Diagnostic Classifiers Research Articles

#6891 MACHINE LEARNING NEPHROPATHOLOGY DIAGNOSTICS OF THROMBOTIC MICROANGIOPATHIES

Abstract Background and Aims Thrombotic microangiopathies, comprising atypical hemolytic uremic syndrome (aHUS) and other diseases, can present with a broad clinical and histopathological spectrum. On our way to an evidence-base for the nephropathological work-up of TMAs, we have chosen a machine-learning approach, thus eliminating suboptimal reproducibility of descriptors for individual lesions in the three decisive compartments artery, arteriole and glomerulus with human experts. Here, we present our results for an end-to-end diagnostic system. Method We collected 50 random biopsies with TMAs of various etiologies (including aHUS, hypertension-associated, systemic sclerosis, anti-phospholipid antibody syndrome and others) and 50 biopsies with Mimickers (differential diagnoses of TMA), including severe hypertensive nephropathy, necrotising arteritis/arteriolitis, cryoglobulinemic vasculitis from the three participating centers Cologne, Weill-Cornell Medical Center, and Turin. Whole slide images (WSIs) from all four nephropathology stainings HE, PAS, trichrome and Jones were included in this study. We developed an instance segmentation Mask-RCNN model with a Swin Transformer (t) backbone on tissue crops detected using a lightweight variant of the U-Net segmentation architecture. For the classification model we used our own MorphSet++ set transformer architecture to process batches of EfficientNetv2s-encoded tissue crops entered in three separate compartment channels. Batches were chosen with Monte Carlo sampling or using our own soft Markov Chain Monte Carlo (MCMC) approach. Results of the classification model are reported with 5-fold internal cross-validation. Results Segmentation performance measured as mIOU, mAP, mAR, mF1, mAS for artery reached 0.565, 0.739, 0.679, 0.704, 0.995, for arteriole 0.342, 0.531, 0.488, 0.490, 0.996, for glomeruli 0.818, 0.880, 0.919, 0.896, 0.993. Classification accuracy reached 90% with no false positives for TMA. Missed cases of TMA could be salvaged by an experienced nephropathologist on the display of decisive compartment crops, which were selected using model confidence averaged across each sampling iteration. Conclusion We have designed and trained architectures capable of segmenting decisive compartments and diagnosing TMAs on renal biopsy sections. This will enable automatic analysis of clinicopathological datasets with TMA in large cohorts. Our ultimate goal is to use large cohorts from collaborating institutions for weakly supervised, case-level-annotated training of diagnostic, prognostic and theranostic classifiers.

#6541 GLOMERULONEPHRITIS DIAGNOSIS BY MACHINE LEARNING ON PERIODIC ACID-SCHIFF (PAS) WHOLE SLIDE IMAGES

Abstract Background and Aims Machine learning (ML) holds great promise for improving diagnostics, prognostication and theranostics in nephropathology. So far, applications have not gone much further than segmentation of tissue compartments on whole slide images (WSIs) of paraffin sections. As a proof-of-concept study, we describe the development of a diagnostic classifier for glomerulephritis based on expert-annotated or automatically segmented glomerular transections from periodic-acid Schiff (PAS) paraffin sections only. Method A total of n = 350 biopsies from 5 institutions with 12 classes of glomerulonephritis IgA nephropathy (IgAN), membranous nephropathy (Membranous), anti-glomerular basement membrane antibody GN (ABMGN), infection-associated GN (IAGN), ANCA-associated GN (ANCA-GN), idiopathic membranoproliferative GN (MPGN), SLE GN class IV (SLE-GN-IV), cryglobulinemic GN (CryoGN), C3 GN (C3-GN), dense deposit disease (DDD), fibrillary GN (FibrillaryGN) and proliferative GN with monoclonal immunoglobulin deposits (PGNMID) were included in the study with their respective PAS sections. Glomerular transections were expert-annotated by a nephropathologist and automatically segmented with our own transformer-based segmentation model trained on 100 biopsies with thrombotic microangiopathies and a range of vascular, vasculitic and glomerular diseases closely resembling/mimicking thrombotic microangiopathies. For classification, we divided the cohort into 5 folds for internal cross-validation, performed sample size augmentation with various methods (including shifts in resolution/scale, AutoAugment and others) and trained our proprietary self-attention-based MILx architecture on an EfficientNet backbone with selection of glomerular crop batches by soft Markov chain Monte Carlo sampling in a semi-supervised fashion, with diagnostic class labels for each biopsy. We compared the performance of our proprietary architecture on both expert-annotated and automatically segmented glomerular crops with a recently published benchmark architecture (CLAM) for multiple-instance learning in histopathology. Results Automatic glomerular segmentation performance was excellent with mean AUC and sensitivity (mean average recall) over all classes at 0.904, with near perfect mean average specificity (0.994), as expected best for Membranous, worst for ABMGN. Classification performance of MILx with expert-annotated glomerular crops as inputs had a mean balanced accuracy of 0.84, with AUC metrics in descending order of 0.97 for Membranous, 0.89 for ABMGN, 0.88 for IgAN, 0.86 for Fibrillary, 0.83 for MPGN, 0.80 for ANCA-GN, 0.79 for DDD, 0.78 for PGNMID, 0.75 for IAGN, 0.73 for SLE-GN-IV and CryoGN, 0.67 for C3-GN. Performance with MILx was similar for automatically segmented glomerular crops as input. On this dataset, MILx outperformed CLAM with both entire WSIs as well as expert-annotated glomerular crops as inputs (mean balanced accuracy of 0.72) by a significant margin. Conclusion This proof-of-concept-study indicates that nephropathology-specific architectures like our MILx can be trained for complex tasks on relatively small biopsy cohorts. We should be able to deliver an end-to-end-pipeline for this diagnostic and other tasks based on training sets with case-labels provided by trusted institutions with only minimal expert labeling or annotation required. PAC and HQ contributed equally to this work.

