A hybrid genetic-neural system for predicting protein secondary structure

Artificial Neural Network to Predict Varicocele Impact on Male Fertility through Testicular Endocannabinoid Gene Expression Profiles

The relationship between varicocele and fertility has always been a matter of debate because of the absence of predictive clinical indicators or molecular markers able to define the severity of this disease. Even though accumulated evidence demonstrated that the endocannabinoid system (ECS) plays a central role in male reproductive biology, particularly in the testicular compartment, to date no data point to a role for ECS in the etiopathogenesis of varicocele. Therefore, the present research has been designed to investigate the relationship between testicular ECS gene expression and fertility, using a validated animal model of experimental varicocele (VAR), taking advantage of traditional statistical approaches and artificial neural network (ANN). Experimental induction of VAR led to a clear reduction of spermatogenesis in left testes ranging from a mild (Johnsen score 7: 21%) to a severe (Johnsen score 4: 58%) damage of the germinal epithelium. However, the mean number of new-borns recorded after two sequential matings was quite variable and independent of the Johnsen score. While the gene expression of biosynthetic and degrading enzymes of AEA (NAPE-PLD and FAAH, respectively) and of 2-AG (DAGLα and MAGL, respectively), as well as their binding cannabinoid receptors (CB1 and CB2), did not change between testes and among groups, a significant downregulation of vanilloid (TRPV1) expression was recorded in left testes of VAR rats and positively correlated with animal fertility. Interestingly, an ANN trained by inserting the left and right testicular ECS gene expression profiles (inputs) was able to predict varicocele impact on male fertility in terms of mean number of new-borns delivered (outputs), with a very high accuracy (average prediction error of 1%). The present study provides unprecedented information on testicular ECS gene expression patterns during varicocele, by developing a freely available predictive ANN model that may open new perspectives in the diagnosis of varicocele-associated infertility.

Exploiting intrastructure information for secondary structure prediction with multifaceted pipelines

Predicting the secondary structure of proteins is still a typical step in several bioinformatic tasks, in particular, for tertiary structure prediction. Notwithstanding the impressive results obtained so far, mostly due to the advent of sequence encoding schemes based on multiple alignment, in our view the problem should be studied from a novel perspective, in which understanding how available information sources are dealt with plays a central role. After revisiting a well-known secondary structure predictor viewed from this perspective (with the goal of identifying which sources of information have been considered and which have not), we propose a generic software architecture designed to account for all relevant information sources. To demonstrate the validity of the approach, a predictor compliant with the proposed generic architecture has been implemented and compared with several state-of-the-art secondary structure predictors. Experiments have been carried out on standard data sets, and the corresponding results confirm the validity of the approach. The predictor is available at http://iasc.diee.unica.it/ssp2/ through the corresponding web application or as downloadable stand-alone portable unpack-and-run bundle.

Improved prediction of MHC class I binders/non-binders peptides through artificial neural network using variable learning rate: SARS corona virus, a case study

Fundamental step of an adaptive immune response to pathogen or vaccine is the binding of short peptides (also called epitopes) to major histocompatibility complex (MHC) molecules. The various prediction algorithms are being used to capture the MHC peptide binding preference, allowing the rapid scan of entire pathogen proteomes for peptide likely to bind MHC, saving the cost, effort, and time. However, the number of known binders/non-binders (BNB) to a specific MHC molecule is limited in many cases, which still poses a computational challenge for prediction. The training data should be adequate to predict BNB using any machine learning approach. In this study, variable learning rate has been demonstrated for training artificial neural network and predicting BNB for small datasets. The approach can be used for large datasets as well. The dataset for different MHC class I alleles for SARS Corona virus (Tor2 Replicase polyprotein 1ab) has been used for training and prediction of BNB. A total of 90 datasets (nine different MHC class I alleles with tenfold cross validation) have been retrieved from IEDB database for BNB. For fixed learning rate approach, the best value of AROC is 0.65, and in most of the cases it is 0.5, which shows the poor predictions. In case of variable learning rate, of the 90 datasets the value of AROC for 76 datasets is between 0.806 and 1.0 and for 7 datasets the value is between 0.7 and 0.8 and for rest of 7 datasets it is between 0.5 and 0.7, which indicates very good performance in most of the cases.

Prediction of MHC class I binding peptides using probability distribution functions

Binding of peptides to specific Major Histo-compatibility Complex (MHC) molecule is important for understanding immunity and has applications to vaccine discovery and design of immunotherapy. Artificial neural networks (ANN) are widely used by predictions tools to classify the peptides as binders or non-binders (BNB). However, the number of known binders to a specific MHC molecule is limited in many cases, which poses a computational challenge for prediction of BNB and hence, needs improvement in learning of ANN. Here, we describe, the application of probability distribution functions to initialize the weights and biases of the artificial neural network in order to predict HLA-A*0201 binders and non-binders. The 10-fold cross validation has been used to validate the results. It is evident from the results that the A(ROC) for 90% of test cases for Weibull, Uniform and Rayleigh distributions is in the range 0.90-1.0. Further, the standard deviation for AROC was minimum for Weibull distribution, and may be used to train the artificial neural network for HLA-A*0201 MHC Class-I binders and non-binders prediction.

Application of expert networks for predicting proteins secondary structure

The present study utilizes expert neural networks for the prediction of proteins secondary structure. We use three independent networks, one for each structure (alpha, beta and coil) as the first-level processing unit; decision upon the chosen structure for each residue is carried out by a second-level, post-processing unit, which utilizes the Chou and Fasman frequency values Falpha and Fbeta in order to strengthen and/or deplete the probability of the specific structure under investigation. The highest prediction case was 76%. Our method requires primitive computational means and a relatively small training set, while still been comparable to previous work. It is not meant to be an alternative to the determination of secondary structure by means of free energy minimization, integration of dynamic equations of motion or crystallography, which are expensive, time-consuming and complicated, but to provide additional constrains, which might be considered and incorporated into larger computing setups in order to reduce the initial search space for the above methods.

Analysing protein–protein interaction networks of human liver cancer cell lines with diverse metastasis potential

PURPOSE

Human hepatocellular carcinoma (HCC) is one of the most mortal tumor. In a previous study, we had constructed glycoprotein expression profiles and glycoprotein databases of three human liver cancer cell lines with diverse metastasis potential. In order to discover vital glycoproteins related to pathogenesis and metastasis of HCC, in this study we analyzed previous data with bioinformatic approach.

METHODS

We took previous data to draw the protein-protein interaction (PPI) networks of liver cell lines by searching IntACT database and then using Pajeck software. Further more, we compared the differences between the three PPI networks by drawing the PPI networks of differential glycoproteins and by naming differential display PPI networks.

RESULTS

Large numbers of proliferation and apoptosis-relative proteins interact with the differential glycoproteins, and among the differential glycoproteins there are many interactions.

CONCLUSIONS

We conclude that neither single nor several proteins cause malignant proliferation of liver cells. "Molecule groups" concept should be introduced into diagnosis and metastasis prediction of the HCC.

Overview of BITS2005, the Second Annual Meeting of the Italian Bioinformatics Society

The BITS2005 Conference brought together about 200 Italian scientists working in the field of Bioinformatics, students in Biology, Computer Science and Bioinformatics on March 17–19 2005, in Milan. This Editorial provides a brief overview of the Conference topics and introduces the peer-reviewed manuscripts accepted for publication in this Supplement.