Shrunken Centroid Research Articles

Enhanced Neonatal Brain Tissue Analysis via Minimum Spanning Tree Segmentation and the Brier Score Coupled Classifier

Automatic assessment of brain regions in an MR image has emerged as a pivotal tool in advancing diagnosis and continual monitoring of neurological disorders through different phases of life. Nevertheless, current solutions often exhibit specificity to particular age groups, thereby constraining their utility in observing brain development from infancy to late adulthood. In our research, we introduce a novel approach for segmenting and classifying neonatal brain images. Our methodology capitalizes on minimum spanning tree (MST) segmentation employing the Manhattan distance, complemented by a shrunken centroid classifier empowered by the Brier score. This fusion enhances the accuracy of tissue classification, effectively addressing the complexities inherent in age-specific segmentation. Moreover, we propose a novel threshold estimation method utilizing the Brier score, further refining the classification process. The proposed approach yields a competitive Dice similarity index of 0.88 and a Jaccard index of 0.95. This approach marks a significant step toward neonatal brain tissue segmentation, showcasing the efficacy of our proposed methodology in comparison to the latest cutting-edge methods.

Decorrelated nearest shrunken centroids for tensor data

AbstractThe nearest shrunken centroids (NSC) method is an efficient and accurate classifier. However, it is incapable of modelling correlation among predictors. Moreover, many contemporary datasets have tensor predictors that cannot be directly handled by NSC. We tackle these challenges by proposing a new distance‐based classifier, tensor decorrelated NSC (TDNSC). TDNSC leverages the popular separable covariance structure on tensor data to decorrelate data and allow easy application of NSC afterwards. Unlike existing tensor classifiers that often rely on complicated iterative algorithms, TDNSC has analytical solutions. The theoretical properties and empirical results suggest that TDNSC is a promising method for tensor classification.

Identification and validation of a novel HOX-related classifier signature for predicting prognosis and immune microenvironment in pediatric gliomas.

Background: Pediatric gliomas (PGs) are highly aggressive and predominantly occur in young children. In pediatric gliomas, abnormal expression of Homeobox (HOX) family genes (HFGs) has been observed and is associated with the development and progression of the disease. Studies have found that overexpression or underexpression of certain HOX genes is linked to the occurrence and prognosis of gliomas. This aberrant expression may contribute to the dysregulation of important pathological processes such as cell proliferation, differentiation, and metastasis. This study aimed to propose a novel HOX-related signature to predict patients' prognosis and immune infiltrate characteristics in PGs. Methods: The data of PGs obtained from publicly available databases were utilized to reveal the relationship among abnormal expression of HOX family genes (HFGs), prognosis, tumor immune infiltration, clinical features, and genomic features in PGs. The HFGs were utilized to identify heterogeneous subtypes using consensus clustering. Then random forest-supervised classification algorithm and nearest shrunken centroid algorithm were performed to develop a prognostic signature in the training set. Finally, the signature was validated in an internal testing set and an external independent cohort. Results: Firstly, we identified HFGs significantly differentially expressed in PGs compared to normal tissues. The individuals with PGs were then divided into two heterogeneous subtypes (HOX-SI and HOX-SII) based on HFGs expression profiles. HOX-SII showed higher total mutation counts, lower immune infiltration, and worse prognosis than HOX-SI. Then, we constructed a HOX-related gene signature (including HOXA6, HOXC4, HOXC5, HOXC6, and HOXA-AS3) based on the cluster for subtype prediction utilizing random forest supervised classification and nearest shrunken centroid algorithm. The signature was revealed to be an independent prognostic factor for patients with PGs by multivariable Cox regression analysis. Conclusion: Our study provides a novel method for the prognosis classification of PGs. The findings also suggest that the HOX-related signature is a new biomarker for the diagnosis and prognosis of patients with PGs, allowing for more accurate survival prediction.

Discovering common pathogenic processes between COVID-19 and HFRS by integrating RNA-seq differential expression analysis with machine learning.