Circulating cell-free messenger RNA secretome characterization of primary sclerosing cholangitis.

Primary sclerosing cholangitis (PSC) is a rare chronic cholestatic liver disease characterized by multifocal bile duct strictures. To date, underlying molecular mechanisms of PSC remain unclear, and therapeutic options are limited. We performed cell-free messenger RNA (cf-mRNA) sequencing to characterize the circulating transcriptome of PSC and noninvasively investigate potentially bioactive signals that are associated with PSC. Serum cf-mRNA profiles were compared among 50 individuals with PSC, 20 healthy controls, and 235 individuals with NAFLD. Tissue and cell type-of-origin genes that are dysregulated in subjects with PSC were evaluated. Subsequently, diagnostic classifiers were developed using PSC dysregulated cf-mRNA genes. Differential expression analysis of the cf-mRNA transcriptomes of PSC and healthy controls resulted in identification of 1407 dysregulated genes. Furthermore, differentially expressed genes between PSC and healthy controls or NAFLD shared common genes known to be involved in liver pathophysiology. In particular, genes from liver- and specific cell type-origin, including hepatocyte, HSCs, and KCs, were highly abundant in cf-mRNA of subjects with PSC. Gene cluster analysis revealed that liver-specific genes dysregulated in PSC form a distinct cluster, which corresponded to a subset of the PSC subject population. Finally, we developed a cf-mRNA diagnostic classifier using liver-specific genes that discriminated PSC from healthy control subjects using gene transcripts of liver origin. Blood-based whole-transcriptome cf-mRNA profiling revealed high abundance of liver-specific genes in sera of subjects with PSC, which may be used to diagnose patients with PSC. We identified several unique cf-mRNA profiles of subjects with PSC. These findings may also have utility for noninvasive molecular stratification of subjects with PSC for pharmacotherapy safety and response studies.

Molecular diagnostic classification for cancers of unknown primary (CUP): Post-testing diagnosis and treatment impact analysis from real-world claims data.

6610 Background: Next-generation sequencing-based molecular diagnostic classifiers can help identify the tissue of origin for CUP, and enable the selection of site-specific therapies (as opposed to empiric chemotherapy) for this vulnerable population with high unmet need. NCCN guidelines do not endorse the use of tissue of origin classifiers as standard of care for CUP patients due to limited clinical evidence. Here, we linked results of a commercial molecular diagnostic classifier with claims data to understand how this test impacted patient care. Methods: We assessed de-identified claims data from the Komodo Healthcare Map (a database including provider visits, laboratory tests, procedures, imaging, and prescriptions) linked to the Tempus Tumor Origin (TO) test—a machine learning classifier that uses RNA-seq data to classify tumors into one of 68 histological subtypes. Eligible patients had pathologist-confirmed CUP and were classified as one of 9 subtypes (each having n>10). Impact was determined by identifying either one or more new diagnostic codes or new subtype-related medication claims following TO testing. Results: We analyzed data from 490 patients: 483 for the diagnosis analysis and 213 for the medication analysis (206 patients appear in both, due to differences in timing of diagnosis vs. medication claims). We found that post-TO testing, 49.9% (n=241) of patients had a diagnostic code change, 63.8% (n=136) had a treatment change, and 41% (n=85) had both. In total, 59.6% (n=292) of patients were impacted by the use of this classifier, with variation according to predicted subtype (Table). Cancer types with specific treatment options were more likely to have changes in diagnostic code or treatment. For CUP predicted as lung adenocarcinomas, 85% of the cases had a new subtype-aligned medication, and 78% were specifically placed on immunotherapy (IO) compared with 24% of overall patients. Conclusions: Using real-world claims data, we show that molecular diagnostic testing utilizing Tempus TO impacted care for a majority of CUP patients. This cohort will be followed to identify how TO testing decisions impact outcomes. [Table: see text]

