Subpopulation shift, characterized by a disparity in subpopulation distribution between the training and target datasets, can significantly degrade the performance of machine learning models. Current solutions to subpopulation shift involve modifying empirical risk minimization with re-weighting strategies to improve generalization. This strategy relies on assumptions about the number and nature of subpopulations and annotations on group membership, which are unavailable for many real-world datasets. Instead, we propose using an ensemble of diverse classifiers to adaptively capture risk associated with subpopulations. Given a feature extractor network, we replace its standard linear classification layer with a mixture of prototypical classifiers, where each member is trained to classify the data while focusing on different features and samples from other members. In empirical evaluation on nine real-world datasets, covering diverse domains and kinds of subpopulation shift, our method of Diverse Prototypical Ensembles (DPEs) often outperforms the prior state-of-the-art in worst-group accuracy. The code is available at https://github.com/minhto2802/dpe4subpop.

BACKGROUND: Accurate cancer classification is essential for correct treatment selection and better prognostication. microRNAs (miRNAs) are small RNA molecules that negatively regulate gene expression, and their dyresgulation is a common disease mechanism in many cancers. Through a clearer understanding of miRNA dysregulation in cancer, improved mechanistic knowledge and better treatments can be sought. RESULTS: We present a topology-preserving deep learning framework to study miRNA dysregulation in cancer. Our study comprises miRNA expression profiles from 3685 cancer and non-cancer tissue samples and hierarchical annotations on organ and neoplasticity status. Using unsupervised learning, a two-dimensional topological map is trained to cluster similar tissue samples. Labelled samples are used after training to identify clustering accuracy in terms of tissue-of-origin and neoplasticity status. In addition, an approach using activation gradients is developed to determine the attention of the networks to miRNAs that drive the clustering. Using this deep learning framework, we classify the neoplasticity status of held-out test samples with an accuracy of 91.07%, the tissue-of-origin with 86.36%, and combined neoplasticity status and tissue-of-origin with an accuracy of 84.28%. The topological maps display the ability of miRNAs to recognize tissue types and neoplasticity status. Importantly, when our approach identifies samples that do not cluster well with their respective classes, activation gradients provide further insight in cancer subtypes or grades. CONCLUSIONS: An unsupervised deep learning approach is developed for cancer classification and interpretation. This work provides an intuitive approach for understanding molecular properties of cancer and has significant potential for cancer classification and treatment selection.

OBJECTIVE: Abnormalities in impulse propagation and cardiac repolarization are frequent in hypertrophic cardiomyopathy (HCM), leading to abnormalities in 12-lead electrocardiograms (ECGs). Computational ECG analysis can identify electrophysiological and structural remodeling and predict arrhythmias. This requires accurate ECG segmentation. It is unknown whether current segmentation methods developed using datasets containing annotations for mostly normal heartbeats perform well in HCM. Here, we present a segmentation method to effectively identify ECG waves across 12-lead HCM ECGs. METHODS: We develop (1) a web-based tool that permits manual annotations of P, P', QRS, R', S', T, T', U, J, epsilon waves, QRS complex slurring, and atrial fibrillation by 3 experts and (2) an easy-to-implement segmentation method that effectively identifies ECG waves in normal and abnormal heartbeats. Our method was tested on 131 12-lead HCM ECGs and 2 public ECG sets to evaluate its performance in non-HCM ECGs. RESULTS: Over the HCM dataset, our method obtained a sensitivity of 99.2% and 98.1% and a positive predictive value of 92% and 95.3% when detecting QRS complex and T-offset, respectively, significantly outperforming a state-of-the-art segmentation method previously employed for HCM analysis. Over public ECG sets, it significantly outperformed 3 state-of-the-art methods when detecting P-onset and peak, T-offset, and QRS-onset and peak regarding the positive predictive value and segmentation error. It performed at a level similar to other methods in other tasks. CONCLUSION: Our method accurately identified ECG waves in the HCM dataset, outperforming a state-of-the-art method, and demonstrated similar good performance as other methods in normal/non-HCM ECG sets.