Zoonotic virus spillover in human hosts including outbreaks of Hantavirus and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) imposes a serious impact on the quality of life of patients. Recent studies provide a shred of evidence that patients with Hantavirus-caused hemorrhagic fever with renal syndrome (HFRS) are at risk of contracting SARS-CoV-2. Both RNA viruses shared a higher degree of clinical features similarity including dry cough, high fever, shortness of breath, and certain reported cases with multiple organ failure. However, there is currently no validated treatment option to tackle this global concern. This study is attributed to the identification of common genes and perturbed pathways by combining differential expression analysis with bioinformatics and machine learning approaches. Initially, the transcriptomic data of hantavirus-infected peripheral blood mononuclear cells (PBMCs) and SARS-CoV-2 infected PBMCs were analyzed through differential gene expression analysis for identification of common differentially expressed genes (DEGs). The functional annotation by enrichment analysis of common genes demonstrated immune and inflammatory response biological processes enriched by DEGs. The protein-protein interaction (PPI) network of DEGs was then constructed and six genes named RAD51, ALDH1A1, UBA52, CUL3, GADD45B, and CDKN1A were identified as the commonly dysregulated hub genes among HFRS and COVID-19. Later, the classification performance of these hub genes were evaluated using Random Forest (RF), Poisson Linear Discriminant Analysis (PLDA), Voom-based Nearest Shrunken Centroids (voomNSC), and Support Vector Machine (SVM) classifiers which demonstrated accuracy >70%, suggesting the biomarker potential of the hub genes. To our knowledge, this is the first study that unveiled biological processes and pathways commonly dysregulated in HFRS and COVID-19, which could be in the next future used for the design of personalized treatment to prevent the linked attacks of COVID-19 and HFRS.

Dynamic Network Construction for Identifying Early Warning Signals Based On a Data-Driven Approach: Early Diagnosis Biomarker Discovery for Gastric Cancer.

During the development of complex diseases, there is a critical transition from one status to another at a tipping point, which can be an early indicator of disease deterioration. To effectively enhance the performance of early risk identification, a novel dynamic network construction algorithm for identifying early warning signals based on a data-driven approach (EWS-DDA) was proposed. In EWS-DDA, the shrunken centroid was introduced to measure dynamic expression changes in assumed pathway reactions during the progression of complex disease for network construction and to define early warning signals by means of a data-driven approach. We applied EWS-DDA to perform a comprehensive analysis of gene expression profiles of gastric cancer (GC) from The Cancer Genome Atlas database and the Gene Expression Omnibus database. Six crucial genes were selected as potential biomarkers for the early diagnosis of GC. The experimental results of statistical analysis and biological analysis suggested that the six genes play important roles in GC occurrence and development. Then, EWS-DDA was compared with other state-of-the-art network methods to validate its performance. The theoretical analysis and comparison results suggested that EWS-DDA has great potential for a more complete presentation of disease deterioration and effective extraction of early warning information.

Machine learning approaches to the human metabolome in sepsis identify metabolic links with survival

BackgroundMetabolic predictors and potential mediators of survival in sepsis have been incompletely characterized. We examined whether machine learning (ML) tools applied to the human plasma metabolome could consistently identify and prioritize metabolites implicated in sepsis survivorship, and whether these methods improved upon conventional statistical approaches.MethodsPlasma gas chromatography–liquid chromatography mass spectrometry quantified 411 metabolites measured ≤ 72 h of ICU admission in 60 patients with sepsis at a single center (Brigham and Women’s Hospital, Boston, USA). Seven ML approaches were trained to differentiate survivors from non-survivors. Model performance predicting 28 day mortality was assessed through internal cross-validation, and innate top-feature (metabolite) selection and rankings were compared across the 7 ML approaches and with conventional statistical methods (logistic regression). Metabolites were consensus ranked by a summary, ensemble ML ranking procedure weighing their contribution to mortality risk prediction across multiple ML models.ResultsMedian (IQR) patient age was 58 (47, 62) years, 45% were women, and median (IQR) SOFA score was 9 (6, 12). Mortality at 28 days was 42%. The models’ specificity ranged from 0.619 to 0.821. Partial least squares regression-discriminant analysis and nearest shrunken centroids prioritized the greatest number of metabolites identified by at least one other method. Penalized logistic regression demonstrated top-feature results that were consistent with many ML methods. Across the plasma metabolome, the 13 metabolites with the strongest linkage to mortality defined through an ensemble ML importance score included lactate, bilirubin, kynurenine, glycochenodeoxycholate, phenylalanine, and others. Four of these top 13 metabolites (3-hydroxyisobutyrate, indoleacetate, fucose, and glycolithocholate sulfate) have not been previously associated with sepsis survival. Many of the prioritized metabolites are constituents of the tryptophan, pyruvate, phenylalanine, pentose phosphate, and bile acid pathways.ConclusionsWe identified metabolites linked with sepsis survival, some confirming prior observations, and others representing new associations. The application of ensemble ML feature-ranking tools to metabolomic data may represent a promising statistical platform to support biologic target discovery.

