Evaluación del desempeño computacional de sistemas de recomendación aplicado a bases de datos farmacológicas : Sistema de recomendación basado en el conocimiento.

Abstract

The drug repositioning was born as an alternative to the great problems that exist in the drug development process, making it possible to cover diseases that does not have a specific treatment or that are in short supply. We propose the development of a knowledge-based recommendation system to be used in drug repositioning. The ChEMBL pharmacological database version 29 was used, it contains information about activities between compounds and cells. Information about interrelationships between cells was also used, for which an intersection value was considered, it indicates which cells are more related to each other. Using this information, the recommendation system generates activity predictions for relationships between compounds and cells. Tests were performed to find an activity for relationships between compounds and cells in the ChEMBL v30 database that were not found in v29 database, showing 100% true positive results, along with 99.33% true negative results. , these were obtained on a confusion matrix with 305 recommendations. Based on the results, it was confirmed that the knowledge-based recommendation system generated can generate activity recommendations for relationships between compounds and cells that are not yet found in the pharmacological databases used.