Temporal enhanced ultrasound (TeUS), comprising the analysis of variations in backscattered signals from a tissue over a sequence of ultrasound frames, has been previously proposed as a new paradigm for tissue characterization. In this paper, we propose to use deep recurrent neural networks (RNN) to explicitly model the temporal information in TeUS. By investigating several RNN models, we demonstrate that long short-term memory (LSTM) networks achieve the highest accuracy in separating cancer from benign tissue in the prostate. We also present algorithms for in-depth analysis of LSTM networks. Our in vivo study includes data from 255 prostate biopsy cores of 157 patients. We achieve area under the curve, sensitivity, specificity, and accuracy of 0.96, 0.76, 0.98, and 0.93, respectively. Our result suggests that temporal modeling of TeUS using RNN can significantly improve cancer detection accuracy over previously presented works.

Segmentation of the wrist bones in CT images has been frequently used in different clinical applications including arthritis evaluation, bone age assessment and image-guided interventions. The major challenges include non-uniformity and spongy textures of the bone tissue as well as narrow inter-bone spaces. In this work, we propose an automatic wrist bone segmentation technique for CT images based on a statistical model that captures the shape and pose variations of the wrist joint across 60 example wrists at nine different wrist positions. To establish the correspondences across the training shapes at neutral positions, the wrist bone surfaces are jointly aligned using a group-wise registration framework based on a Gaussian Mixture Model. Principal component analysis is then used to determine the major modes of shape variations. The variations in poses not only across the population but also across different wrist positions are incorporated in two pose models. An intra-subject pose model is developed by utilizing the similarity transforms at all wrist positions across the population. Further, an inter-subject pose model is used to model the pose variations across different wrist positions. For segmentation of the wrist bones in CT images, the developed model is registered to the edge point cloud extracted from the CT volume through an expectation maximization based probabilistic approach. Residual registration errors are corrected by application of a non-rigid registration technique. We validate the proposed segmentation method by registering the wrist model to a total of 66 unseen CT volumes of average voxel size of 0.38 mm. We report a mean surface distance error of 0.33 mm and a mean Jaccard index of 0.86.

Recent studies highlight temporal ultrasound data as highly promising in differentiating between malignant and benign tissues in prostate cancer patients. Since Hidden Markov Models can be used for capturing order and patterns in time varying signals, we employ them to model temporal aspects of ultrasound data that are typically not incorporated in existing models. By comparing order-preserving and order-altering models, we demonstrate that the order encoded in the series is necessary to model the variability in ultrasound data of prostate tissues. In future studies, we will investigate the influence of order on the differentiation between malignant and benign tissues.

UNLABELLED: Protein interaction network-based pathway analysis (PINBPA) for genome-wide association studies (GWAS) has been developed as a Cytoscape app, to enable analysis of GWAS data in a network fashion. Users can easily import GWAS summary-level data, draw Manhattan plots, define blocks, prioritize genes with random walk with restart, detect enriched subnetworks and test the significance of subnetworks via a user-friendly interface. AVAILABILITY AND IMPLEMENTATION: PINBPA app is freely available in Cytoscape app store. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

UNLABELLED: This paper presents the results of a computer-aided intervention solution to demonstrate the application of RF time series for characterization of prostate cancer, in vivo. METHODS: We pre-process RF time series features extracted from 14 patients using hierarchical clustering to remove possible outliers. Then, we demonstrate that the mean central frequency and wavelet features extracted from a group of patients can be used to build a nonlinear classifier which can be applied successfully to differentiate between cancerous and normal tissue regions of an unseen patient. RESULTS: In a cross-validation strategy, we show an average area under receiver operating characteristic curve (AUC) of 0.93 and classification accuracy of 80%. To validate our results, we present a detailed ultrasound to histology registration framework. CONCLUSION: Ultrasound RF time series results in differentiation of cancerous and normal tissue with high AUC.

