Abordagens computacionais para identificação e classificação de toxinas em aracnídeos e predição de peptídeos bioativos antivirais

Abstract

The process of discovering and developing new drugs costs billions of dollars and can take more than ten years. The search for new drugs begins with some form of screening of libraries of chemical compounds that often contain a variety of natural products. Venoms contain toxins with extremely specific functions, therefore being a secretion of great potential in the search for new bioactive compounds. In the era of Next Generation Sequencing and shotgun proteomics, a large number of data of toxins and venom protein sequences from venomous animals have been generated. Thus, there is great interest in applying computational methodologies to facilitate parts of the screening of peptide and proteins for biological activity. To assist in the characterization and search for new toxins and antiviral peptides in arachnid venoms, two computational methods were developed: ArachnoFamTox and EnAVPClass. ArachnoFamTox performs the prediction and classification of arachnid toxins and venom proteins with an approach that uses evolutionary conservation information with Position Specific Scoring Matrices and Hidden Markov Models and is the most specific predictor of toxins among the existing methods and one of the few that classify toxins in families. EnAVPClass predicts antiviral peptides and classifies them according to their mechanism of action. The method uses an approach based on Supervised Machine Learning with Random Forest and Support Vector Machines models and Deep Learning with Long Short Term Memory neural networks. As an example of application of the developed tools, transcriptomes of seven tick species were processed and assembled de novo and the ArachnoFamTox and EnAVPClass tools were used for the prediction and classification of toxins and prediction of Antivirals. A total of 2.069 toxins and venom proteins classified in 13 different families were identified and 47.160 peptides were generated, of which 3.559 antiviral peptides (7.54%) were predicted. Of these, 1.765 were identified as Antivirals that act on virus membranes, 755 on viral replication and 64 on viral assembly. The results show the importance of ticks as promising organisms in the discovery of new antiviral compounds based on their saliva and venom proteins. The tools developed in this work will be available as packages open source in https://github.com/yutakajr after publication and will help in the process of discovering new drugs in silico, classifying families of toxins in arachnids and in the development of new tools.