A Transcriptome-wide Translational Program Defined by LIN28B Expression Level.

LIN28 RNA binding proteins are dynamically expressed throughout mammalian development and during disease. However, it remains unclear how changes in LIN28 expression define patterns of post-transcriptional gene regulation. Here we show that LIN28 expression level is a key variable that sets the magnitude of protein translation. By systematically varying LIN28B protein levels in human cells, we discovered a dose-dependent divergence in transcriptome-wide ribosome occupancy that enabled the formation of two discrete translational subpopulations composed of nearly all expressed genes. This bifurcation in gene expression was mediated by a redistribution in Argonaute association, from let-7 to non-let-7 microRNA families, resulting in a global shift in cellular miRNA activity. Post-transcriptional effects were scaled across the physiological LIN28 expression range. Together, these data highlight the central importance of RBP expression level and its ability to encode regulation.

Identification of a robust DNA methylation signature for Fanconi anemia.

Fanconi anemia (FA) is a clinically variable and genetically heterogeneous cancer-predisposing disorder representing the most common bone marrow failure syndrome. It is caused by inactivating predominantly biallelic mutations involving >20 genes encoding proteins with roles in the FA/BRCA DNA repair pathway. Molecular diagnosis of FA is challenging due to the wide spectrum of the contributing gene mutations and structural rearrangements. The assessment of chromosomal fragility after exposure to DNA cross-linking agents is generally required to definitively confirm diagnosis. We assessed peripheral blood genome-wide DNA methylation (DNAm) profiles in 25 subjects with molecularly confirmed clinical diagnosis of FA (FANCA complementation group) using Illumina's Infinium EPIC array. We identified 82 differentially methylated CpG sites that allow to distinguish subjects with FA from healthy individuals and subjects with other genetic disorders, defining an FA-specific DNAm signature. The episignature was validated using a second cohort of subjects with FA involving different complementation groups, documenting broader genetic sensitivity and demonstrating its specificity using the EpiSign Knowledge Database. The episignature properly classified DNA samples obtained from bone marrow aspirates, demonstrating robustness. Using the selected probes, we trained a machine-learning model able to classify EPIC DNAm profiles in molecularly unsolved cases. Finally, we show that the generated episignature includes CpG sites that do not undergo functional selective pressure, allowing diagnosis of FA in individuals with reverted phenotype due to gene conversion. These findings provide a tool to accelerate diagnostic testing in FA and broaden the clinical utility of DNAm profiling in the diagnostic setting.

Exon-Skipping-Based Subtyping of Colorectal Cancers.

BACKGROUND & AIMS: The identification of colorectal cancer (CRC) molecular subtypes has prognostic and potentially diagnostic value for patients, yet reliable subtyping remains unavailable in the clinic. The current consensus molecular subtype (CMS) classification in CRCs is based on complex RNA expression patterns quantified at the gene level. The clinical application of these methods, however, is challenging due to high uncertainty of single-sample classification and associated costs. Alternative splicing, which strongly contributes to transcriptome diversity, has rarely been used for tissue type classification. Here, we present an AS-based CRC subtyping framework sensitive to differential exon use that can be adapted for clinical application. METHODS: Unsupervised clustering was used to measure the strength of association between different categories of alternative splicing and CMS. To build a classifier, the ground truth for CMS labels was derived from expression data quantified at the gene level. Feature selection was achieved through bootstrapping and L1-penalized estimation. The resulting feature space was used to construct a subtype prediction framework applicable to single and multiple samples. The performance of the models was evaluated on unseen CRCs from 2 independent sources (Indivumed, n = 129; The Cancer Genome Atlas, n = 99). RESULTS: We developed a CRC subtype identifier based on 29 exon-skipping events that accurately classifies unseen tumors and enables more precise differentiation of subtypes characterized by distinct biological and prognostic features as compared to classifiers based on gene expression. CONCLUSIONS: Here, we demonstrate that a small number of exon-skipping events can reliably classify CRC subtypes using individual patient specimens in a manner suitable to clinical application.

Quantitative annotations of T-Cell repertoire specificity.