A genomic breed assignment test for traceability of meat of Dual-Purpose Blue

Assigning meat to its breed of origin for traceability purposes is not always straightforward if the breed from which products are derived is closely related to another one. The objective of this study was to determine if a genomic breed assignment test could distinguish meat of Dual-Purpose Blue, a local endangered breed, from meat of Beef Belgian Blue, a heavily used breed in the Belgian meat industry which is related to Dual-Purpose Blue. For this purpose, a genomic breed assignment test based on a panel of 2,005 SNPs and the nearest shrunken centroids method was used to classify 32 meat samples from Dual-Purpose Blue (n = 16), Beef Belgian Blue (n = 8) and Holstein (n = 8) into their breed of origin. From this SNP panel, 167 SNPs allowed to detect meat of Dual-Purpose Blue and 173 SNPs allowed to detect meat of Beef Belgian Blue. The genomic breed assignment test correctly allocated all the meat samples to their breed of origin with a probability of one. Therefore, the use of the genomic breed assignment test in routine as one step of the certification process of Dual-Purpose Blue meat seemed possible.

A Dual Level Analysis with Evolutionary Computing and Swarm Models for Classification of Leukemia.

One of the major reasons of mortality in human beings is cancer, and there is an absolute necessity for doctors to identify and treat a person suffering from it. Leukemia is a group of blood cancers that usually originates in the bone marrow and results in very high number of abnormal cells. For the diagnosis of cancer, microarray data serves as an important clinical application and serves as a great aid to the entire medical community. The dimensionality of the microarray data is too high, and so selection of suitable genes is quite an important step for the improvement of data classification. Therefore, for the prediction and diagnosis of cancer, there is an utmost necessity to select the most informative genes. In this work, Minimum Redundancy Maximum Relevance (MRMR), Signal to Noise Ratio (SNR), Multivariate Error Weight Uncorrelated Shrunken Centroid (EWUSC), and multivariate correlation-based feature selection (CFS) are chosen as initial feature selection techniques. Then, to select the most informative genes, five different kinds of evolutionary optimization techniques too are incorporated here such as African Buffalo Optimization (ABO), Artificial Bee Colony Optimization (ABCO), Cockroach Swarm Optimization (CSO), Imperialist Competitive Optimization (ICO), and Social Spider Optimization (SSO). Finally, the optimized values are fed through classification process and the best results are obtained when multivariate CFS with SSO is utilized and classified with Probabilistic Neural Network (PNN), and a high classification accuracy of 95.70% is obtained.

Exploration of the pathogenesis of Sjögren's syndrome via DNA methylation and transcriptome analyses.