Label-free atherosclerosis diagnosis through a blood drop of apolipoprotein E knockout mouse model using surface-enhanced Raman spectroscopy validated by machine learning algorithm.

The direct preventative detection of flow-induced atherosclerosis remains a significant challenge, impeding the development of early treatments and prevention measures. This study proposes a method for diagnosing atherosclerosis in the carotid artery using nanometer biomarker measurements through surface-enhanced Raman spectroscopy (SERS) from single-drop blood samples. Atherosclerotic acceleration is induced in apolipoprotein E knockout mice which underwent a partial carotid ligation and were fed a high-fat diet to rapidly induce disturbed flow-induced atherosclerosis in the left common carotid artery while using the unligated, contralateral right carotid artery as control. The progressive atherosclerosis development of the left carotid artery was verified by micro-magnetic resonance imaging (micro-MRI) and histology in comparison to the right carotid artery. Single-drop blood samples are deposited on chips of gold-coated ZnO nanorods grown on silicon wafers that filter the nanometer markers and provide strong SERS signals. A diagnostic classifier was established based on principal component analysis (PCA), which separates the resultant spectra into the atherosclerotic and control groups. Scoring based on the principal components enabled the classification of samples into control, mild, and severe atherosclerotic disease. The PCA-based analysis was validated against an independent test sample and compared against the PCA-PLS-DA machine learning algorithm which is known for applicability to Raman diagnosis. The accuracy of the PCA modification-based diagnostic criteria was 94.5%, and that of the machine learning algorithm 97.5%. Using a mouse model, this study demonstrates that diagnosing and classifying the severity of atherosclerosis is possible using a single blood drop, SERS technology, and machine learning algorithm, indicating the detectability of biomarkers and vascular factors in the blood which correlate with the early stages of atherosclerosis development.

Diagnostic Classifiers for Explaining a Neural Model with Hierarchical Attention for Aspect-based Sentiment Classification

The current models proposed for aspect-based sentiment classification (ABSC) are mainly developed with the purpose of providing high rates of accuracy, regardless of the inner working which is usually difficult to understand. Considering the state-of-art model LCR-Rot-hop++ for ABSC, we use diagnostic classifiers to gain insights into the encoded information of each layer. Starting from a set of various hypotheses, we test how sentiment-related information is captured by different layers of the model. Given the model architecture, information about the related words to the target is easily extracted. Also, the model is able to detect to some extent information about the sentiments of the words and, in particular, sentiments of the words related to the target. However, the model is less effective in extracting the aspect mentions associated with a word and the general structure of the sentence.

A Diagnostic Classifier Based on Circulating miRNA Pairs for COPD Using a Machine Learning Approach

Chronic obstructive pulmonary disease (COPD) is highly underdiagnosed, and early detection is urgent to prevent advanced progression. Circulating microRNAs (miRNAs) have been diagnostic candidates for multiple diseases. However, their diagnostic value has not yet been fully established in COPD. The purpose of this study was to develop an effective model for the diagnosis of COPD based on circulating miRNAs. We included circulating miRNA expression profiles of two independent cohorts consisting of 63 COPD and 110 normal samples, and then we constructed a miRNA pair-based matrix. Diagnostic models were developed using several machine learning algorithms. The predictive performance of the optimal model was validated in our external cohort. In this study, the diagnostic values of miRNAs based on the expression levels were unsatisfactory. We identified five key miRNA pairs and further developed seven machine learning models. The classifier based on LightGBM was selected as the final model with the area under the curve (AUC) values of 0.883 and 0.794 in test and validation datasets, respectively. We also built a web tool to assist diagnosis for clinicians. Enriched signaling pathways indicated the potential biological functions of the model. Collectively, we developed a robust machine learning model based on circulating miRNAs for COPD screening.

CCS-GAN: COVID-19 CT Scan Generation and Classification with Very Few Positive Training Images.