The specificity of a T-cell receptor (TCR) repertoire determines personalized immune capacity. Existing methods have modeled the qualitative aspects of TCR specificity, while the quantitative aspects remained unaddressed. We developed a package, TCRanno, to quantify the specificity of TCR repertoires. We created deep-learning-based, epitope-aware vector embeddings to infer individual TCR specificity. Then we aggregated clonotype frequencies of TCRs to obtain a quantitative profile of repertoire specificity at epitope, antigen and organism levels. Applying TCRanno to 4195 TCR repertoires revealed quantitative changes in repertoire specificity upon infections, autoimmunity and cancers. Specifically, TCRanno found cytomegalovirus-specific TCRs in seronegative healthy individuals, supporting the possibility of abortive infections. TCRanno discovered age-accumulated fraction of severe acute respiratory syndrome coronavirus 2 specific TCRs in pre-pandemic samples, which may explain the aggressive symptoms and age-related severity of coronavirus disease 2019. TCRanno also identified the encounter of Hepatitis B antigens as a potential trigger of systemic lupus erythematosus. TCRanno annotations showed capability in distinguishing TCR repertoires of healthy and cancers including melanoma, lung and breast cancers. TCRanno also demonstrated usefulness to single-cell TCRseq+gene expression data analyses by isolating T-cells with the specificity of interest.

Machine learning aided single cell image analysis improves understanding of morphometric heterogeneity of human mesenchymal stem cells.

The multipotent stem cells of our body have been largely harnessed in biotherapeutics. However, as they are derived from multiple anatomical sources, from different tissues, human mesenchymal stem cells (hMSCs) are a heterogeneous population showing ambiguity in their in vitro behavior. Intra-clonal population heterogeneity has also been identified and pre-clinical mechanistic studies suggest that these cumulatively depreciate the therapeutic effects of hMSC transplantation. Although various biomarkers identify these specific stem cell populations, recent artificial intelligence-based methods have capitalized on the cellular morphologies of hMSCs, opening a new approach to understand their attributes. A robust and rapid platform is required to accommodate and eliminate the heterogeneity observed in the cell population, to standardize the quality of hMSC therapeutics globally. Here, we report our primary findings of morphological heterogeneity observed within and across two sources of hMSCs namely, stem cells from human exfoliated deciduous teeth (SHEDs) and human Wharton jelly mesenchymal stem cells (hWJ MSCs), using real-time single-cell images generated on immunophenotyping by imaging flow cytometry (IFC). We used the ImageJ software for identification and comparison between the two types of hMSCs using statistically significant morphometric descriptors that are biologically relevant. To expand on these insights, we have further applied deep learning methods and successfully report the development of a Convolutional Neural Network-based image classifier. In our research, we introduced a machine learning methodology to streamline the entire procedure, utilizing convolutional neural networks and transfer learning for binary classification, achieving an accuracy rate of 97.54%. We have also critically discussed the challenges, comparisons between solutions and future directions of machine learning in hMSC classification in biotherapeutics.

A DNA Methylation Signature from Buccal Swabs to Identify Tuberculosis Infection.

BACKGROUND: Tuberculosis (TB) is amongst the largest infectious causes of death worldwide and there is a need for a time- and resource-effective diagnostic method. In this novel and exploratory study, we show the potential of using buccal swabs to collect human DNA and investigate the DNA methylation (DNAm) signatures as a diagnostic tool for TB. METHODS: Buccal swabs were collected from pulmonary TB patients (n= 7), TB exposed (n= 7), and controls (n= 9) in Sweden. Using Illumina MethylationEPIC array the DNAm status was determined. RESULTS: We identified 5644 significant differentially methylated CpG sites between the patients and controls. Performing the analysis on a validation cohort of samples collected in Kenya and Peru (patients, n=26; exposed, n=9; control, n=10) confirmed the DNAm signature. We identified a TB consensus disease module, significantly enriched in TB-associated genes. Lastly, we used machine learning to identify a panel of seven CpG sites discriminative for TB and developed a TB classifier. In the validation cohort the classifier performed with an AUC of 0.94, sensitivity of 0.92, and specificity of 1. CONCLUSION: In summary, the result from this study shows clinical implications of using DNAm signatures from buccal swabs to explore new diagnostic strategies for TB.

MEM-FET: Essential protein prediction using membership feature and machine learning approach.

