Search

Knowing the function of proteins is essential for understanding several biological systems. The experiments in laboratory to determine protein class are costly and require a long time to be done. Therefore, it is necessary to provide efficient computational models to identify the class to which a protein belongs. Nowadays, a significant volume of information regarding proteins and their structure is continually being made available in public data repositories. For example, the STING_DB database has a lot of information extracted from all protein structural levels (primary, secondary, tertiary, and quaternary), which are frequently used in classification models for this type of problem. However, it is unknown which physical-chemical properties are the most relevant ones to contribute to the prediction of the class. Therefore, there is a need to identify the subset of more suitable properties. In this work, we propose an approach based on a multi-objective genetic algorithm with the classifier k-NN to select the best physical-chemical properties. Our strategy uses a multi-objective genetic algorithm to obtain a smaller subset of features that contribute significantly to the prediction problem. To improve the prediction’s performance, we choose to perform a post enrichment process, then we compare the performance of our methodology with several classifiers: ANN, SVM, Random Forest, and k-NN. Our method achieved an average F-measure value of 70.22% with the Random Forest classifier. Finally, a comparative analysis, with statistical significance, shows the relevance of our approach in relation to other methodologies.

Fundo Mackenzie de Pesquisa Society currently seeks competitiveness for its business, high performance and a low cost support platform, idealizing the contemporary scenario where the world is interconnected by communication networks. Thus the need for optical networks arises because of its advantages of reaching long distances and high speeds with good bandwidth compared to the wired system, but the optical system itself is limited in its resources favoring the need for research to soften the data. damage caused by distortion of optical signals. The purpose of this paper will use machine learning and artificial intelligence techniques to construct a conceptual model capable of predicting signal distortions in optical link designs and their regeneration and thus ensuring their autonomous optimization, with the aim of reducing the project cost of implementing fiber optic links. The computational potential has increased in the last decades favoring the execution of machine learning algorithms and promoting the conditions for this work. It will be made a comparative analysis between three algorithms already used in the literature in optical communications application seeking to find the most suitable algorithm for the construction of this machine learning model that needs a composite output in its predictions, considering the range of variables necessary to elaborate a optical link. The results presented are motivating, showing a high accuracy of predictions of machine learning algorithms around 99% and in the validation of predictions made an optimized link with a BER 1.10−06 evidencing the application of machine learning algorithms in the projects of optical links. A sociedade atualmente busca competitividade para seus negócios, alta performance e uma plataforma de sustentação com baixo custo, isso idealiza o cenário contemporâneo onde o mundo está interconectado por redes de comunicação. Surge com isso a necessidade das redes ópticas pelas suas vantagens de alcançar grandes distância e altas velocidade com uma boa largura de banda comparado com o sistema cabeado, mas o próprio sistema óptico é limitado em seus recursos favorecendo a necessidade de pesquisas para amenizar os danos causados pelas distorções dos sinais ópticos. A proposta do presente trabalho utilizará as técnicas de aprendizagem de máquina e inteligência artificial para construir um modelo conceitual capaz de predizer distorções de sinais nos projetos de enlaces ópticos e sua regeneração e com isso garanta sua otimização autônoma, com o objetivo de reduzir o custo do projeto de implementação de enlaces de fibra óptica. O potencial computacional tem aumentado nas ultimas décadas favorecendo a executabilidade dos algoritmos de machine learning e promovendo as condições para realização deste trabalho. Será feita uma análise comparativa entre três algoritmos já utilizados na literatura em aplicacão de comunicações ópticas procurando encontrar o algoritmo mais adequado para construção desse modelo de machine learning que necessita de uma saída composta em suas predições, visto a abrangência de variáveis necessárias para elaboração de um enlace óptico. Os resultados apresentados, são motivadores, mostrando uma acurácia de predições alta dos algoritmos de machine learning em torno de 99% e a na validação das predições realizadas um enlace otimizado com uma BER 1.10−06 evidenciando a aplicação dos algoritmos de machine learning nos projetos de enlaces ópticos. Palavras

Depression is the leading cause of disability in the world and, according to the World Health Organization, is the leading cause of illness and disability in the adolescence. Previous research has found that half of all people who developed mental disorders had their first symptoms up to age 14. However, what are the most important characteristics to determinate whether or not someone has depression? Genetic algorithms have been used quite efficiently in the problem of feature selection from a dataset. Thus, the main objective of this work is to use genetic algorithms to search for the most relevant features to improve the performance of classifiers to assist in the diagnosis of depression in children and adolescents. The dataset used in this work contains information on 166 children and adolescents between 10 and 16 years of age, of whom 67 are males and 99 are females with different depressive symptoms. The genetic algorithm found, in its best result, a set of 55 features out of the 112 in the dataset. The feature referring to how one feels with her/his own appearance was the most used in the solutions returned by the genetic algorithms. The use of the dataset composed of the 55 features selected by the genetic algorithms improved the performance of the classifiers by 6 to 20 percentage points, reaching f-measure values between 90% and 96%, which is a very significant result in the depression diagnosis context.

Knowing the function of proteins is essential in several areas such as bioinformatics, agriculture, and others. The processes to determine protein function that is realized in laboratories are costly and require a long time to be done. Therefore, it is necessary to provide efficient computational models that aim to find the function of a protein. There are currently several kinds of researches that deal with the prediction problem of protein function. However, each of them presents a different methodology, employing different classifiers as well. Based on this problem, we propose a methodology using a multi-objective genetic algorithm with the classifier k-NN to select the best characteristics and then apply several classifiers such as Artificial Neural Network, SVM, Random Forest, and k-NN, in order to compare their performance in the same methodology. Our methodology found the best performance to be the k-NN classifier, with an f-Measure value of 86.07%.

With the recent advances of genomic sequencing techniques, the number of protein sequences available for analysis has increased greatly. Wet-lab processes to predict the function of a protein are too high-cost to answer to this demand. However, knowing the function of a protein is extremely important in several fields such as medicine and agriculture. Therefore, it is necessary to find computational models able to predict protein function. That is an open research field in Bioinformatics, since the existing models don't perform well enough yet. The Sting DB database gathers relevant informations related to the proteins' physico-chemical characteristics, which have been considered in some researches, most of those limited to a few set of characteristics. In this work, we proposed a methodology utilizing a multi-objective genetic algorithm to find the ideal subset of characteristics to identify the classes of a dataset of enzymes. After the feature selection process, we performed a dataset enhancement by adding new variables, in order to construct a SVM classifier. The proposed methodology achieved 77.3% precision and 72.7% F-Measure averages.