Publication: Application of information extraction techniques to pharmacological domain : extracting drug-drug interactions
Loading...
Identifiers
Publication date
2010
Defense date
2010-04-23
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Una interacción farmacológica ocurre cuando los efectos de un fármaco se modifican por la presencia de otro. Las consecuencias pueden ser perjudiciales si la interacción causa un aumento de la toxicidad del fármaco o la disminución de su efecto, pudiendo provocar incluso la muerte del paciente en los peores casos. Las interacciones farmacológicas no sólo suponen un grave problema para la seguridad del paciente, sino que además también conllevan un importante incremento en el gasto médico. En la actualidad, el personal sanitario tiene a su disposición diversas bases de datos sobre interacciones que permiten evitar posibles interacciones a la hora de prescribir un determinado tratamiento, sin embargo, estas bases de datos no están completas. Por este motivo, médicos y farmacéuticos se ven obligados a revisar una gran cantidad de artículos científicos e informes sobre seguridad de medicamentos para estar al día de todo lo publicado en relación al tema. Desgraciadamente, el gran volumen de información al respecto hace que estos profesionales estén desbordados ante tal avalancha. El desarrollo de métodos automáticos que permitan recopilar, mantener e interpretar toda esta información es crucial a la hora de conseguir una mejora real en la detección temprana de las interacciones entre fármacos. Por tanto, la extracción de información podría reducir el tiempo empleado por el personal médico en la revisión de la literatura médica. Sin embargo, la extracción de interacciones farmacológicas a partir textos biomédicos no ha sido dirigida hasta el momento. Motivados por estos aspectos, en esta tesis hemos realizado un estudio detallado sobre diversas técnicas de extracción de información aplicadas al dominio farmacológico. Basándonos en este estudio, hemos propuesto dos aproximaciones distintas para la extracción de interacciones farmacológicas de los textos. Nuestra primera aproximación propone un enfoque híbrido, que combina análisis sintáctico superficial y la aplicación de patrones léxicos definidos por un farmacéutico. La segunda aproximación se aborda mediante aprendizaje supervisado, concretamente, el uso de métodos kernels. Además, se han desarrollado las siguientes tareas auxiliares: (1) el análisis de los textos utilizando la herramienta UMLS MetaMap Transfer (MMTx), que proporciona información sintáctica y semántica, (2) un proceso para identificar y clasificar los nombres de fármacos que ocurren en los textos, y (3) un proceso para reconoger las expresiones anafóricas que se refieren a fármacos. Un prototipo ha sido desarrollado para integrar y combinar las distintas técnicas propuestas en esta tesis. Para la evaluación de las dos propuestas, con la ayuda de un farmacéutico desarrollamos y anotamos un corpus con interacciones farmacológicas. El corpus DrugDDI es una de las principales aportaciones de la tesis, ya que es el primer corpus en el dominio biomédico anotado con este tipo de información y porque creemos que puede alentar la investigación sobre extracción de información en el dominio farmacológico. Los experimentos realizados demuestran que el enfoque basado en kernels consigue mejores resultados que los reportados por el enfoque que utiliza información sintáctica y patrones léxicos. Además, los kernels consiguen resultados comparables a los obtenidos en dominios similares como son las interacciones entre proteínas. Esta tesis se ha llevado a cabo en el marco del consorcio de investigación MAVIRCM (Mejorando el acceso y visibilidad de la información multilingüe en red para la Comunidad de Madrid, www.mavir.net) dentro del Programa de Actividades de I+D en Tecnologías 2005-2008 de la Comunidad de Madrid (S-0505/TIC-0267) así como en el proyecto de investigación BRAVO: ”Búsqueda de Respuestas Avanzada Multimodal y Multilingüe” (TIN2007-67407-C03-01).----------------------------------------------------------------------------------------
A drug-drug interaction occurs when one drug influences the level or activity of another drug. The detection of drug interactions is an important research area in patient safety since these interactions can become very dangerous and increase health care costs. Although there are different databases supporting health care professionals in the detection of drug interactions, this kind of resource is rarely complete. Drug interactions are frequently reported in journals of clinical pharmacology, making medical literature the most effective source for the detection of drug interactions. However, the increasing volume of the literature overwhelms health care professionals trying to keep an up-to-date collection of all reported drug-drug interactions. The development of automatic methods for collecting, maintaining and interpreting this information is crucial for achieving a real improvement in their early detection. Information Extraction (IE) techniques can provide an interesting way of reducing the time spent by health care professionals on reviewing the literature. Nevertheless, no approach has been carried out to extract drug-drug interactions from biomedical texts. In this thesis, we have conducted a detailed study on various IE techniques applied to biomedical domain. Based on this study, we have proposed two different approximations for the extraction of drug-drug interactions from texts. The first approximation proposes a hybrid approach, which combines shallow parsing and pattern matching to extract relations