Proteins are played key roles in different functionalities in our daily life. All functional roles of a protein are a bit enhanced in interaction compared to individuals. Identification of essential proteins of an organism is a time consume and costly task during observation in the wet lab. The results of observation in wet lab always ensure high reliability and accuracy in the biological ground. Essential protein prediction using computational approaches is an alternative choice in research. It proves its significance rapidly in day-to-day life as well as reduces the experimental cost of wet lab effectively. Existing computational methods were implemented using Protein interaction networks (PPIN), Sequence, Gene Expression Dataset (GED), Gene Ontology (GO), Orthologous groups, and Subcellular localized datasets. Machine learning has diverse categories of features that enable to model and predict essential macromolecules of understudied organisms. A novel methodology MEM-FET (membership feature) is predicted based on features, that is, edge clustering coefficient, Average clustering coefficient, subcellular localization, and Gene Ontology within a compartment of common neighbors. The accuracy (ACC) values of the predicted true positive (TP) essential proteins are 0.79, 0.74, 0.78, and 0.71 for YHQ, YMIPS, YDIP, and YMBD datasets. An enriched set of essential proteins are also predicted using the MEM-FET algorithm. Ensemble ML also validated the proposed model with an accuracy of 60%. It has been predicted that MEM-FET algorithms outperform other existing algorithms with an ACC value of 80% for the yeast dataset.

Characterization of fission and fusion mitochondrial dynamics in HD fibroblasts according to patient's severity status.

Huntington's Disease (HD) is an inheritable neurodegenerative condition caused by an expanded CAG trinucleotide repeat in the HTT gene with a direct correlation between CAG repeats expansion and disease severity with earlier onset-of- disease. Previously we have shown that primary skin fibroblasts from HD patients exhibit unique phenotype disease features, including distinct nuclear morphology and perturbed actin cap linked with cell motility, that are correlated with the HD patient disease severity. Here we provide further evidence that mitochondrial fission-fusion morphology balance dynamics, classified using a custom image-based high-content analysis (HCA) machine learning tool, that improved correlation with HD severity status. This mitochondrial phenotype is supported by appropriate changes in fission-fusion biomarkers (Drp1, MFN1, MFN2, VAT1) levels in the HD patients' fibroblasts. These findings collectively point towards a dysregulation in mitochondrial dynamics, where both fission and fusion processes may be disrupted in HD cells compared to healthy controls. This study shows for the first time a methodology that enables identification of HD phenotype before patient's disease onset (Premanifest). Therefore, we believe that this tool holds a potential for improving precision in HD patient's diagnostics bearing the potential to evaluate alterations in mitochondrial dynamics throughout the progression of HD, offering valuable insights into the molecular mechanisms and drug therapy evaluation underlying biological differences in any disease stage.

Morphologic patterns observed in prostate biopsy cases with discrepant grade group and molecular risk classification.

BACKGROUND: Molecular-based risk classifier tests are increasingly being utilized by urologists and radiation oncologists to guide clinical decision making. The Decipher prostate biopsy test is a 22-gene RNA biomarker assay designed to predict likelihood of high-grade disease at radical prostatectomy and risk of metastasis and mortality. The test provides a risk category of low, intermediate, or high. We investigated histologic features of biopsies in which the Grade Group (GG) and Decipher risk category (molecular risk) were discrepant. METHODS: Our institutional urologic outcomes database was searched for men who underwent prostate biopsies with subsequent Decipher testing from 2016 to 2020. We defined discrepant GG and molecular risk as either GG1-2 with high Decipher risk category or GG ≥ 3 with low Decipher risk category. The biopsy slide on which Decipher testing was performed was re-reviewed for GG and various histologic features, including % Gleason pattern 4, types of Gleason pattern 4 and 5, other "high risk" features (e.g., complex papillary, ductal carcinoma, intraductal carcinoma [IDC]), and other unusual and often "difficult to grade" patterns (e.g., atrophic carcinoma, mucin rupture, pseudohyperplastic carcinoma, collagenous fibroplasia, foamy gland carcinoma, carcinoma with basal cell marker expression, carcinoma with prominent vacuoles, and stromal reaction). Follow-up data was also obtained from the electronic medical record. RESULTS: Of 178 men who underwent prostate biopsies and had Decipher testing performed, 41 (23%) had discrepant GG and molecular risk. Slides were available for review for 33/41 (80%). Of these 33 patients, 23 (70%) had GG1-2 (GG1 n = 5, GG2 n = 18) with high Decipher risk, and 10 (30%) had GG ≥ 3 with low Decipher risk. Of the 5 GG1 cases, one case was considered GG2 on re-review; no other high risk features were identified but each case showed at least one of the following "difficult to grade" patterns: 3 atrophic carcinoma, 1 collagenous fibroplasia, 1 carcinoma with mucin rupture, and 1 carcinoma with basal cell marker expression. Of the 18 GG2 high Decipher risk cases, 2 showed GG3 on re-review, 5 showed large cribriform and/or other high risk features, and 10 showed a "difficult to grade" pattern. Of the 10 GG ≥ 3 low Decipher risk cases, 5 had known high risk features including 2 with large cribriform, 1 with IDC, and 1 with Gleason pattern 5. CONCLUSIONS: In GG1-2 high Decipher risk cases, difficult to grade patterns were frequently seen in the absence of other known high risk morphologic features; whether these constitute true high risk cases requires further study. In the GG ≥ 3 low Decipher risk cases, aggressive histologic patterns such as large cribriform and IDC were observed in half (50%) of cases; therefore, the molecular classifier may not capture all high risk histologic patterns.

