Search

Two categories of immune responses—innate and adaptive immunity—have both polygenic backgrounds and a significant environmental component. The goal of the reported study was to define candidate genes and mutations for the immune traits of interest in chickens using machine learning–based sensitivity analysis for single-nucleotide polymorphisms (SNPs) located in candidate genes defined in quantitative trait loci regions. Here the adaptive immunity is represented by the specific antibody response toward keyhole limpet hemocyanin (KLH), whereas the innate immunity was represented by natural antibodies toward lipopolysaccharide (LPS) and lipoteichoic acid (LTA). The analysis consisted of 3 basic steps: an identification of candidate SNPs via feature selection, an optimisation of the feature set using recursive feature elimination, and finally a gene-level sensitivity analysis for final selection of models. The predictive model based on 5 genes (MAPK8IP3 CRLF3, UNC13D, ILR9, and PRCKB) explains 14.9% of variance for KLH adaptive response. The models obtained for LTA and LPS use more genes and have lower predictive power, explaining respectively 7.8 and 4.5% of total variance. In comparison, the linear models built on genes identified by a standard statistical analysis explain 1.5, 0.5, and 0.3% of variance for KLH, LTA, and LPS response, respectively. The present study shows that machine learning methods applied to systems with a complex interaction network can discover phenotype-genotype associations with much higher sensitivity than traditional statistical models. It adds contribution to evidence suggesting a role of MAPK8IP3 in the adaptive immune response. It also indicates that CRLF3 is involved in this process as well. Both findings need additional verification.

Motivation Drug-induced liver injury (DILI) is one of the primary problems in drug development. Early prediction of DILI can bring a significant reduction in the cost of clinical trials. In this work we examined whether occurrence of DILI can be predicted using gene expression profile in cancer cell lines and chemical properties of drugs. Methods We used gene expression profiles from 13 human cell lines, as well as molecular properties of drugs to build Machine Learning models of DILI. To this end, we have used a robust cross-validated protocol based on feature selection and Random Forest algorithm. In this protocol we first identify the most informative variables and then use them to build predictive models. The models are first built using data from single cell lines, and chemical properties. Then they are integrated using Super Learner method with several underlying methods for integration. The entire modelling process is performed using nested cross-validation. Results We have obtained weakly predictive ML models when using either molecular descriptors, or some individual cell lines (AUC ∈(0.55−0.61)). Models obtained with the Super Learner approach have a significantly improved accuracy (AUC=0.73), which allows to divide substances in two categories: low-risk and high-risk.

Additional file 1 Supplementary materials for “Integration of human cell lines gene expression and chemical properties of drugs for Drug Induced Liver Injury prediction”

Super learner algorithm was created to combine results of multiple base learners with the use of cross validation. However, in many cases it does not outperform significantly a simple average of the base results. We propose to apply multiple repeats of cross validation to improve the performance of super learning. Two approaches to application of repeated cross validation were tested on artificial data sets and on real-life, biomedical data sets. One of the approaches, MEAN OUTPUT strategy, proved to significantly improve the results. To reduce the computational complexity of the algorithm, we suggest the use of 3-fold, rather than the previously recommended 10-fold validation. The tests showed, that this simplification does not affect the super learning results.

Motivation: Drug-induced liver injury (DILI) is one of the primary problems in drug development. Early prediction of DILI can bring a significant reduction in the cost of clinical trials. In this work we examined whether occurrence of DILI can be predicted using gene expression profile in cancer cell lines and chemical properties of drugs. Methods: We used gene expression profiles from 13 human cell lines, as well as molecular properties of drugs to build Machine Learning models of DILI. To this end, we have used a robust cross-validated protocol based on feature selection and Random Forest algorithm. In this protocol we first identify the most informative variables and then use them to build predictive models. The models are first built using data from single cell lines, and chemical properties. Then they are integrated using Super Learner method with several underlying methods for integration. The entire modelling process is performed using nested cross-validation. Results: We have obtained weakly predictive ML models when using either molecular descriptors, or some individual cell lines (AUC: 0.55-0.61). Models obtained with the Super Learner approach have a significantly improved accuracy (AUC=0.73), which allows to divide substances in two categories: low-risk and high-risk.

Super learner algorithm can be applied to combine results of multiple base learners to improve quality of predictions. The default method for verification of super learner results is by nested cross validation; however, this technique is very expensive computationally.

Poincaré plot is a return map which can help perform graphical analysis of data. We can also fit an ellipse to the plot shape by determining descriptors SD1, SD2 and SD1/SD2 ratio to study the data quantitatively. In this paper we show examples of application of Poincaré plots in analysis of various kinds of biomedical signals: RR intervals, EMG, gait data and EHG.

Ruby and Perl are programming languages used in many fields. In this paper we would like to present their usefulness with regard to basic bioinformatic problems. We concentrate on a comparison of widely used Perl and relatively rarely used Ruby to show that Ruby can be a very efficient tool in bioinformatics. Both Perl and Ruby have a built-in regular expressions (or regexp) engine, which is essential in solving many problems in bioinformatics. We present some selected examples: printing the file content, removing comments from a FASTA file, using hashes, printing nucleotides included in a sequence, searching for a specific nucleotide in sequence and translating nucleotide sequences into protein sequences obtained in GenBank format. It is our belief that Ruby’s popularity will rise because of its simple syntax and the richness of its methods. Programs in Ruby are very easy to read and therefore easier to maintain and debug, which are the most important characteristics for a programming language.

No-reference image quality assessment (NRIQA) methods estimate image quality degradations without any information about the “perfect-quality” reference image. In this paper, we propose an NRIQA algorithm based on the idea of comparison two blurred variants of the original image to be estimated.