between drugs from biomedical texts. The second approximation is based on a supervised machine learning approach, in particular, kernel methods. In addition, we have created and annotated the first corpus, DrugDDI, annotated with drug-drug interactions, which allow us to evaluate and compare both approximations. To the best of our knowledge, the DrugDDI corpus is the only available corpus annotated for drug-drug interactions and this thesis is the first work which addresses the problem of extracting drug-drug interactions from biomedical texts. We believe the DrugDDI corpus is an important contribution because it could encourage other research groups to research into this problem. We have also defined three auxiliary processes to provide crucial information, which will be used by the aforementioned approximations. These auxiliary tasks are as follows: (1) a process for text analysis based on the UMLS MetaMap Transfer tool (MMTx) to provide shallow syntactic and semantic information from texts, (2) a process for drug name recognition and classification, and (3) a process for drug anaphora resolution. Finally, we have developed a pipeline prototype which integrates the different auxiliary processes. The pipeline architecture allows us to easily integrate these modules with each of the approaches proposed in this thesis: pattern-matching or kernels. Several experiments were performed on the DrugDDI corpus. They show that while the first approximation based on pattern matching achieves low performance, the approach based on kernel-methods achieves a performance comparable to those obtained by approaches which carry out a similar task such as the extraction of protein-protein interactions. This work has been partially supported by the Spanish research projects: MAVIR consortium (S-0505/TIC-0267, www.mavir.net), a network of excellence funded by the Madrid Regional Government and TIN2007-67407-C03-01 (BRAVO: Advanced Multimodal and Multilingual Question Answering).
A drug-drug interaction occurs when one drug influences the level or activity of another drug. The detection of drug interactions is an important research area in patient safety since these interactions can become very dangerous and increase health care costs. Although there are different databases supporting health care professionals in the detection of drug interactions, this kind of resource is rarely complete. Drug interactions are frequently reported in journals of clinical pharmacology, making medical literature the most effective source for the detection of drug interactions. However, the increasing volume of the literature overwhelms health care professionals trying to keep an up-to-date collection of all reported drug-drug interactions. The development of automatic methods for collecting, maintaining and interpreting this information is crucial for achieving a real improvement in their early detection. Information Extraction (IE) techniques can provide an interesting way of reducing the time spent by health care professionals on reviewing the literature. Nevertheless, no approach has been carried out to extract drug-drug interactions from biomedical texts. In this thesis, we have conducted a detailed study on various IE techniques applied to biomedical domain. Based on this study, we have proposed two different approximations for the extraction of drug-drug interactions from texts. The first approximation proposes a hybrid approach, which combines shallow parsing and pattern matching to extract relations between drugs from biomedical texts. The second approximation is based on a supervised machine learning approach, in particular, kernel methods. In addition, we have created and annotated the first corpus, DrugDDI, annotated with drug-drug interactions, which allow us to evaluate and compare both approximations. To the best of our knowledge, the DrugDDI corpus is the only available corpus annotated for drug-drug interactions and this thesis is the first work which addresses the problem of extracting drug-drug interactions from biomedical texts. We believe the DrugDDI corpus is an important contribution because it could encourage other research groups to research into this problem. We have also defined three auxiliary processes to provide crucial information, which will be used by the aforementioned approximations. These auxiliary tasks are as follows: (1) a process for text analysis based on the UMLS MetaMap Transfer tool (MMTx) to provide shallow syntactic and semantic information from texts, (2) a process for drug name recognition and classification, and (3) a process for drug anaphora resolution. Finally, we have developed a pipeline prototype which integrates the different auxiliary processes. The pipeline architecture allows us to easily integrate these modules with each of the approaches proposed in this thesis: pattern-matching or kernels. Several experiments were performed on the DrugDDI corpus. They show that while the first approximation based on pattern matching achieves low performance, the approach based on kernel-methods achieves a performance comparable to those obtained by approaches which carry out a similar task such as the extraction of protein-protein interactions. This work has been partially supported by the Spanish research projects: MAVIR consortium (S-0505/TIC-0267, www.mavir.net), a network of excellence funded by the Madrid Regional Government and TIN2007-67407-C03-01 (BRAVO: Advanced Multimodal and Multilingual Question Answering).
Description
Keywords
Gestión de información, Técnicas de búsqueda y uso de información, Interacciones farmacológicas, Drug interactions, Information extracting techniques