Latent pathway-based Bayesian models to identify intervenable factors of racial disparities in breast cancer stage at diagnosis.

PURPOSE: We built Bayesian Network (BN) models to explain roles of different patient-specific factors affecting racial differences in breast cancer stage at diagnosis, and to identify healthcare related factors that can be intervened to reduce racial health disparities. METHODS: We studied women age 67-74 with initial diagnosis of breast cancer during 2006-2014 in the National Cancer Institute's SEER-Medicare dataset. Our models included four measured variables (tumor grade, hormone receptor status, screening utilization and biopsy delay) expressed through two latent pathways-a tumor biology path, and health-care access/utilization path. We used various Bayesian model assessment tools to evaluate these two latent pathways as well as each of the four measured variables in explaining racial disparities in stage-at-diagnosis. RESULTS: Among 3,010 Black non-Hispanic (NH) and 30,310 White NH breast cancer patients, respectively 70.2% vs 76.9% were initially diagnosed at local stage, 25.3% vs 20.3% with regional stage, and 4.56% vs 2.80% with distant stage-at-diagnosis. Overall, BN performed approximately 4.7 times better than Classification And Regression Tree (CART) (Breiman L, Friedman JH, Stone CJ, Olshen RA. Classification and regression trees. CRC press; 1984) in predicting stage-at-diagnosis. The utilization of screening mammography is the most prominent contributor to the accuracy of the BN model. Hormone receptor (HR) status and tumor grade are useful for explaining racial disparity in stage-at diagnosis, while log-delay in biopsy impeded good prediction. CONCLUSIONS: Mammography utilization had a significant effect on racial differences in breast cancer stage-at-diagnosis, while tumor biology factors had less impact. Biopsy delay also aided in predicting local and regional stages-at-diagnosis for Black NH women but not for white NH women.

RdRp-based sensitive taxonomic classification of RNA viruses for metagenomic data.

With advances in library construction protocols and next-generation sequencing technologies, viral metagenomic sequencing has become the major source for novel virus discovery. Conducting taxonomic classification for metagenomic data is an important means to characterize the viral composition in the underlying samples. However, RNA viruses are abundant and highly diverse, jeopardizing the sensitivity of comparison-based classification methods. To improve the sensitivity of read-level taxonomic classification, we developed an RNA-dependent RNA polymerase (RdRp) gene-based read classification tool RdRpBin. It combines alignment-based strategy with machine learning models in order to fully exploit the sequence properties of RdRp. We tested our method and compared its performance with the state-of-the-art tools on the simulated and real sequencing data. RdRpBin competes favorably with all. In particular, when the query RNA viruses share low sequence similarity with the known viruses ($\sim 0.4$), our tool can still maintain a higher F-score than the state-of-the-art tools. The experimental results on real data also showed that RdRpBin can classify more RNA viral reads with a relatively low false-positive rate. Thus, RdRpBin can be utilized to classify novel and diverged RNA viruses.

A novel fusion technology utilizing complex network and sequence information for FAD-binding site identification.