We present a novel algorithm that is able to generate deep synthetic COVID-19 pneumonia CT scan slices using a very small sample of positive training images in tandem with a larger number of normal images. This generative algorithm produces images of sufficient accuracy to enable a DNN classifier to achieve high classification accuracy using as few as 10 positive training slices (from 10 positive cases), which to the best of our knowledge is one order of magnitude fewer than the next closest published work at the time of writing. Deep learning with extremely small positive training volumes is a very difficult problem and has been an important topic during the COVID-19 pandemic, because for quite some time it was difficult to obtain large volumes of COVID-19-positive images for training. Algorithms that can learn to screen for diseases using few examples are an important area of research. Furthermore, algorithms to produce deep synthetic images with smaller data volumes have the added benefit of reducing the barriers of data sharing between healthcare institutions. We present the cycle-consistent segmentation-generative adversarial network (CCS-GAN). CCS-GAN combines style transfer with pulmonary segmentation and relevant transfer learning from negative images in order to create a larger volume of synthetic positive images for the purposes of improving diagnostic classification performance. The performance of a VGG-19 classifier plus CCS-GAN was trained using a small sample of positive image slices ranging from at most 50 down to as few as 10 COVID-19-positive CT scan images. CCS-GAN achieves high accuracy with few positive images and thereby greatly reduces the barrier of acquiring large training volumes in order to train a diagnostic classifier for COVID-19.

Viewing both sides of the coin for infectious disease diagnosis.

Optimal management of lower respiratory tract infection relies on distinguishing infectious from noninfectious etiologies and identifying the microbiologic cause if applicable. This process is complicated by overlapping clinical symptoms and the colonizing lung microbiota. In a recent issue of the JCI, Mick, Tsitsiklis, and colleagues apply RNA-Seq to tracheal aspirates from critically ill children and demonstrate how integration of the host response with microbial identification results in a harmonious and accurate diagnostic classifier. Though promising, there are numerous barriers to realizing a combined host and pathogen diagnostic.

UCseek: ultrasensitive early detection and recurrence monitoring of urothelial carcinoma by shallow-depth genome-wide bisulfite sequencing of urinary sediment DNA.

Current methods for the detection and surveillance of urothelial carcinomas (UCs) are often invasive, costly, and not effective for low-grade, early-stage, and minimal residual disease (MRD) tumors. We aimed to develop and validate a model from urine sediments to predict different grade and stage UCs with low cost and high accuracy. We collected 167 samples, including 90 tumors and 77 individuals without tumors, as a discovery cohort. We assessed copy number variations and methylation values for them and constructed a diagnostic classifier to detect UC, UCseek, by using an individual read-based method and support vector machine. The performance of UCseek was validated in an independent cohort derived from three hospitals (n=206) and a relapse cohort (n=42) for monitoring recurrence. We constructed UCseek, which could predict UCs with high sensitivity (92.7%), high specificity (90.7%), and high accuracy (91.7%) in the independent validation set. The accuracy of UCseek in low-grade and early-stage patients reached 91.8% and 94.3%, respectively. Notably, UCseek retained great performance at ultralow sequencing depths (0.3X-0.5X). It also demonstrated a powerful ability to monitor recurrence in a surveillance cohort compared with cystoscopy (90.91% vs. 59.09%). We optimized an improved approach named UCseek for the noninvasive diagnosis and monitoring of UCs in both low- and high-grade tumors and in early- and advanced-stage tumors, even at ultralow sequencing depths, which may reduce the burden of cystoscopy and blind second surgery. A full list of funding bodies that contributed to this study can be found in the Acknowledgments section.

Level of Attention to Motherese Speech as an Early Marker of Autism Spectrum Disorder

