UC3M researchers developing a prototype for extracting drug interactions

A pharmacological interaction occurs when the effects of a one drug are modified by the presence of another.  The consequences can be harmful if the interaction causes an increase in the toxicity of the drug or if it diminishes its effect, possibly even causing the patient's death, in the worst case scenario.  At present, there are databases for checking possible interactions between drugs administered to patients, but the main problem is that many of the databases are not updated for three years.  "The biomedical literature is the best system for staying up-to-date with respect to new interactions, but each year 300,000 articles are published just within the pharmacology domain, which is an avalanche of information overwhelming medical personnel", remarked one of the researchers, Isabel Segura, from  the  Bases de Datos Avanzadas  (advanced Database Group) (LABDA) of UC3M.

Once the problem is located, a solution must be sought.  In this case, it is necessary to develop automated tools which offer medical personnel efficient administration and access to all of this information, according to the researchers, who have created a system to detect the names of drugs in biomedical texts with 90% accuracy, according to the scientific article published in the journal Drug Discovery Today by professors Isabel Segura and Paloma Martínez, of the Department of Information Technology at UC3M, and María Segura of the University Hospital of Móstoles. The system combines terminological resources and the rules of nomenclature recommended by the World Health Organization.

"The system we describe permits identification of drug names and their classification within drug families in scientific texts", professor Segura explained. "In our case", she pointed out, "it is a prior step, crucial for locating interactions between drugs from the related literature which facilitates keeping the specialized databases up-to-date". Notwithstanding, the system could have immediate usefulness to improve upon the current search engines, whose functioning is based exclusively on the terms which appear in the documents, adding something of significance. A semantic search engine, for example, could improve the results of the search for a drug family by including the articles which mention the concrete drugs of that family.

At present the researchers are working on developing a prototype for extracting interactions, employing this system of drug recognition and classification. "We have already developed two different prototypes, one based on the use of linguistic patterns and information and a second based on techniques of automatic learning", explained the director of LABDA, Paloma Martínez. "We are evaluating comparing its results with what would be proposed by a specialist and for weighing the results we have manually composed a collection of documents with interactions which we hope to publish with an open license next spring".  However, she also pointed out that in an environment as sensitive as health, it is important to take into account that automated solutions can only facilitate routine tasks and serve as support, while the final decisions are up to the experts.

In today's society, many drugs are consumed and there is a marked trend toward polypharmacy (drugs composed of one or two active ingredients) or polytherapy (the combination of one or two drugs).  Both are influential factors in the occurrence of drug interactions since the more drugs a patient takes, the greater the risk that an adverse reaction is produced.  The main risk groups are persons over the age of 65 and polymedicated patients.  In our country, the percent of the population over age 65 is 11.37%, with an average of 4 prescription drugs prescribed per individual, and that number increases with the patient's age. In addition to affecting the safety of the patient, these pharmacological interactions represent a significant medical cost. 

http://www.uc3m.es/portal/page/portal/actualidad_cientifica/noticias/searching_drug_interact