Flavin adenine dinucleotide (FAD) binding sites play an increasingly important role as useful targets for inhibiting bacterial infections. To reveal protein topological structural information as a reasonable complement for the identification FAD-binding sites, we designed a novel fusion technology according to sequence and complex network. The specially designed feature vectors were combined and fed into CatBoost for model construction. Moreover, due to the minority class (positive samples) is more significant for biological researches, a random under-sampling technique was applied to solve the imbalance. Compared with the previous methods, our methods achieved the best results for two independent test datasets. Especially, the MCC obtained by FADsite and FADsite_seq were 14.37 %-53.37 % and 21.81 %-60.81 % higher than the results of existing methods on Test6; and they showed improvements ranging from 6.03 % to 21.96 % and 19.77 %-35.70 % on Test4. Meanwhile, statistical tests show that our methods significantly differ from the state-of-the-art methods and the cross-entropy loss shows that our methods have high certainty. The excellent results demonstrated the effectiveness of using sequence and complex network information in identifying FAD-binding sites. It may be complementary to other biological studies. The data and resource codes are available at https://github.com/Kangxiaoneuq/FADsite.

A mathematical model of competition between fiber and mucin degraders in the gut provides a possible explanation for mucus thinning.

The human gut microbiota relies on complex carbohydrates (glycans) for energy and growth, primarily dietary fiber and host-derived mucins. We introduce a mathematical model of a glycan generalist and a mucin specialist in a two-compartment chemostat model of the human colon. Our objective is to characterize the influence of dietary fiber and mucin supply on the abundance of mucin-degrading species within the gut ecosystem. Current mathematical gut reactor models that include the enzymatic degradation of glycans do not differentiate between glycan types and their degraders. The model we present distinguishes between a generalist that can degrade both dietary fiber and mucin, and a specialist species that can only degrade mucin. The integrity of the colonic mucus barrier is essential for overall human health and well-being, with the mucin specialist Akkermanisa muciniphila being associated with a healthy mucus layer. Competition, particularly between the specialist and generalists like Bacteroides thetaiotaomicron, may lead to mucus layer erosion, especially during periods of dietary fiber deprivation. Our model treats the colon as a gut reactor system, dividing it into two compartments that represent the lumen and the mucus of the gut, resulting in a complex system of ordinary differential equations with a large and uncertain parameter space. To understand the influence of model parameters on long-term behavior, we employ a random forest classifier, a supervised machine learning method. Additionally, a variance-based sensitivity analysis is utilized to determine the sensitivity of steady-state values to changes in model parameter inputs. By constructing this model, we can investigate the underlying mechanisms that control gut microbiota composition and function, free from confounding factors.

Establishing the utility of multi-platform liquid biopsy by integrating the CSF methylome and proteome in CNS tumours.

BACKGROUND: Liquid biopsy represents a major development in cancer research, with significant translational potential. Similarly, it is increasingly recognized that multi-omic molecular approaches are a powerful avenue through which to understand complex and heterogeneous disease biology. We hypothesize that merging these two promising frontiers of cancer research will improve the discriminatory capacity of current models and allow for improved clinical utility. METHODS: We have compiled a cohort of patients with glioblastoma, brain metastasis, and primary central nervous system lymphoma. Cell-free methylated DNA immunoprecipitation (cfMeDIP) and shotgun proteomic profiling was obtained from the cerebrospinal fluid (CSF) of each patient and used to build tumour-specific classifiers. RESULTS: We show that the DNA methylation and protein profiles of cerebrospinal fluid can be integrated to fully discriminate lymphoma from its diagnostic counterparts with perfect AUC of 1 (95% confidence interval 1-1) and 100% specificity, significantly outperforming single-platform classifiers. CONCLUSIONS: We present the most specific and accurate CNS lymphoma classifier to date and demonstrates the synergistic capability of multi-platform liquid biopsies. This has far-reaching translational utility for patients with newly diagnosed intra-axial brain tumours.

Metabolic signatures derived from whole-brain MR-spectroscopy identify early tumor progression in high-grade gliomas using machine learning.

