Aigli Korfiati

Non-coding RNA Sequences Identification and Classification Using a Multi-class and Multi-label Ensemble Technique

High throughput sequencing RNA-sequencing technologies and modern in silico techniques have expanded our knowledge on short non-coding RNAs. These sequences were initially split into various categories based on their cellular functionality and their sequential, thermodynamic and structural properties believing that their sequence can be used as an identifier to distinguish them. However, recent evidence has indicated that the same sequences can act and function as more than one type of non-coding RNAs with a striking example of mature microRNA sequences which can also be transfer RNA fragments. Most of the existing computational methods for the prediction of non-coding RNA sequences have emphasized on the prediction of only one type of noncoding RNAs and even the ones designed for multiclassification do not support multiple labeling and are thus not able to assign a sequence to more than one non-coding RNA type. In the present paper, we introduce a new multilabel- multiclass method based on the combination of multiobjective evolutionary algorithms and multi-label implementations of Random Forests to optimize the feature selection process and assign short RNA sequences to one or more non-coding RNA types. The overall methodology clearly outperformed other machine learning techniques which were used for the same purpose and it is applicable to data coming from RNA-sequencing experiments.

On the Computational Prediction of miRNA Promoters

MicroRNAs transcription regulation is an open topic in molecular biology and the identification of the promoters of microRNAs would give us relevant insights on cellular regulatory mechanisms. In the present study, we introduce a new computational methodology for the prediction of microRNA promoters, which is based on the hybrid combination of an adaptive genetic algorithm with a nu-Support Vector Regression (nu-SVR) classifier. This methodology uses genetic algorithms to locate the optimal features set and to optimize the parameters of the nu-SVR classifier. The main advantage of the proposed solution is that it systematically studies and calculates a vast number of features that can be used for promoters prediction including frequency-based properties, regulatory elements and epigenetic features. The proposed method also handles efficiently the issues of over-fitting, feature selection, convergence and class imbalance. Experimental results give accuracy over 87 % in the miRNA promoter prediction.

An asynchronous interface relaxation method for multi-domain/multi-physics problems

An approach for the solution of multi-domain and multi-physics problems is the application of an interface relaxation (IR) method to treat the solution on the common boundaries between domains of the original problem. This solution process is more efficient than other techniques, but still remains quite computationally intensive and the inherently parallel solution of the underlying problems does not scale to the overall method. This paper presents an asynchronous parallel algorithm of a specific IR method, named GEO. The performance results in terms of convergence speed and execution time demonstrate the efficiency of the proposed algorithm towards the solution of large-scale multi-domain and multi-physics problems.

Predicting and classifying short non-coding RNAs using a multiclass evolutionary methodology

High throughput sequencing technologies alongside with advanced bioinformatics methods have uncovered a vast number of short non-coding RNAs. They are split in various categories based on their cellular functionality and their sequential, thermodynamic and structural properties. Existing computational methods emphasize on the prediction of only one type of non-coding RNAs and thus their applicability in full transcriptome studies is limited. Only a few methods so far have been proposed for predicting multiple short non-coding RNAs and they do not cover the most significant classes of them. In the present paper, we introduce a new multiclass method based on the combination of genetic algorithms and support vector machines which is able to distinguish among tRNAs, miRNAs, snoRNAs, rRNAs and other RNA sequences with accuracy over 93%. Finally, the advanced feature selection mechanism of the proposed method uncovers significant characteristics for each one of the studied short non-coding RNAs.

Dynamic Cloud Resources Allocation on Multidomain/Multiphysics Problems

The solution of multidomain/multiphysics problems is a computationally and memory demanding process, especially for large-scale differential equations. In this paper, we propose a cloud application that provides users a solution environment for multiphysics/multidomain problems utilizing cloud technologies that manage pre-existing hardware, network, operating system and applications. Particularly, according to the problem and its computational demands, the user can have the results from any place and any device without any other concern. The user sets the problems parameters, chooses the solution method that fits better to the specific problem and finally gets the problem solution. The application dynamically allocates the minimum possible resources automatically in the background without the users interference.

Workflow Coordinated Resources Allocation for Big Data Analytics in the Cloud

Cloud computing consists of a set of new technologies that permit the dynamic allocation of computational resources (storage, CPU, memory) when performing high demanding data analysis. In the modern world of information data, cloud computing can provide valuable solutions for the Big Data Analytics domain. The correct allocation of resources in a Big Data analysis problem can both increase performance and decrease cost. This article proposes a system architecture for allocating computational resources according to the problem demands in a cloud infrastructure. Workflows are utilized in order to coordinate the execution of complex data analysis pipelines.