Caregivers have long captured the attention of their infants by speaking in motherese, a playful speech style characterized by heightened affect. Reduced attention to motherese in toddlers with autism spectrum disorder (ASD) may be a contributor to downstream language and social challenges and could be diagnostically revealing. To investigate whether attention toward motherese speech can be used as a diagnostic classifier of ASD and is associated with language and social ability. This diagnostic study included toddlers aged 12 to 48 months, spanning ASD and non-ASD diagnostic groups, at a research center. Data were collected from February 2018 to April 2021 and analyzed from April 2021 to March 2022. Gaze-contingent eye-tracking test. Using gaze-contingent eye tracking wherein the location of a toddler's fixation triggered a specific movie file, toddlers participated in 1 or more 1-minute eye-tracking tests designed to quantify attention to motherese speech, including motherese vs traffic (ie, noisy vehicles on a highway) and motherese vs techno (ie, abstract shapes with music). Toddlers were also diagnostically and psychometrically evaluated by psychologists. Levels of fixation within motherese and nonmotherese movies and mean number of saccades per second were calculated. Receiver operating characteristic (ROC) curves were used to evaluate optimal fixation cutoff values and associated sensitivity, specificity, positive predictive value (PPV), and negative predictive value. Within the ASD group, toddlers were stratified based on low, middle, or high levels of interest in motherese speech, and associations with social and language abilities were examined. A total of 653 toddlers were included (mean [SD] age, 26.45 [8.37] months; 480 males [73.51%]). Unlike toddlers without ASD, who almost uniformly attended to motherese speech with a median level of 82.25% and 80.75% across the 2 tests, among toddlers with ASD, there was a wide range, spanning 0% to 100%. Both the traffic and techno paradigms were effective diagnostic classifiers, with large between-group effect sizes (eg, ASD vs typical development: Cohen d, 1.0 in the techno paradigm). Across both paradigms, a cutoff value of 30% or less fixation on motherese resulted in an area under the ROC curve (AUC) of 0.733 (95% CI, 0.693-0.773) and 0.761 (95% CI, 0.717-0.804), respectively; specificity of 98% (95% CI, 95%-99%) and 96% (95% CI, 92%-98%), respectively; and PPV of 94% (95% CI, 86%-98%). Reflective of heterogeneity and expected subtypes in ASD, sensitivity was lower at 18% (95% CI, 14%-22%) and 29% (95% CI, 24%-34%), respectively. Combining metrics increased the AUC to 0.841 (95% CI, 0.805-0.877). Toddlers with ASD who showed the lowest levels of attention to motherese speech had weaker social and language abilities. In this diagnostic study, a subset of toddlers showed low levels of attention toward motherese speech. When a cutoff level of 30% or less fixation on motherese speech was used, toddlers in this range were diagnostically classified as having ASD with high accuracy. Insight into which toddlers show unusually low levels of attention to motherese may be beneficial not only for early ASD diagnosis and prognosis but also as a possible therapeutic target.

P106 Identification of urine and serum diagnostic biomarkers of inflammatory bowel disease using a proteomic approach

Abstract Background Inflammatory bowel diseases (IBD) are chronic, heterogeneous, and inflammatory conditions mainly affecting the gastrointestinal tract. Currently, endoscopy is the gold standard test for assessing mucosal activity and healing in clinical practice; however, it is a costly, time-consuming, invasive, and uncomfortable procedure for patients. There is, therefore, a need for sensitive, specific, fast and non-invasive biomarkers for the diagnosis of IBD. Methods Label-free quantification by nanoscale liquid chromatography coupled to tandem mass spectrometry (nLC MS/MS) was performed to profile the urinary and serum proteomes of 100 patients with newly diagnosed IBD, before starting any treatment [50 patients with Crohn′s disease (CD) and 50 patients with ulcerative colitis (UC)] and 50 healthy controls (HC). Data Mining and Pattern Recognition techniques were used to identify hidden relationships that are not detectable by using the classical and most commonly used lineal classifiers; and thus, identifying potential markers able to discriminate the studied cohorts. Finally, we applied Ingenuity Pathway Analysis (IPA) to analyze the pathways and functions of differentially expressed proteins. Results Serum proteomics results revealed 45 differentially expressed proteins in the comparison between UC and HC groups, 32 proteins significantly expressed in CD versus HC, and 12 proteins in CD compared with UC. In urine samples, 110 proteins were significantly changed in UC versus HC, 50 proteins were differentially expressed between CD and HC, and a total of 31 proteins were significantly changed in CD compared with UC patients. Receiver operating characteristic curve (ROC) analysis found multiple proteins with high area under the curve values, up to 0.94, indicating that these serum and urine proteins are of value as new non-invasive diagnostic classifiers of IBD patients. For each study comparative, the 5 most significant classifiers were selected (Figures 1 and 2). IPA revealed multiple signaling pathways, including prothrombin activation, acute phase response signaling, complement and coagulation system and liver X receptor/retinoid X receptor (LXR/RXR) activation, altered in IBD patients compared to HC (p-value < 0.05) (Figures 3 and 4). Conclusion Our findings indicate that analysis of the urine and serum proteome using nLC MS/MS is a feasible approach for biomarker discovery. We identified several serum and urine proteins that could serve as new non-invasive markers for the diagnosis of IBD patients after further validation. After bioinformatic analyses, we found multiple proteins that may play important roles in the pathogenesis of IBD.

