Abstract:Identifying drug-drug interactions is an important and challenging problem in computational biology and healthcare research. There are accurate, structured but limited domain knowledge and noisy, unstructured but abundant textual information available for building predictive models. The difficulty lies in mining the true patterns embedded in text data and developing efficient and effective ways to combine heterogenous types of information. We demonstrate a novel approach of leveraging augmented text-mining fea… Show more
“…In contrast, the work by [16] achieved 45% with predictions based on drug metabolism. In terms of accuracy, which measures the percentage of correct predictions combining both the similar and dissimilar predictions, our system comes out at over 80% compared to 69% where drug predictions are based on the relationship between drug targets [18]. To illustrate the conceptual framework of this study, the same model can be used to decide if the drug is suitable for prescription.…”
Section: Resultsmentioning
confidence: 99%
“…For example, [16] has developed a method that combines text mining and automated reasoning to predict enzyme-specific DDIs. [18] also uses text mining techniques to create features based on relevant information such as genes and disease names extracted from drug databases to augment limited domain knowledge. These features are then used to build a logistic regression model to predict drug-drug interaction (DDI).…”
Abstract. With poly-pharmacy becoming more common, it is important for health providers to be aware of the drug profile of patients before prescribing. Although there are many methods on extracting information on drug interactions, they do not integrate with the patients' medical history. This paper describes state-of-the art approaches in extracting the term frequencies of drug properties, and using this knowledge to decide if a drug is suitable for prescription after considering if there is any drug allergy the patient may have and the drugs that the patient is currently taking. An experiment is conducted to evaluate the accuracy of associating the similarity ratio in terms of their term frequencies to the similarity between them. Experimental evaluation of our model yields an accuracy of over 80% which is superior to models that use other methods. Since a drug is to be avoided if it is similar to a drug that patient is allergic to, our model will help dentist decide if a drug is suitable for prescription to the patient. Hence such an approach, when integrated within the clinical workflow will reduce prescription errors thereby increasing the health outcome of the patients.
“…In contrast, the work by [16] achieved 45% with predictions based on drug metabolism. In terms of accuracy, which measures the percentage of correct predictions combining both the similar and dissimilar predictions, our system comes out at over 80% compared to 69% where drug predictions are based on the relationship between drug targets [18]. To illustrate the conceptual framework of this study, the same model can be used to decide if the drug is suitable for prescription.…”
Section: Resultsmentioning
confidence: 99%
“…For example, [16] has developed a method that combines text mining and automated reasoning to predict enzyme-specific DDIs. [18] also uses text mining techniques to create features based on relevant information such as genes and disease names extracted from drug databases to augment limited domain knowledge. These features are then used to build a logistic regression model to predict drug-drug interaction (DDI).…”
Abstract. With poly-pharmacy becoming more common, it is important for health providers to be aware of the drug profile of patients before prescribing. Although there are many methods on extracting information on drug interactions, they do not integrate with the patients' medical history. This paper describes state-of-the art approaches in extracting the term frequencies of drug properties, and using this knowledge to decide if a drug is suitable for prescription after considering if there is any drug allergy the patient may have and the drugs that the patient is currently taking. An experiment is conducted to evaluate the accuracy of associating the similarity ratio in terms of their term frequencies to the similarity between them. Experimental evaluation of our model yields an accuracy of over 80% which is superior to models that use other methods. Since a drug is to be avoided if it is similar to a drug that patient is allergic to, our model will help dentist decide if a drug is suitable for prescription to the patient. Hence such an approach, when integrated within the clinical workflow will reduce prescription errors thereby increasing the health outcome of the patients.
“…For example, Tari et al [9] developed a method that combines text mining and automated reasoning to predict enzyme-specific DDI. Yan et al also used text mining techniques to create features based on relevant information such as genes and disease names extracted from drug databases to augment limited domain knowledge [10]. These features were then used to build a logistic regression model to predict DDI.…”
Section: Drug Interactionsmentioning
confidence: 99%
“…These relations were then mapped with the general knowledge about drug metabolism and interactions to derive the DDI. Just like our work, DrugBank was also used by [10]. However, one of the methods in their preparation of data was to represent each drug by a vector of drug targets.…”
Section: Baseline Modelsmentioning
confidence: 99%
“…Table 4 shows the results based on the aggregated similarity ratio obtained from the normalised feature vectors. In terms of accuracy, the percentage of correct predictions combining both the similar and dissimilar predictions, our system achieved 76% compared to 69% where drug predictions were based on the relationship between drug targets [10]. Table 5 shows the results of the experiment trained using the word embeddings approach.…”
Section: Experiments Using Term Similaritymentioning
The rapid increase in the volume and variety of data poses a challenge to safe drug prescription for the dentist. The increasing number of patients that take multiple drugs further exerts pressure on the dentist to make the right decision at point-of-care. Hence, a robust decision support system will enable dentists to make decisions on drug prescription quickly and accurately. Based on the assumption that similar drug-pairs have a higher similarity ratio, this paper suggests an innovative approach to obtain the similarity ratio between the drug that the dentist is going to prescribe and the drug that the patient is currently taking. We conducted experiments to obtain the similarity ratios of both positive and negative drug-pairs, by using feature vectors generated from term similarities and word embeddings of bio-medical text corpus. This model can be easily adapted and implemented for use in a dental clinic to assist the dentist in deciding if a drug is suitable for prescription, taking into consideration the medical profile of the patients. Experimental evaluation of our model's association of the similarity ratio between two drugs yielded a superior F score of 89%.Hence, such an approach, when integrated within the clinical work-flow, will reduce prescription errors and thereby increase the health outcomes of patients.
Rapid increases in data volume and variety pose a challenge to safe drug prescription for health professionals like doctors and dentists. This is addressed by our study, which presents innovative approaches in mining data from drug corpus and extracting feature vectors to combine this knowledge with individual patient medical profiles. Within our three-tiered frameworkthe prediction layer, the knowledge layer and the presentation layer-we describe multiple approaches in computing similarity ratios from the feature vectors, illustrated with an example of applying the framework in a typical medical clinic. Experimental evaluation shows that the word embedding model performs better than the adverse network model, with a F score of 0.75. The F score is a common metrics used for evaluating the performance of classification algorithms. Similarity to a drug the patient is allergic to or is taking are important considerations for the suitability of a drug for prescription. Hence, such an approach, when integrated within the clinical work-flow, will reduce prescription errors thereby increasing patient health outcomes. Keywords Feature vector • Similarity ratio • Word embedding • Adverse network model • Personalised drug prescription This work is partially supported by Glory Dental Surgery Pte Ltd and undertaken collaboratively with their panel of dentists. We would like to thank Dr. Xueling Oh and Dr. Elizabeth Goh for enriching the authors understanding of drug prescription within the medical domain.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.