PURPOSE: Recurrence for high-grade gliomas is inevitable despite maximal safe resection and adjuvant chemoradiation, and current imaging techniques fall short in predicting future progression. However, we introduce a novel whole-brain magnetic resonance spectroscopy (WB-MRS) protocol that delves into the intricacies of tumor microenvironments, offering a comprehensive understanding of glioma progression to inform expectant surgical and adjuvant intervention. METHODS: We investigated five locoregional tumor metabolites in a post-treatment population and applied machine learning (ML) techniques to analyze key relationships within seven regions of interest: contralateral normal-appearing white matter (NAWM), fluid-attenuated inversion recovery (FLAIR), contrast-enhancing tumor at time of WB-MRS (Tumor), areas of future recurrence (AFR), whole-brain healthy (WBH), non-progressive FLAIR (NPF), and progressive FLAIR (PF). Five supervised ML classification models and a neural network were developed, optimized, trained, tested, and validated. Lastly, a web application was developed to host our novel calculator, the Miami Glioma Prediction Map (MGPM), for open-source interaction. RESULTS: Sixteen patients with histopathological confirmation of high-grade glioma prior to WB-MRS were included in this study, totaling 118,922 whole-brain voxels. ML models successfully differentiated normal-appearing white matter from tumor and future progression. Notably, the highest performing ML model predicted glioma progression within fluid-attenuated inversion recovery (FLAIR) signal in the post-treatment setting (mean AUC = 0.86), with Cho/Cr as the most important feature. CONCLUSIONS: This study marks a significant milestone as the first of its kind to unveil radiographic occult glioma progression in post-treatment gliomas within 8 months of discovery. These findings underscore the utility of ML-based WB-MRS growth predictions, presenting a promising avenue for the guidance of early treatment decision-making. This research represents a crucial advancement in predicting the timing and location of glioblastoma recurrence, which can inform treatment decisions to improve patient outcomes.

Integrated biomarker profiling for predicting the response of type 2 diabetes to metformin.

AIM: To explore biomarkers that can predict the response of type 2 diabetes (T2D) patients to metformin at an early stage to provide better treatment for T2D. METHODS: T2D patients with (responders) or without response (non-responders) to metformin were recruited, and their serum samples were used for metabolomic analysis to identify candidate biomarkers. Moreover, the efficacy of metformin was verified by insulin-resistant mice, and the candidate biomarkers were verified to determine the biomarkers. Five different machine learning methods were used to construct the integrated biomarker profiling (IBP) with the biomarkers to predict the response of T2D patients to metformin. RESULTS: A total of 73 responders and 63 non-responders were recruited, and 88 differential metabolites were identified in the serum samples. After being verified in mice, 19 of the 88 were considered as candidate biomarkers. Next, after metformin regulation, nine candidate biomarkers were confirmed as the biomarkers. After comparing five machine learning models, the nine biomarkers were constructed into the IBP for predicting the response of T2D patients to metformin based on the Naïve Bayes classifier, which was verified with an accuracy of 89.70%. CONCLUSIONS: The IBP composed of nine biomarkers can be used to predict the response of T2D patients to metformin, enabling clinicians to start a combined medication strategy as soon as possible if T2D patients do not respond to metformin.

Prediction of hypertension risk based on multiple feature fusion.

OBJECTIVE: In the application of machine learning to the prediction of hypertension, many factors have seriously affected the classification accuracy and generalization performance. We propose a pulse wave classification model based on multi-feature fusion for accuracy prediction of hypertension. METHODS AND MATERIALS: We propose an ensemble under-sampling model with dynamic weights to decrease the influence of class imbalance on classification, further to automatically classify of hypertension on inquiry diagnosis. We also build a deep learning model based on hybrid attention mechanism, which transforms pulse waves to feature maps for extraction of in-depth features, so as to automatically classify hypertension on pulse diagnosis. We build the multi-feature fusion model based on dynamic Dempster/Shafer (DS) theory combining inquiry diagnosis and pulse diagnosis to enhance fault tolerance of prediction for multiple classifiers. In addition, this study calculates feature importance ranking of scale features on inquiry diagnosis and temporal and frequency-domain features on pulse diagnosis. RESULTS: The accuracy, sensitivity, specificity, F1-score and G-mean after 5-fold cross-validation were 94.08%, 93.43%, 96.86%, 93.45% and 95.12%, respectively, based on the hypertensive samples of 409 cases from Longhua Hospital affiliated to Shanghai University of Traditional Chinese Medicine and Hospital of Integrated Traditional Chinese and Western Medicine. We find the key factors influencing hypertensive classification accuracy, so as to assist in the prevention and clinical diagnosis of hypertension. CONCLUSION: Compared with the state-of-the-art models, the multi-feature fusion model effectively utilizes the patients' correlated multimodal features, and has higher classification accuracy and generalization performance.

Deep learning uncertainty quantification for clinical text classification.