Sjögren's syndrome (SS), a systemic autoimmune disorder, is characterized by dry mouth and eyes. However, SS pathogenesis is poorly understood. We performed bioinformatics analysis to investigate the potential targets and molecular pathogenesis of SS. Gene expression profiles (GSE157159) and methylation data (GSE110007) associated with SS patients were obtained from the Gene Expression Omnibus (GEO) database. Differentially methylated positions (DMPs) and differentially expressed genes (DEGs) were identified by the R package limma. The potential biological functions of DEGs were determined using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. Key DMPs were selected by overlap and the shrunken centroid algorithm, and corresponding genes were identified as hub genes, with their diagnostic value assessed by receiver operating characteristic (ROC) curves. The potential molecular mechanisms of hub genes were analyzed by protein-protein interaction (PPI) networks and single-gene gene set enrichment analysis (GSEA). Peripheral blood mononuclear cells (PBMCs) were collected from control and SS patients atThe Affiliated Hospital of Southwest Medical Universityand Dazhou Central Hospital. The mRNA levels of hub genes were verified by quantitative real-time polymerase chain reaction (qRT-PCR). We identified 788 DMPs and 2457 DEGs between the two groups. Functional enrichment analysis suggested that the DEGs were significantly enriched in T cell activation, leukocyte cell-cell adhesion, and cytokine-cytokine receptor interaction. TSS200, TSS1500, and 1stExon were identified as highly enriched areas of differentially methylated promoter CpG islands (DMCIs). In total, 61 differentially methylated genes (DMGs) were identified by the overlap of 2457 DEGs and 507 genes related to DMPs (DMPGs), of which 21 genes located near TSS200, TSS1500, and 1stExon were selected. Then, three key DMPs and the corresponding hub genes (RUNX3, HLA-DPA1, and CD6) were screened by the shrunken centroid algorithm and calculated to have areas under the ROC curve of 1.000, 0.931, and 0.986, respectively, indicating good diagnostic value. The GSEA results suggested that all three hub genes were highly associated with the immune response. Finally, positive mRNA expression of the three hub genes in clinical SS samples was verified by qRT-PCR, consistent with the GSE157159 data. The identification of three hub genes provides novel insight into molecular mechanisms and therapeutic targets for SS. Key Points • Hub genes were screened by DNA methylation and transcriptome analyses. • The relative expression of hub genes in peripheral blood samples was verified by qRT-PCR. • HLA-DPA1 was correlated with the pathogenic mechanism of SS.

Coordinatewise Gaussianization: Theories and Applications

In statistical analysis, researchers often perform coordinatewise Gaussianization such that each variable is marginally normal. The normal score transformation is a method for coordinatewise Gaussianization and is widely used in statistics, econometrics, genetics and other areas. However, few studies exist on the theoretical properties of the normal score transformation, especially in high-dimensional problems where the dimension p diverges with the sample size n. In this article, we show that the normal score transformation uniformly converges to its population counterpart even when log p = o ( n / log n ) . Our result can justify the normal score transformation prior to any downstream statistical method to which the theoretical normal transformation is beneficial. The same results are established for the Winsorized normal transformation, another popular choice for coordinatewise Gaussianization. We demonstrate the benefits of coordinatewise Gaussianization by studying its applications to the Gaussian copula model, the nearest shrunken centroids classifier and distance correlation. The benefits are clearly shown in theory and supported by numerical studies. Moreover, we also point out scenarios where coordinatewise Gaussinization does not help and even causes damages. We offer a general recommendation on how to use coordinatewise Gaussianization in applications. Supplementary materials for this article are available online.

Different thresholding methods on Nearest Shrunken Centroid algorithm

This article considers the impact of different thresholding methods to the Nearest Shrunken Centroid algorithm, which is popularly referred as the Prediction Analysis of Microarrays (PAM) for high-dimensional classification. PAM uses soft thresholding to achieve high computational efficiency and high classification accuracy but in the price of retaining too many features. When applied to microarray human cancers, PAM selected 2611 features on average from 10 multi-class datasets. Such a large number of features make it difficult to perform follow up study. One reason behind this problem is the soft thresholding, which is known to produce biased parameter estimate in regression analysis. In this article, we extend the PAM algorithm with two other thresholding methods, hard and order thresholding, and a deep search algorithm to achieve better thresholding parameter estimate. The modified algorithms are extensively tested and compared to the original one based on real data and Monte Carlo studies. In general, the modification not only gave better cancer status prediction accuracy, but also resulted in more parsimonious models with significantly smaller number of features.

BRCA1 Variant Assessment Using a Simple Analytic Assay.

