Sequence mining consists of finding statistically relevant patterns in data collections represented sequentially. These, are an important type of data, where it matters the order that occupy the elements in the set and that finds a wide range of applications in Bioinformatics and Computational Biology. The prediction of protein structures is one of these applications. Where, a protein is no more than a sequence of amino acids forming patterns known as alpha helices, beta sheets and turns. For purposes of our investigation, these collections or secondary structures would be the itemsets, while the amino acids that make up the entire sequence, the items. Despite multiple attempts to predict protein folding, the algorithms developed to date only reach a 35% effectiveness. That is why we propose SPMCcm, an algorithm based on the prediction of frequent sequences and a scheme of classifiers. Which uses the information provided by the amino acid sequence, in two stages. Where, the first stage learns of the interactions between the secondary structures of the proteins, which it extracts as frequent sequences or itemsets. Meanwhile, the second stage learns of the interaction between the amino acids present in the interacting structures or items.The experimental evaluation showed that SPMCcm behaves in a similar way, independently of the base classifier used, reaching accuracies in the prediction of up to 48%, higher than the 35% reported by the literature, without using large computational resources and possessing explanatory capacity.

Security of the data consumed, generated and stored is crucial to the quality of life in today's society. For this reason, this paper proposes a comparative study of different combination schemes of multiple classifiers based on decision trees, due to its scalability and easy implementation. As a result, precision and recall values of about 97% and 100% were obtained, showing their high reliability, reducing false alarms and high generalization capacity. A comparison with a deep learning based algorithm showed that tree combination strategies are competitive and with statistically similar and superior results to the state-of-the-art. In the end, the results suggest that adaptive strategies such as XGBoost or highly randomized strategies such as Random Forest or Extra-Tree can be alternatives for the protection of precious data on the network.

Understanding the folding of proteins is one of the most interesting research field for the Bioinformatics. The contact maps constitute an intermediate step in the prediction of the 3D structure of the proteins and allow to represent folding patterns. Currently, the methods used to predict contact maps achieve low precision results, only about 25% of long-range (L/5) contacts are correctly predicted, and their knowledge base is not humanly interpretable. In this paper, we propose an easy implementation multiple classifier for contact maps, which is based on patterns of interaction between secondary structures and employed decision trees as base classifiers. This method is able to naturally reduce the level of imbalance between contact/non-contact classes. In addition, a set of interpretable rules are extracted as a complement to the prediction. The validation of method performance shows that an average of 45% of general contacts are correctly predicted. Moreover, a Z-score comparison of its longrange contacts predictions (L/5) with participant methods in CASP11 competition shows that it is competitive with the state-of-the-art methods.