Development of new approaches to evaluate protein/amino acid sequences with artificial intelligence

Studies and research on protein or amino acids in the field of bioinformatics are of great importance in terms of understanding the structures of proteins and determining the roles of proteins in cellular activities. When the researches are examined, it is observed that studies related to determining protein functions, predicting interactions between proteins, finding protein secondary structures, classifying protein families, identifying coding and non-coding mRNAs, predicting viral-host interactions and detecting drug-target interactions have been actively conducted. In the thesis, new artificial intelligence-based analysis approaches have been developed for these research areas. In this framework, a new algorithmic-based categorical approach has been developed to map protein sequences. As part of the algorithmic-based approach, three different protein mapping methods have been proposed. Accordingly, the proposed methods of quantifying protein sequences are: • An AVL-based protein mapping method has been proposed to convert protein sequences, and the success of this method has been tested on interactions between proteins of the SARS-CoV-2 virus that causes COVID-19 disease. This method has been evaluated in the algorithmic-based category due to the binary search tree structure it contains. • In another study, entropy-based protein mapping method has been proposed and used to identify protein-protein interactions on viral genomes. Since Shannon entropy is used in this method, it is evaluated in the algorithmic-based category. • Another protein mapping method (FIBHASH), which includes both Fibonacci numbers and hash tables, has been developed and the performance of this method has been evaluated in determining protein families. This method is a hybrid method that includes Fibonacci numbers and hash table. Since the hash table is frequently used in data structures and algorithm analysis, the FIBHASH method is considered an algorithmic-based method. • The proposed protein mapping methods were used to predict the interactions between SARS-CoV-2 virus and host cells, and both the proposed protein mapping methods and the state of art methods in the literature were compared. The successes of artificial intelligence obtained with the proposed protein mapping methods have been promising for new studies in the fields of protein analysis.