We previously developed a biological assay to accurately predict BRCA1 (BRCA1 DNA repair associated) mutation status, based on gene expression profiles of Epstein-Barr virus-transformed lymphoblastoid cell lines. The original work was done using whole genome expression microarrays, and nearest shrunken centroids analysis. While these approaches are appropriate for model building, they are difficult to implement clinically, where more targeted testing and analysis are required for time and cost savings. Here, we describe adaptation of the original predictor to use the NanoString nCounter platform for testing, with analysis based on the k-top scoring pairs (k-TSP) method. Assessing gene expression using the nCounter platform on a set of lymphoblastoid cell lines yielded 93.8% agreement with the microarray-derived data, and 87.5% overall correct classification of BRCA1 carriers and controls. Using the original gene expression microarray data used to develop our predictor with nearest shrunken centroids, we rebuilt a classifier based on the k-TSP method. This classifier relies on the relative expression of 10 pairs of genes, compared to the original 43 identified by nearest shrunken centroids (NSC), and was 96.2% concordant with the original training set prediction, with a 94.3% overall correct classification of BRCA1 carriers and controls. The k-TSP classifier was shown to accurately predict BRCA1 status using data generated on the nCounter platform and is feasible for initiating a clinical validation.

Unsupervised Cross-User Adaptation in Taste Sensation Recognition Based on Surface Electromyography

Human taste sensation can be qualitatively described with surface electromyography. However, the pattern recognition models trained on one subject (the source domain) do not generalize well on other subjects (the target domain). To improve the generalizability and transferability of taste sensation models developed with sEMG data, two methods were innovatively applied in this study: domain regularized component analysis (DRCA) and conformal prediction with shrunken centroids (CPSC). The effectiveness of these two methods was investigated independently in an unlabeled data augmentation process with the unlabeled data from the target domain, and the same crossuser adaptation pipeline was conducted on six subjects. The results show that DRCA improved the classification accuracy on six subjects (p < 0.05), compared with the baseline models trained only with the source domain data, while CPSC did not guarantee the accuracy improvement. Furthermore, the combination of DRCA and CPSC presented statistically significant improvement (p < 0.05) in classification accuracy on six subjects. The proposed strategy of combining DRCA and CPSC showed its effectiveness in addressing the cross-user data distribution drift in sEMGbased taste sensation recognition application. It also shows the potential for more cross-user adaptation applications.

CPSC: Conformal Prediction With Shrunken Centroids for Efficient Prediction Reliability Quantification and Data Augmentation, a Case in Alternative Herbal Medicine Classification With Electronic Nose

In machine learning applications, the reliability of predictions is significant for assisted decision and risk control. As an effective framework to quantify the prediction reliability, conformal prediction (CP) was commonly developed with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> -nearest neighbor, known as CPKNN. However, the conventional CPKNN suffers from high variance and bias, and long computational time as the feature dimensionality increases. To address these limitations, a new CP framework-CP with shrunken centroids (CPSC) is proposed. It regularizes the class centroids to attenuate the irrelevant features and shrink the sample space for predictions and reliability quantification. We have also compared the CPSC with CP based on SVM, LightGBM, RF on the herbal medicine dataset, with the electronic nose as a case and assessed them in two tasks: 1) off-line prediction: the training set was fixed, and the accuracy on the testing set was evaluated and 2) online prediction with data augmentation: CPKNN and CPSC filtered unlabeled data to augment the training data based on the prediction reliability, and the final accuracy of the testing set was compared. The result shows that CPSC significantly outperformed CPKNN in both two tasks: 1) CPSC reached a significantly higher accuracy with lower computation cost, and with the same credibility output, CPSC generally achieved higher accuracy and 2) the data augmentation process with CPSC robustly manifested a statistically significant improvement in prediction accuracy with different reliability thresholds, and the augmented data were more balanced in classes. This novel CPSC provides higher prediction accuracy and better reliability quantification, which can assist in reliable decision support.