Machine Learning and MRI-based Diagnostic Models for ADHD: Are We There Yet?

Objective: Machine learning (ML) has been applied to develop magnetic resonance imaging (MRI)-based diagnostic classifiers for attention-deficit/hyperactivity disorder (ADHD). This systematic review examines this literature to clarify its clinical significance and to assess the implications of the various analytic methods applied. Methods: A comprehensive literature search on MRI-based diagnostic classifiers for ADHD was performed and data regarding the utilized models and samples were gathered. Results: We found that, although most studies reported the classification accuracies, they varied in choice of MRI modalities, ML models, cross-validation and testing methods, and sample sizes. We found that the accuracies of cross-validation methods inflated the performance estimation compared with those of a held-out test, compromising the model generalizability. Test accuracies have increased with publication year but were not associated with training sample sizes. Improved test accuracy over time was likely due to the use of better ML methods along with strategies to deal with data imbalances. Conclusion: Ultimately, large multi-modal imaging datasets, and potentially the combination with other types of data, like cognitive data and/or genetics, will be essential to achieve the goal of developing clinically useful imaging classification tools for ADHD in the future.

FTGAN: A Novel GAN-Based Data Augmentation Method Coupled Time–Frequency Domain for Imbalanced Bearing Fault Diagnosis

For imbalanced bearing fault diagnosis, generative adversarial networks (GANs) are a common data augmentation (DA) approach. Nevertheless, current GAN-based methods cannot update the generator from time–frequency domain simultaneously, downgrading the authenticity of signal time–frequency character. In this article, Fourier-like transform GAN (FTGAN), a novel GAN method, is proposed by introducing a Fourier-like transformer (FLT) based on autoencoder (AE) to improve synthetic data quality. FLT approximates the discrete Fourier transform (DFT) by the neural network, learning a universal map from time to frequency domain during training. FTGAN with FLT can decouple input into a time–frequency domain, fitting the distribution of time and frequency of data simultaneously. Multidomain distribution is manipulated in FTGAN without introducing additional signal transformation means. Furthermore, train on real, test on synthetic (TRTS) and train on synthetic, test on real (TSTR) analyses of 1-D data are introduced to evaluate data quality. Real and synthetic data are applied as training or test sets of diagnostic classifiers by turns so that data quality can be analyzed through diagnosis results. Experiment results show that the proposed method can generate bearing fault signals closer to real data in the time and frequency domains, effectively improving the performance under an imbalanced dataset.

Exploration of diagnostic biomarkers, microenvironment characteristics, and ursolic acid's therapeutic effect for benign prostate hyperplasia.

Benign prostatic hyperplasia (BPH) and early-stage prostate cancer (PC) have similar symptoms, making it challenging to differentially diagnose these two conditions. The study used Weighted Gene Co-Expression Network Analysis, as well as two machine learning strategies to identify BPH-specific biomarkers based on an integrated transcriptome data from 922 samples. Eight prognostic genes (ALCAM, COL6A2, CRISP2, FOXF2, IGF1, PTN, SCN7A, and UAP1) were identified to be BPH-specific biomarkers with high accuracy and specificity. Moreover, we constructed a seven-gene diagnostic classifier to distinguish BPH from PC. The infiltrations of plasmacytoid dendritic cells and neutrophil cells showed distinct differences between BPH and non-BPH groups. Additionally, ursolic acid can reverse transcriptional features associated with the occurrence and progression of BPH. Both in vivo and in vitro experiments have confirmed that it induces apoptosis of BPH cells and inhibits cell proliferation by promoting cell cycle S-phase arrest. The diagnostic biomarkers, microenvironment characteristics, and therapeutic effect of ursolic acid explored in this study offer new diagnostic and therapeutic strategies for BPH.

Grading Classification of Tongue Cancer in Oral Images Using CNN Transfer Learning