Objective In physiological research, there are not too many studies on multivariate data sets, containing two or more simultaneously recorded time series. It is important to examine synchronization in these kinds of signals. The aim of this study is to present the linear measures: the cross-correlation function, the coherence function, the wavelet cross-correlation and the wavelet coherence to assess synchronization between contractions in different topographic regions of the uterus. Study design Spontaneous uterine activity was recorded directly by a dual micro-tip catheter (Millar Instruments, Inc., USA). The device consisted of two ultra-miniature pressure sensors. One sensor was placed in the fundus, the other one in the cervix. For this analysis, a healthy patient with normal contractions, a patient with dysmenorrhea, a patient with fibromyomas in the follicular phase, and the patient with endometriosis were selected. Results For each method the values of synchronization parameters for normal contractions were higher than the values of these parameters for other pairs of signals. The differences between these four groups of the uterine contraction signals were clear. The lowest values of the synchronization measures were in the case of dysmenorrheic patient. Conclusion The analysis of synchronization of the uterine contractions signals may have a diagnostic value. For intrauterine pressure signals results obtained by means of different synchronization methods are different, but consistent.

For medical images diagnostically important information often lies in the texture. Fractal dimension may be used as an index of irregularity. In this paper we describe the adaptation of the intensity difference scaling method for assessment of the fractal dimension D in the irregular regions of interest (irregular ROIs). This is of great importance because the investigated regions are often small. It is difficult to fit entire regular region of interest within the examined organ with simultaneous inclusion of the relevant fragment, and at the same time to avoid the influence of boundaries. Fractal analysis of various kinds of medical images: panoramic radiography and nuclear medicine scan showed the validity of assessment of D in irregular ROIs.

The fractal dimension D may be calculated in many ways, since its strict definition, the Hausdorff definition is too complicated for practical estimation. In this paper we perform a comparative study often methods of fractal analysis of time series. In Benoit, a commercial program for fractal analysis, five methods of computing fractal dimension of time series (rescaled range analysis, power spectral analysis, roughness-length, variogram methods and wavelet method) are available. We have implemented some other algorithms for calculating D: Higuchi's fractal dimension, relative dispersion analysis, running fractal dimension, method based on mathematical morphology and method based on intensity differences. For biomedical signals results obtained by means of different algorithms are different, but consistent.

In physiological research, we often study multivariate data sets, containing two or more simultaneously recorded time series. The aim of this paper is to present the cross-correlation and the wavelet cross-correlation methods to assess synchronization between contractions in different topographic regions of the uterus. From a medical point of view, it is important to identify time delays between contractions, which may be of potential diagnostic significance in various pathologies. The cross-correlation was computed in a moving window with a width corresponding to approximately two or three contractions. As a result, the running cross-correlation function was obtained. The propagation% parameter assessed from this function allows quantitative description of synchronization in bivariate time series. In general, the uterine contraction signals are very complicated. Wavelet transforms provide insight into the structure of the time series at various frequencies (scales). To show the changes of the propagation% parameter along scales, a wavelet running cross-correlation was used. At first, the continuous wavelet transforms as the uterine contraction signals were received and afterwards, a running cross-correlation analysis was conducted for each pair of transformed time series. The findings show that running functions are very useful in the analysis of uterine contractions.

The analysis of the uterine contraction signals in nonpregnant states gives information about physiological changes during the menstrual cycle. Spontaneous uterine activity was recorded directly by a dual microtip catheter. The device consisted of two ultra-miniature pressure sensors. One sensor was placed in the fundus, the other in the cervix. It was important to identify time delays between contractions in two topographic locations, which may be of potential diagnostic significance in various pathologies: dysmenorrhea, endometriosis, and fecundity disorders. In this study the following synchronization measures-the cross-correlation, the semblance, the mutual information-were used to visualize the time delay changes over time. These measures were computed in a moving window with a width corresponding to approximately two or three contractions. As a result, the running synchronization functions were obtained. The running synchronization functions visualize changes in the propagation of the two simultaneously recorded signals. The propagation% parameter assessed from these functions allows for quantitative description of synchronization. Finally, we illustrate the use of running synchronization functions to investigate the effect of treatment with tamoxifen on primary dysmenorrhea.

The aim of this study is to evaluate synchronization of signals representing uterine contraction activity by means of linear signal processing methods – power spectrum, coherence measure, wavelet power spectrum and wavelet coherence measure. We present results of this analysis for patients with primary dysmenorrhea, uterine myomas, endometriosis and for a healthy woman. Using these methods we have found that primary dysmenorrhea, endometriosis and uterine myomas are accompanied by the uterine contraction disorders. Our analysis will allow more precise diagnostics and better control of treatment.

The irregularity or ”roughness” of medical images is quantified by means of fractal dimension D. For medical images diagnostically important information often lies in the texture. In this paper we describe the application of the intensity difference scaling method for assessment of the fractal dimension D in the irregular regions of interest (irregular ROI-s). Near boundary between different tissues or structures the values of fractal dimensions changed significantly. It is difficult to fit entire regular region of interest within the examined organ with simultaneous inclusion of the relevant fragment, and at the same time to avoid the influence of boundaries.

It was shown earlier that nonlinear processes probably generate uterine contractions. In this study, the nonlinear synchronization measures (the mutual correlation dimension, the cross-approximate entropy, the mutual information and the nonlinear interdependencies) were employed to analyze the association between two time series representing the uterine contraction activity. Here the notion of synchronization is used in a loose sense as the synonym of correlation, the similarity of the signals or the similarity of their dynamics. The signals were recorded from the different topographic regions of the uterus: the cervix and the fundus. The results obtained by means of different algorithms are different but qualitatively similar for the checked methods.