In surface-based registration for image-guided interventions, the presence of missing data can be a significant issue. This often arises with real-time imaging modalities such as ultrasound, where poor contrast can make tissue boundaries difficult to distinguish from surrounding tissue. Missing data poses two challenges: ambiguity in establishing correspondences; and extrapolation of the deformation field to those missing regions. To address these, we present a novel non-rigid registration method. For establishing correspondences, we use a probabilistic framework based on a Gaussian mixture model (GMM) that treats one surface as a potentially partial observation. To extrapolate and constrain the deformation field, we incorporate biomechanical prior knowledge in the form of a finite element model (FEM). We validate the algorithm, referred to as GMM-FEM, in the context of prostate interventions. Our method leads to a significant reduction in target registration error (TRE) compared to similar state-of-the-art registration algorithms in the case of missing data up to 30%, with a mean TRE of 2.6 mm. The method also performs well when full segmentations are available, leading to TREs that are comparable to or better than other surface-based techniques. We also analyze robustness of our approach, showing that GMM-FEM is a practical and reliable solution for surface-based registration.

In the last decade many computational approaches have been introduced to model networks of molecular interactions from gene expression data. Such networks can provide an understanding of the regulatory mechanisms in the cells. System identification algorithms refer to a group of approaches that capture the dynamic relationship between the input and output of a system, and provide a deterministic model of its function. These approaches have been extensively developed for engineering systems, and have reasonable computational requirements. In this paper, we present two system identification methods applied to reverse engineering of gene regulatory networks. Gene regulatory networks are constructed as systems where the output to be estimated is an expression profile of a gene, and the inputs are the potential regulators of that gene. The first reverse engineering method is based on orthogonal search and selects terms from a predefined set of gene expression profiles to best fit the expression levels of a given output gene. The second method consists of multiple cascade models; each cascade includes a dynamic component and a static component. Several cascades are used in parallel to reduce the difference of the estimated expression profiles with the actual ones. To assess the performance of the proposed methods, they are applied to a temporal synthetic dataset, a simulated gene expression time series of songbird brain, and yeast Saccharomyces Cerevisiae cell cycle. Results are compared to known mechanisms of the underlying data and the literature, and demonstrate that the proposed approaches capture the underlying interactions as networks.

BACKGROUND: The speed at which biological datasets are being accumulated stands in contrast to our ability to integrate them meaningfully. Large-scale biological databases containing datasets of genes, proteins, cells, organs, and diseases are being created but they are not connected. Integration of these vast but heterogeneous sources of information will allow the systematic and comprehensive analysis of molecular and clinical datasets, spanning hundreds of dimensions and thousands of individuals. This integration is essential to capitalize on the value of current and future molecular- and cellular-level data on humans to gain novel insights about health and disease. RESULTS: We describe a new open-source Cytoscape plugin named iCTNet (integrated Complex Traits Networks). iCTNet integrates several data sources to allow automated and systematic creation of networks with up to five layers of omics information: phenotype-SNP association, protein-protein interaction, disease-tissue, tissue-gene, and drug-gene relationships. It facilitates the generation of general or specific network views with diverse options for more than 200 diseases. Built-in tools are provided to prioritize candidate genes and create modules of specific phenotypes. CONCLUSIONS: iCTNet provides a user-friendly interface to search, integrate, visualize, and analyze genome-scale biological networks for human complex traits. We argue this tool is a key instrument that facilitates systematic integration of disparate large-scale data through network visualization, ultimately allowing the identification of disease similarities and the design of novel therapeutic approaches.The online database and Cytoscape plugin are freely available for academic use at: http://www.cs.queensu.ca/ictnet.

Dynamic Bayesian networks are of particular interest to reverse engineering of gene regulatory networks from temporal transcriptional data. However, the problem of learning the structure of these networks is quite challenging. This is mainly due to the high dimensionality of the search space that makes exhaustive methods for structure learning not practical. Consequently, heuristic techniques such as Hill Climbing are used for DBN structure learning. Hill Climbing is not an efficient method for this purpose as it is prone to get trapped in local optima and the learned network is not very accurate.

Inference of causal interactions among genes known to be involved in the regulation of cell cycle, has received considerable attention in recent years. Capturing the mechanism of gene regulation in the cell cycle is necessary to elucidate both normal and abnormal cell reproduction. Within the last few years, many reverse engineering approaches have been applied to the yeast Saccharomyces cerevisiae. Among these approaches, Dynamic Bayesian Networks (DBNs) are of particular interest. However, learning the structure of these networks is an NP-hard problem. In this paper, we apply DBN with an evolutionary structure learning strategy, M-CMA-ES, to 14 cell cycle regulated genes in the yeast Saccharomyces cerevisiae dataset. The resulting interactions are evaluated and compared with the KEGG pathway as the target network. Precision and sensitivity are also used as evaluation criteria for comparing our inferred network with two previous studies of yeast cell cycle data. The results indicate markedly improved scores for M-CMA-ES approach compared to previous methods.