- Oral cancer is one of the top 10 diseases in cancer incidence worldwide and its prognosis is not good as early diagnosis is not made and so the survival rate from it has not been improved greatly. Oral cancer occurs most commonly in tongue, floor of mouth, and lower lip and 5-year survival rate is very low at 50% and if diagnosed early, the average time of survival becomes longer. Therefore, to improve the survival rate of oral cancer patients, early diagnosis and discovery and patient habit improvement including smoking are very important. Oral cancer can be diagnosed with imaging equipment like CT and MRI or through invasive biopsy, but it is not easy to make an approach and generally difficult to discover early. Therefore, to identify if early diagnosis would be possible using the intraoral images obtained through ordinary camera, the reliability of tongue cancer grade classification was evaluated by applying deep learning analysis. The collected images of oral cavity were classified into four groups: normal, inflammatory lesion, pre-cancerous lesion, and malignant tumor depending on clinical importance and three (3) types of image pretreatment were applied. In case of deep neural network, Inception-ResNet-v2 was used and 70%(1,307 images) of the pretreated images were used as training data and after image enhancement processing, transfer-learned with SGDM technique. As test data, the remaining 30% data (561 images) were used. In confusion matrix, it appeared that sensitivity was 90.37%, specificity 95.03%, accuracy 93.21%, positive predictability 92.06%, and negative predictability 93.93%, and in ROC curve, AUC was normal because normal was 0.9781, inflammatory lesion 0.9102, precancerous lesion 0.8944, and malignant tumor 0.9688, and the classification performance of inflammatory lesion or pre-cancerous lesion was inferior to that of malignant tumor. The accuracy depending on pretreatment did not show a statistically significant difference in HE-1 and HE-2 compared to standard image. The performance of the diagnostic classifier applying deep learning into intraoral images showed accuracy at the average rate of 82.9% and so showed more excellent performance than the precedent studies using different RGB images. Deep learning study using such RGB images tends to be rapidly commercialized as it has disclosed big data whose images were easy to obtain and that were preclassified and so noninvasive data can be easily obtained unlike biopsy or blood test. Intraoral image does not have preclassified and disclosed big data yet, but have a distortion in an artifact or lesion in obtaining an image and so if we can standardize the image acquisition and organize big data through correct preclassification in cooperation with clinicians although this is limited, it will be helpful for early diagnosis of tongue cancer.

Small RNA Sequencing Identifies a Six-MicroRNA Signature Enabling Classification of Brain Metastases According to their Origin.

Brain metastases (BMs) are the most frequent intracranial tumors in adults and one of the greatest challenges for modern oncology. Most are derived from lung, breast, renal cell, and colorectal carcinomas and melanomas. Up to 14% of patients are diagnosed with BMs of unknown primary, which are commonly characterized by an early and aggressive metastatic spread. It is important to discover novel biomarkers for early identification of BM origin, allowing better management of patients with this disease. Our study focused on microRNAs (miRNAs), which are very stable in frozen native and FFPE tissues and have been shown to be sensitive and specific diagnostic biomarkers of cancer. We aimed to identify miRNAs with significantly different expression in the five most frequent groups of BMs and develop a diagnostic classifier capable of sensitive and specific classification of BMs. Total RNA enriched for miRNAs was isolated using the mirVana miRNA Isolation Kit from 71 fresh-frozen histopathologically confirmed BM tissues originating in 5 cancer types. Sequencing libraries were prepared using the QIAseq miRNA Library Kit and sequenced on the NextSeq 500 platform. MiRNA expression was further validated by RT-qPCR. Differential analysis identified 373 miRNAs with significantly different expression between 5 BM groups (p<0.001). A classifier model was developed based on the expression of 6 miRNAs (hsa-miR-141-3p, hsa-miR-141-5p, hsa-miR-146a-5p, hsa-miR-194-5p, hsa-miR-200b-3p and hsa-miR-365b-5p) with the ability to correctly classify 91.5% of samples. Subsequent validation confirmed both significantly different expression of selected miRNAs in 5 BM groups as well as their diagnostic potential. To date, our study is the first to analyze miRNA expression in various types of BMs using small RNA sequencing to develop a diagnostic classifier and, thus, to help stratify BMs of unknown primary. The presented results confirm the importance of studying the dysregulated expression of miRNAs in BMs and the diagnostic potential of the validated 6-miRNA signature.

Identification of a possible proteomic biomarker in Parkinson's disease: discovery and replication in blood, brain and cerebrospinal fluid.

