking features have a higher likelihood of being useful biomarkers. Besides providing the user with an experiment-wide measure of a molecular species' biomarker potential, our workflow delivers spatially localized explanations of the classification model's decision-making process in the form of a novel representation called SHAP maps. SHAP maps deliver insight into the spatial specificity of biomarker candidates by highlighting in which regions of the tissue sample each feature provides discriminative information and in which regions it does not. SHAP maps also enable one to determine whether the relationship between a biomarker candidate and a biological state of interest is correlative or anticorrelative. Our automated approach to estimating a molecular species' potential for characterizing a user-provided biological class, combined with the untargeted and multiplexed nature of IMS, allows for the rapid screening of thousands of molecular species and the obtention of a broader biomarker candidate shortlist than would be possible through targeted manual assessment. Our biomarker candidate discovery workflow is demonstrated on mouse-pup and rat kidney case studies.

Implications of Peak Selection in the Interpretation of Unsupervised Mass Spectrometry Imaging Data Analyses

Mass spectrometry imaging can produce large amounts of complex spectral and spatial data. Such data sets are often analyzed with unsupervised machine learning approaches, which aim at reducing their complexity and facilitating their interpretation. However, choices made during data processing can impact the overall interpretation of these analyses. This work investigates the impact of the choices made at the peak selection step, which often occurs early in the data processing pipeline. The discussion is done in terms of visualization and interpretation of the results of two commonly used unsupervised approaches: t-distributed stochastic neighbor embedding and k-means clustering, which differ in nature and complexity. Criteria considered for peak selection include those based on hypotheses (exemplified herein in the analysis of metabolic alterations in genetically engineered mouse models of human colorectal cancer), particular molecular classes, and ion intensity. The results suggest that the choices made at the peak selection step have a significant impact in the visual interpretation of the results of either dimensionality reduction or clustering techniques and consequently in any downstream analysis that relies on these. Of particular significance, the results of this work show that while using the most abundant ions can result in interesting structure-related segmentation patterns that correlate well with histological features, using a smaller number of ions specifically selected based on prior knowledge about the biochemistry of the tissues under investigation can result in an easier-to-interpret, potentially more valuable, hypothesis-confirming result. Findings presented will help researchers understand and better utilize unsupervised machine learning approaches to mine high-dimensionality data.

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.

Analysis of Hashimoto's thyroiditis on fine needle aspiration samples by MALDI-Imaging

Matrix-Assisted Laser Desorption/Ionization (MALDI)-Mass Spectrometry imaging (MSI) has been applied in various diseases aimed to biomarkers discovery. In this study diagnosis and prognosis of Hashimoto Thyroiditis (HT) in cytopathology by MALDI-MSI has been investigated. Specimens from a routine series of subjects who underwent UltraSound-guided thyroid Fine Needle Aspirations (FNAs) were used. The molecular classifier trained in a previous study was modified to include HT as a separate entity in the group of benign lesions, in the diagnostic proteomic triage of thyroid nodules. The statistical analysis confirmed the existence of signals that HT shares with hyperplastic lesions and others that are specific and characterize this subgroup. Statistically relevant HT-related peaks were included in the model. Then, the discriminatory capability of the classifier was tested in a second validation phase, showing a good agreement with cytological diagnoses. The possibility to overlap the molecular signatures of both the lymphocytes and epithelial cells components (ROIs or pixel-by-pixel analysis) confirmed the composite proteomic background of HT. These results open the way to their possible translation as alternative serum biomarkers of this autoimmune condition.

MALDI-MSI as a Complementary Diagnostic Tool in Cytopathology: A Pilot Study for the Characterization of Thyroid Nodules

The present study applies for the first time as Matrix-Assisted Laser Desorption/Ionization (MALDI) Mass Spectrometry Imaging (MSI) on real thyroid Fine Needle Aspirations (FNAs) to test its possible complementary role in routine cytology in the diagnosis of thyroid nodules. The primary aim is to evaluate the potential employment of MALDI-MSI in cytopathology, using challenging samples such as needle washes. Firstly, we designed a statistical model based on the analysis of Regions of Interest (ROIs), according to the morphological triage performed by the pathologist. Successively, the capability of the model to predict the classification of the FNAs was validated in a different group of patients on ROI and pixel-by-pixel approach. Results are very promising and highlight the possibility to introduce MALDI-MSI as a complementary tool for the diagnostic characterization of thyroid nodules.

Update on: proteome analysis in thyroid pathology - part II: overview of technical and clinical enhancement of proteomic investigation of the thyroid lesions

INTRODUCTION

An accurate diagnostic classification of thyroid lesions remains an important clinical aspect that needs to be addressed in order to avoid 'diagnostic' thyroidectomies. Among the several 'omics' techniques, proteomics is playing a pivotal role in the search for diagnostic markers. In recent years, different approaches have been used, taking advantage of the technical improvements related to mass spectrometry that have occurred. Areas covered: The review provides an update of the recent findings in diagnostic classification, in genetic definition and in the investigation of thyroid lesions based on different proteomics approaches and on different type of specimens: cytological, surgical and biofluid samples. A brief section will discuss how these findings can be integrated with those obtained by metabolomics investigations. Expert commentary: Among the several proteomics approaches able to deepen our knowledge of the molecular alterations of the different thyroid lesions, MALDI-MSI is strongly emerging above all. In fact, MS-imaging has also been demonstrated to be capable of distinguishing thyroid lesions, based on their different molecular signatures, using cytological specimens. The possibility to use the material obtained by the fine needle aspiration makes MALDI-MSI a highly promising technology that could be implemented into the clinical and pathological units.

Automated Morphological and Morphometric Analysis of Mass Spectrometry Imaging Data: Application to Biomarker Discovery

Mass spectrometry imaging datasets are mostly analyzed in terms of average intensity in regions of interest. However, biological tissues have different morphologies with several sizes, shapes, and structures. The important biological information, contained in this highly heterogeneous cellular organization, could be hidden by analyzing the average intensities. Finding an analytical process of morphology would help to find such information, describe tissue model, and support identification of biomarkers. This study describes an informatics approach for the extraction and identification of mass spectrometry image features and its application to sample analysis and modeling. For the proof of concept, two different tissue types (healthy kidney and CT-26 xenograft tumor tissues) were imaged and analyzed. A mouse kidney model and tumor model were generated using morphometric - number of objects and total surface - information. The morphometric information was used to identify m/z that have a heterogeneous distribution. It seems to be a worthwhile pursuit as clonal heterogeneity in a tumor is of clinical relevance. This study provides a new approach to find biomarker or support tissue classification with more information. Graphical Abstract ᅟ.

Using k-dependence causal forest to mine the most significant dependency relationships among clinical variables for thyroid disease diagnosis

Numerous data mining models have been proposed to construct computer-aided medical expert systems. Bayesian network classifiers (BNCs) are more distinct and understandable than other models. To graphically describe the dependency relationships among clinical variables for thyroid disease diagnosis and ensure the rationality of the diagnosis results, the proposed k-dependence causal forest (KCF) model generates a series of submodels in the framework of maximum spanning tree (MST) and demonstrates stronger dependence representation. Friedman test on 12 UCI datasets shows that KCF has classification accuracy advantage over the other state-of-the-art BNCs, such as Naive Bayes, tree augmented Naive Bayes, and k-dependence Bayesian classifier. Our extensive experimental comparison on 4 medical datasets also proves the feasibility and effectiveness of KCF in terms of sensitivity and specificity.