Search

High throughput technologies in genomics enable the analysis of small alterations in gene expression levels. Patterns of such deviations are an important starting point for the discovery and verification of new biomarker candidates. Identifying such patterns is a challenging task that requires sophisticated machine learning approaches. Currently, there are a variety of classification models, and a common approach is to compare the performance and select the best one for a given classification problem. Since the association between the features of a data set and the performance of a particular classification method is still not fully understood, the main contribution of this work is to provide a new methodology for predicting the prediction results of different classifiers in the field of biomarker discovery. We propose here a three-steps computational workflow that includes an analysis of the data set characteristics, the calculation of the classification accuracy and, finally, the prediction of the resulting classification error. The experiments were carried out on synthetic and microarray datasets. Using this method, we showed that the predictability strongly depends on the discriminatory ability of the features, e.g., sets of genes, in two or multi-class datasets. If a dataset has a certain discriminatory ability, this method enables prediction of the classification performance before applying a learning model. Thus, our results contribute to a better understanding of the relationship between dataset characteristics and the corresponding performance of a machine learning method, and suggest the optimal classification method for a given dataset based on its discriminatory ability.

Hypothermia has a profound impact on the electrophysiological mechanisms of the heart. Experimental investigations provide a better understanding of electrophysiological alterations associated with cooling. However, there is a lack of computer models suitable for simulating the effects of hypothermia in cardio-electrophysiology. In this work, we propose a model that describes the cooling-induced electrophysiological alterations in ventricular tissue in a temperature range from 27°C to 37°C. To model the electrophysiological conditions in a 3D left ventricular tissue block it was essential to consider the following anatomical and physiological parameters in the model: the different cell types (endocardial, M, epicardial), the heterogeneous conductivities in longitudinal, transversal and transmural direction depending on the prevailing temperature, the distinct fiber orientations and the transmural repolarization sequences. Cooling-induced alterations on the morphology of the action potential (AP) of single myocardial cells thereby are described by an extension of the selected Bueno-Orovio model for human ventricular tissue using Q10 temperature coefficients. To evaluate alterations on tissue level, the corresponding pseudo electrocardiogram (pECG) was calculated. Simulations show that cooling-induced AP and pECG-related parameters, i.e. AP duration, morphology of the notch of epicardial AP, maximum AP upstroke velocity, AP rise time, QT interval, QRS duration and J wave formation are in good accordance with literature and our experimental data. The proposed model enables us to further enhance our knowledge of cooling-induced electrophysiological alterations from cellular to tissue level in the heart and may help to better understand electrophysiological mechanisms, e.g. in arrhythmias, during hypothermia.

Metabolic biomarkers may play an important role in the diagnosis, prognostication and assessment of response to pharmacological therapy in complex diseases. The process of discovering new metabolic biomarkers is a non-trivial task which involves a number of bioanalytical processing steps coupled with a computational approach for the search, prioritization and verification of new biomarker candidates. Kinetic analysis provides an additional dimension of complexity in time-series data, allowing for a more precise interpretation of biomarker dynamics in terms of molecular interaction and pathway modulation. A novel network-based computational strategy for the discovery of putative dynamic biomarker candidates is presented, enabling the identification and verification of unexpected metabolic signatures in complex diseases such as myocardial infarction. The novelty of the proposed method lies in combining metabolic time-series data into a superimposed graph representation, highlighting the strength of the underlying kinetic interaction of preselected analytes. Using this approach, we were able to confirm known metabolic signatures and also identify new candidates such as carnosine and glycocholic acid, and pathways that have been previously associated with cardiovascular or related diseases. This computational strategy may serve as a complementary tool for the discovery of dynamic metabolic or proteomic biomarkers in the field of clinical medicine.

BackgroundLittle is known about the effect of cardiac resynchronization therapy (CRT) on endo- and epicardial ventricular activation. Noninvasive imaging of cardiac electrophysiology (NICE) is a novel imaging tool for visualization of both epi- and endocardial ventricular electrical activation.Methodology/principal findingsNICE was performed in ten patients with congestive heart failure (CHF) undergoing CRT and in ten patients without structural heart disease (control group). NICE is a fusion of data from high-resolution ECG mapping with a model of the patient's individual cardiothoracic anatomy created from magnetic resonance imaging. Beat-to-beat endocardial and epicardial ventricular activation sequences were computed during native rhythm as well as during ventricular pacing using a bidomain theory-based heart model to solve the related inverse problem. During right ventricular (RV) pacing control patients showed a deterioration of the ventricular activation sequence similar to the intrinsic activation pattern of CHF patients. Left ventricular propagation velocities were significantly decreased in CHF patients as compared to the control group (1.6±0.4 versus 2.1±0.5 m/sec; pConclusions/significanceEndocardial and epicardial ventricular activation can be visualized noninvasively by NICE. Identification of individual ventricular activation properties may help identify responders to CRT and to further improve response to CRT by facilitating a patient-specific lead placement and device programming.