INTRODUCTION: Machine learning algorithms are expected to work side-by-side with humans in decision-making pipelines. Thus, the ability of classifiers to make reliable decisions is of paramount importance. Deep neural networks (DNNs) represent the state-of-the-art models to address real-world classification. Although the strength of activation in DNNs is often correlated with the network's confidence, in-depth analyses are needed to establish whether they are well calibrated. METHOD: In this paper, we demonstrate the use of DNN-based classification tools to benefit cancer registries by automating information extraction of disease at diagnosis and at surgery from electronic text pathology reports from the US National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) population-based cancer registries. In particular, we introduce multiple methods for selective classification to achieve a target level of accuracy on multiple classification tasks while minimizing the rejection amount-that is, the number of electronic pathology reports for which the model's predictions are unreliable. We evaluate the proposed methods by comparing our approach with the current in-house deep learning-based abstaining classifier. RESULTS: Overall, all the proposed selective classification methods effectively allow for achieving the targeted level of accuracy or higher in a trade-off analysis aimed to minimize the rejection rate. On in-distribution validation and holdout test data, with all the proposed methods, we achieve on all tasks the required target level of accuracy with a lower rejection rate than the deep abstaining classifier (DAC). Interpreting the results for the out-of-distribution test data is more complex; nevertheless, in this case as well, the rejection rate from the best among the proposed methods achieving 97% accuracy or higher is lower than the rejection rate based on the DAC. CONCLUSIONS: We show that although both approaches can flag those samples that should be manually reviewed and labeled by human annotators, the newly proposed methods retain a larger fraction and do so without retraining-thus offering a reduced computational cost compared with the in-house deep learning-based abstaining classifier.

A machine learning-based prediction model for postoperative delirium in cardiac valve surgery using electronic health records.

BACKGROUND: Previous models for predicting delirium after cardiac surgery remained inadequate. This study aimed to develop and validate a machine learning-based prediction model for postoperative delirium (POD) in cardiac valve surgery patients. METHODS: The electronic medical information of the cardiac surgical intensive care unit (CSICU) was extracted from a tertiary and major referral hospital in southern China over 1 year, from June 2019 to June 2020. A total of 507 patients admitted to the CSICU after cardiac valve surgery were included in this study. Seven classical machine learning algorithms (Random Forest Classifier, Logistic Regression, Support Vector Machine Classifier, K-nearest Neighbors Classifier, Gaussian Naive Bayes, Gradient Boosting Decision Tree, and Perceptron.) were used to develop delirium prediction models under full (q = 31) and selected (q = 19) feature sets, respectively. RESULT: The Random Forest classifier performs exceptionally well in both feature datasets, with an Area Under the Curve (AUC) of 0.92 for the full feature dataset and an AUC of 0.86 for the selected feature dataset. Additionally, it achieves a relatively lower Expected Calibration Error (ECE) and the highest Average Precision (AP), with an AP of 0.80 for the full feature dataset and an AP of 0.73 for the selected feature dataset. To further evaluate the best-performing Random Forest classifier, SHAP (Shapley Additive Explanations) was used, and the importance matrix plot, scatter plots, and summary plots were generated. CONCLUSIONS: We established machine learning-based prediction models to predict POD in patients undergoing cardiac valve surgery. The random forest model has the best predictive performance in prediction and can help improve the prognosis of patients with POD.

Parietal cortical alpha/beta suppression during prospective memory retrieval.

Prospective memory (PM) impairment is among the most frequent memory complaints, yet little is known about the underlying neural mechanisms. PM for a planned intention may be achieved through strategic monitoring of the environment for cues, involving ongoing attentional processes, or through spontaneous retrieval. We hypothesized that parietal spectral power modulation accompanies prospectively encoded intention retrieval, irrespective of PM retrieval approach. A cognitively engaging arithmetic-based ongoing task (OGT) was employed to encourage spontaneous retrieval, with a focal, internally generated PM cue to eliminate OGT/PM trial differentiation based on perceptual or conceptual PM cue features. Two PM repetition frequencies were used to vary the extent of strategic monitoring. We observed a transient parietal alpha/beta spectral power reduction directly preceding the response, which was distinguishable on a single trial basis, as revealed by an OGT/PM trial classification rate exceeding 70% using linear discriminant analysis. The alpha/beta idling rhythm reflects cortical inhibition. A disengagement of task-relevant neural assemblies from this rhythm, reflected in alpha/beta power reduction, is deemed to increase information content, facilitate information integration, and enable engagement of neural assemblies in task-related cortical networks. The observed power reduction is consistent with the Dual Pathways model, where PM strategies converge at the PM retrieval stage.