High throughput genomic techniques produce datasets involving thousands of gene expression profiles. In order to infer biologically meaningful regulatory interactions, a dimensionality reduction must take place to identify genes or groups of genes that are important to the biological system being analyzed. Here we provide a systematic approach to remove dispersible genes from consideration based on their gene expression profiles, and to identify a smaller set of coordinately expressed genes, or metagenes that are biologically related to one and other based on previous biological knowledge. We then apply neural network based reverse engineering techniques to demonstrate that through these dimensionality reduction techniques novel genetic interactions can be identified

High throughput genomic techniques produce datasets involving thousands of gene expression profiles. In order to infer biologically meaningful regulatory interactions, a dimensionality reduction must take place to identify genes or groups of genes that are important to the biological system being analyzed. Here we provide a systematic approach to remove dispersible genes from consideration based on their gene expression profiles, and to identify a smaller set of coordinately expressed genes, or metagenes that are biologically related to one and other based on previous biological knowledge. We then apply neural network based reverse engineering techniques to demonstrate that through these dimensionality reduction techniques novel genetic interactions can be identified

Recombinant interferon beta (IFNβ) is one of the most commonly prescribed treatments for multiple sclerosis; however, the treatment results in partial success producing no benefit in almost half of the patients. We address the problem of identifying minimal and robust sets of molecular biomarkers that are able to present predictive models of response to treatment in multiple sclerosis patients. To achieve this, we utilize a multivariate feature selection and classification framework; OSeMA (orthogonal search model analysis) integrates fast orthogonal search algorithm for feature selection and discriminant analysis for classification. Feature-selection and classification performance of OSeMA are evaluated through comparative studies with two wrapper-approach feature-selection/classification systems. It is demonstrated that the feature-selection of OSeMA significantly reduces the computational time of exhaustive searches while identifying complex gene-gene relationships. Utilizing OSeMA, we are able to construct classification models that are highly predictive of therapy response in MS patients, based on their gene expression data acquired prior to initiation of IFNp treatment

In this paper, we investigate the underlying common gene expression signatures in related cancer types. Shared expression signatures are investigated in breast and ovarian cancers specifically through the definition of four progressively more difficult classification problems. SHEBA, a stochastic Bayesian inference approach, is introduced to identify highly predictive gene sets in the defined classification problems. The heuristics reduce the computation time required to identify the most informative groups of features in the gene space, while providing a good approximation of comparable exhaustive approaches. The breast and ovarian cancer class could be distinguished well from the other classes of cancers using SHEBA in three of the four classification problems, suggesting the existence of a commonality between their gene expressions. Extensive statistical validation and preliminary biological review of the most predictive gene sets demonstrate their robustness and specificity.

Multi-feature analysis of human chromosome images is a major step towards classification of homologous chromosomes. An automatic quantitative classification method is proposed for homolog differentiation using multiple features. This method is based on mutual information maximization applied to an unsupervised neural network architecture. The neural network consists of separate modules which are trained to classify homologs using independent features. Mutual information is then maximized between the outputs of the modules forcing them to produce the same classification results, for a given chromosome. The proposed method was successfully applied to classify homologs of chromosome 16 with 100% accuracy.

Classification of homologous human chromosomes is essential to advanced studies of cancer genetics. This paper describes novel segmentation and classification algorithms to extract multiple features, from microscopy images of chromosomes, for classification purposes. Multicolour images of metaphase chromosomes prepared by applying PNA probes are used for this purpose. Centromeres are segmented using an iterative fuzzy algorithm as well as a gradient method. Moreover, telomere length measurements are performed on chromosome images and normalized for the image database. Multiple intensity features are calculated as a result of the developed algorithms. Heteromorphic chromosomes (such as 16 and 22) are then successfully classified into their parental homologues, based on the calculated multiple features, and used to verify the developed methods.