The robust nearest shrunken centroids classifier for high-dimensional heavy-tailed data

The robust nearest shrunken centroids classifier for high-dimensional heavy-tailed data

Genes and regulatory mechanisms associated with experimentally-induced bovine respiratory disease identified using supervised machine learning methodology

Bovine respiratory disease (BRD) is a multifactorial disease involving complex host immune interactions shaped by pathogenic agents and environmental factors. Advancements in RNA sequencing and associated analytical methods are improving our understanding of host response related to BRD pathophysiology. Supervised machine learning (ML) approaches present one such method for analyzing new and previously published transcriptome data to identify novel disease-associated genes and mechanisms. Our objective was to apply ML models to lung and immunological tissue datasets acquired from previous clinical BRD experiments to identify genes that classify disease with high accuracy. Raw mRNA sequencing reads from 151 bovine datasets (n = 123 BRD, n = 28 control) were downloaded from NCBI-GEO. Quality filtered reads were assembled in a HISAT2/Stringtie2 pipeline. Raw gene counts for ML analysis were normalized, transformed, and analyzed with MLSeq, utilizing six ML models. Cross-validation parameters (fivefold, repeated 10 times) were applied to 70% of the compiled datasets for ML model training and parameter tuning; optimized ML models were tested with the remaining 30%. Downstream analysis of significant genes identified by the top ML models, based on classification accuracy for each etiological association, was performed within WebGestalt and Reactome (FDR ≤ 0.05). Nearest shrunken centroid and Poisson linear discriminant analysis with power transformation models identified 154 and 195 significant genes for IBR and BRSV, respectively; from these genes, the two ML models discriminated IBR and BRSV with 100% accuracy compared to sham controls. Significant genes classified by the top ML models in IBR (154) and BRSV (195), but not BVDV (74), were related to type I interferon production and IL-8 secretion, specifically in lymphoid tissue and not homogenized lung tissue. Genes identified in Mannheimia haemolytica infections (97) were involved in activating classical and alternative pathways of complement. Novel findings, including expression of genes related to reduced mitochondrial oxygenation and ATP synthesis in consolidated lung tissue, were discovered. Genes identified in each analysis represent distinct genomic events relevant to understanding and predicting clinical BRD. Our analysis demonstrates the utility of ML with published datasets for discovering functional information to support the prediction and understanding of clinical BRD.

Development of a genomic tool for breed assignment by comparison of different classification models: Application to three local cattle breeds.

Assignment of individual cattle to a specific breed can often not rely on pedigree information. This is especially the case for local breeds for which the development of genomic assignment tools is required to allow individuals of unknown origin to be included to their herd books. A breed assignment model can be based on two specific stages: (a) the selection of breed-informative markers and (b) the assignment of individuals to a breed with a classification method. However, the performance of combination of methods used in these two stages has been rarely studied until now. In this study, the combination of 16 different SNP panels with four classification methods was developed on 562 reference genotypes from 12 cattle breeds. Based on their performances, best models were validated on three local breeds of interest. In cross-validation, 14 models had a global cross-validation accuracy higher than 90%, with a maximum of 98.22%. In validation, best models used 7,153 or 2,005 SNPs, based on a partial least squares-discriminant analysis (PLS-DA) and assigned individuals to breeds based on nearest shrunken centroids. The average validation sensitivity of the first two best models for the three local breeds of interest were 98.33% and 97.5%. Moreover, results reported in this study suggest that further studies should consider the PLS-DA method when selecting breed-informative SNPs.

IGlioSub: an integrative transcriptomic and epigenomic classifier for glioblastoma molecular subtypes

BackgroundGlioblastoma (GBM) is the most aggressive and prevalent primary brain tumor, with a median survival of 15 months. Advancements in multi-omics profiling combined with computational algorithms have unraveled the existence of three GBM molecular subtypes (Classical, Mesenchymal, and Proneural) with clinical relevance. However, due to the costs of high-throughput profiling techniques, GBM molecular subtyping is not currently employed in clinical settings.MethodsUsing Random Forest and Nearest Shrunken Centroid algorithms, we constructed transcriptomic, epigenomic, and integrative GBM subtype-specific classifiers. We included gene expression and DNA methylation (DNAm) profiles from 304 GBM patients profiled in the Cancer Genome Atlas (TCGA), the Human Glioblastoma Cell Culture resource (HGCC), and other publicly available databases.ResultsThe integrative Glioblastoma Subtype (iGlioSub) classifier shows better performance (mean AUC = 95.9%) stratifying patients than gene expression (mean AUC = 91.9%) and DNAm-based classifiers (AUC = 93.6%). Also, to expand the understanding of the molecular differences between the GBM subtypes, this study shows that each subtype presents unique DNAm patterns and gene pathway activation.ConclusionsThe iGlioSub classifier provides the basis to design cost-effective strategies to stratify GBM patients in routine pathology laboratories for clinical trials, which will significantly accelerate the discovery of more efficient GBM subtype-specific treatment approaches.