Biomarkers to aid diagnosis and delineate the progression of Parkinson's disease are vital for targeting treatment in the early phases of the disease. Here, we aim to discover a multi-protein panel representative of Parkinson's and make mechanistic inferences from protein expression profiles within the broader objective of finding novel biomarkers. We used aptamer-based technology (SomaLogic®) to measure proteins in 1599 serum samples, 85 cerebrospinal fluid samples and 37 brain tissue samples collected from two observational longitudinal cohorts (the Oxford Parkinson's Disease Centre and Tracking Parkinson's) and the Parkinson's Disease Brain Bank, respectively. Random forest machine learning was performed to discover new proteins related to disease status and generate multi-protein expression signatures with potential novel biomarkers. Differential regulation analysis and pathway analysis were performed to identify functional and mechanistic disease associations. The most consistent diagnostic classifier signature was tested across modalities [cerebrospinal fluid (area under curve) = 0.74, P = 0.0009; brain area under curve = 0.75, P = 0.006; serum area under curve = 0.66, P = 0.0002]. Focusing on serum samples and using only those with severe disease compared with controls increased the area under curve to 0.72 (P = 1.0 × 10-4). In the validation data set, we showed that the same classifiers were significantly related to disease status (P < 0.001). Differential expression analysis and weighted gene correlation network analysis highlighted key proteins and pathways with known relationships to Parkinson's. Proteins from the complement and coagulation cascades suggest a disease relationship to immune response. The combined analytical approaches in a relatively large number of samples, across tissue types, with replication and validation, provide mechanistic insights into the disease as well as nominate a protein signature classifier that deserves further biomarker evaluation.

A 2-Gene Host Signature for Improved Accuracy of COVID-19 Diagnosis Agnostic to Viral Variants.

The continued emergence of SARS-CoV-2 variants is one of several factors that may cause false-negative viral PCR test results. Such tests are also susceptible to false-positive results due to trace contamination from high viral titer samples. Host immune response markers provide an orthogonal indication of infection that can mitigate these concerns when combined with direct viral detection. Here, we leverage nasopharyngeal swab RNA-seq data from patients with COVID-19, other viral acute respiratory illnesses, and nonviral conditions (n = 318) to develop support vector machine classifiers that rely on a parsimonious 2-gene host signature to diagnose COVID-19. We find that optimal classifiers include an interferon-stimulated gene that is strongly induced in COVID-19 compared with nonviral conditions, such as IFI6, and a second immune-response gene that is more strongly induced in other viral infections, such as GBP5. The IFI6+GBP5 classifier achieves an area under the receiver operating characteristic curve (AUC) greater than 0.9 when evaluated on an independent RNA-seq cohort (n = 553). We further provide proof-of-concept demonstration that the classifier can be implemented in a clinically relevant RT-qPCR assay. Finally, we show that its performance is robust across common SARS-CoV-2 variants and is unaffected by cross-contamination, demonstrating its utility for improved accuracy of COVID-19 diagnostics. IMPORTANCE In this work, we study upper respiratory tract gene expression to develop and validate a 2-gene host-based COVID-19 diagnostic classifier and then demonstrate its implementation in a clinically practical qPCR assay. We find that the host classifier has utility for mitigating false-negative results, for example due to SARS-CoV-2 variants harboring mutations at primer target sites, and for mitigating false-positive viral PCR results due to laboratory cross-contamination. Both types of error carry serious consequences of either unrecognized viral transmission or unnecessary isolation and contact tracing. This work is directly relevant to the ongoing COVID-19 pandemic given the continued emergence of viral variants and the continued challenges of false-positive PCR assays. It also suggests the feasibility of pan-respiratory virus host-based diagnostics that would have value in congregate settings, such as hospitals and nursing homes, where unrecognized respiratory viral transmission is of particular concern.

Nucleus Composition in Transition-based Dependency Parsing

Abstract Dependency-based approaches to syntactic analysis assume that syntactic structure can be analyzed in terms of binary asymmetric dependency relations holding between elementary syntactic units. Computational models for dependency parsing almost universally assume that an elementary syntactic unit is a word, while the influential theory of Lucien Tesnière instead posits a more abstract notion of nucleus, which may be realized as one or more words. In this article, we investigate the effect of enriching computational parsing models with a concept of nucleus inspired by Tesnière. We begin by reviewing how the concept of nucleus can be defined in the framework of Universal Dependencies, which has become the de facto standard for training and evaluating supervised dependency parsers, and explaining how composition functions can be used to make neural transition-based dependency parsers aware of the nuclei thus defined. We then perform an extensive experimental study, using data from 20 languages to assess the impact of nucleus composition across languages with different typological characteristics, and utilizing a variety of analytical tools including ablation, linear mixed-effects models, diagnostic classifiers, and dimensionality reduction. The analysis reveals that nucleus composition gives small but consistent improvements in parsing accuracy for most languages, and that the improvement mainly concerns the analysis of main predicates, nominal dependents, clausal dependents, and coordination structures. Significant factors explaining the rate of improvement across languages include entropy in coordination structures and frequency of certain function words, in particular determiners. Analysis using dimensionality reduction and diagnostic classifiers suggests that nucleus composition increases the similarity of vectors representing nuclei of the same syntactic type.