Comparative Study of Several Machine Learning Algorithms for Classification of Unifloral Honeys.

Unifloral honeys are highly demanded by honey consumers, especially in Europe. To ensure that a honey belongs to a very appreciated botanical class, the classical methodology is palynological analysis to identify and count pollen grains. Highly trained personnel are needed to perform this task, which complicates the characterization of honey botanical origins. Organoleptic assessment of honey by expert personnel helps to confirm such classification. In this study, the ability of different machine learning (ML) algorithms to correctly classify seven types of Spanish honeys of single botanical origins (rosemary, citrus, lavender, sunflower, eucalyptus, heather and forest honeydew) was investigated comparatively. The botanical origin of the samples was ascertained by pollen analysis complemented with organoleptic assessment. Physicochemical parameters such as electrical conductivity, pH, water content, carbohydrates and color of unifloral honeys were used to build the dataset. The following ML algorithms were tested: penalized discriminant analysis (PDA), shrinkage discriminant analysis (SDA), high-dimensional discriminant analysis (HDDA), nearest shrunken centroids (PAM), partial least squares (PLS), C5.0 tree, extremely randomized trees (ET), weighted k-nearest neighbors (KKNN), artificial neural networks (ANN), random forest (RF), support vector machine (SVM) with linear and radial kernels and extreme gradient boosting trees (XGBoost). The ML models were optimized by repeated 10-fold cross-validation primarily on the basis of log loss or accuracy metrics, and their performance was compared on a test set in order to select the best predicting model. Built models using PDA produced the best results in terms of overall accuracy on the test set. ANN, ET, RF and XGBoost models also provided good results, while SVM proved to be the worst.

MicroRNA expression profiling with a droplet digital PCR assay enables molecular diagnosis and prognosis of cancers of unknown primary.

Metastasis is responsible for the majority of cancer‐related deaths. Particularly, challenging is the management of metastatic cancer of unknown primary site (CUP), whose tissue of origin (TOO) remains undetermined even after extensive investigations and whose therapy is rather unspecific and poorly effective. Molecular approaches to identify the most probable TOO of CUPs can overcome some of these issues. In this study, we applied a predetermined set of 89 microRNAs (miRNAs) to infer the TOO of 53 metastatic cancers of unknown or uncertain origin. The miRNA expression was assessed with droplet digital PCR in 159 samples, including primary tumors from 17 tumor classes (reference set) and metastases of known and unknown origin (test set). We combined two different statistical models for class prediction to obtain the most probable TOOs: the nearest shrunken centroids approach of Prediction Analysis of Microarrays (PAMR) and the least absolute shrinkage and selection operator (LASSO) models. The molecular test was successful for all formalin‐fixed paraffin‐embedded samples and provided a TOO identification within 1 week from the biopsy procedure. The most frequently predicted origins were gastrointestinal, pancreas, breast, lung, and bile duct. The assay was applied also to multiple metastases from the same CUP, collected from different metastatic sites: The predictions showed a strong agreement, intrinsically validating our assay. The final CUPs' TOO prediction was compared with the clinicopathological hypothesis of primary site. Moreover, a panel of 13 miRNAs proved to have prognostic value and be associated with overall survival in CUP patients. Our study demonstrated that miRNA expression profiling in CUP samples could be employed as diagnostic and prognostic test. Our molecular analysis can be performed on request, concomitantly with standard diagnostic workup and in association with genetic profiling, to offer valuable indications about the possible primary site, thereby supporting treatment decisions.