Dataset Information

Determining molecular predictors of adverse drug reactions with causality analysis based on structure learning.

ABSTRACT: Adverse drug reaction (ADR) can have dire consequences. However, our current understanding of the causes of drug-induced toxicity is still limited. Hence it is of paramount importance to determine molecular factors of adverse drug responses so that safer therapies can be designed.We propose a causality analysis model based on structure learning (CASTLE) for identifying factors that contribute significantly to ADRs from an integration of chemical and biological properties of drugs. This study aims to address two major limitations of the existing ADR prediction studies. First, ADR prediction is mostly performed by assessing the correlations between the input features and ADRs, and the identified associations may not indicate causal relations. Second, most predictive models lack biological interpretability.CASTLE was evaluated in terms of prediction accuracy on 12 organ-specific ADRs using 830 approved drugs. The prediction was carried out by first extracting causal features with structure learning and then applying them to a support vector machine (SVM) for classification. Through rigorous experimental analyses, we observed significant increases in both macro and micro F1 scores compared with the traditional SVM classifier, from 0.88 to 0.89 and 0.74 to 0.81, respectively. Most importantly, identified links between the biological factors and organ-specific drug toxicities were partially supported by evidence in Online Mendelian Inheritance in Man.The proposed CASTLE model not only performed better in prediction than the baseline SVM but also produced more interpretable results (ie, biological factors responsible for ADRs), which is critical to discovering molecular activators of ADRs.

SUBMITTER: Liu M

PROVIDER: S-EPMC3932464 | biostudies-other | 2014 Mar-Apr

REPOSITORIES: biostudies-other

ACCESS DATA

Publications

Determining molecular predictors of adverse drug reactions with causality analysis based on structure learning.

Liu Mei M Cai Ruichu R Hu Yong Y Matheny Michael E ME Sun Jingchun J Hu Jun J Xu Hua H

Journal of the American Medical Informatics Association : JAMIA 20131211 2

<h4>Objective</h4>Adverse drug reaction (ADR) can have dire consequences. However, our current understanding of the causes of drug-induced toxicity is still limited. Hence it is of paramount importance to determine molecular factors of adverse drug responses so that safer therapies can be designed.<h4>Methods</h4>We propose a causality analysis model based on structure learning (CASTLE) for identifying factors that contribute significantly to ADRs from an integration of chemical and biological p ...[more]

PMID: 24334612

Similar Datasets

Project description:Background and objectivesAdverse drug events (ADEs) are the fifth leading cause of death and thus responsible for a large number of hospital admissions in all over the globe. This study was aimed to assess the antibiotics associated preventability of ADEs and causality of adverse drug reactions (ADRs) among hospitalized patients.MethodsA prospective, cross-sectional, observational study was conducted in four tertiary care public sector hospitals of Lahore, Pakistan. Study population consisted of hospitalized patients who were prescribed with one or more antibiotics. Data were collected between 1st January, 2017 and 31st June, 2017 from 1,249 patients (384 patients aged ≤ 18 years and 865 patients aged >18 years). Schumock and Thornton scale was used to assess the preventability of the ADEs. Medication errors (MEs) that caused preventable ADEs were assessed by MEs tracking form while Naranjo score was used to evaluate the causal relation of ADRs with the antibiotics. Statistical Package for the Social Sciences (IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.) and Microsoft Excel (MS Office, 2010) were used for data analysis.Results2,686 antibiotics were prescribed to 1,249 patients. Among them, fluoroquinolones (11.8%), macrolides (11.6%) and cephalosporins (10.9%) were the most frequently prescribed antibiotics. The most affected organ system by antibiotics associated ADEs was gastrointestinal tract. A total of 486 ADEs were found. The preventability assessment revealed that most of the ADEs (58.4%) were preventable (43.6% of the ADEs were definitely preventable while 14.8% were probably preventable) and caused by MEs including wrong drug (40.1%) and monitoring errors (25.0%), during the stage of physician ordering (22.2%) and patient monitoring (21.1%). The errors were caused due to non-adherence of policies (38.4%) and lack of information about antibiotics (32%). Most of the non-preventable ADEs or ADRs among adults and children were "probable" (35.5%) and "possible" (35.8%), respectively. Logistic regression analysis revealed that ADEs were significantly less among females (OR = 0.047, 95%CI = 0.018-0.121, p-value = <0.001), patients aged 18-52 years (OR = 0.041, 95%CI = 0.013-0.130, p-value = <0.001), tuberculosis patients (OR = 0.304, 95%CI = 0.186-0.497, p-value = <0.001), patients with acute respiratory tract infections (OR = 0.004, 95%CI = 0.01-0.019, p-value = <0.001) and among the patients prescribed with 2 antibiotics per prescription (OR = 0.455, 95%CI = 0.319-0.650, p-value = <0.001).ConclusionAccording to preventability assessment most of the ADEs were definitely preventable and caused by MEs due to non-adherence of policies and lack of information about antibiotics. The causality assessment of non-preventable ADEs showed that most of the ADRs were probable and possible.

Dataset Information

Determining molecular predictors of adverse drug reactions with causality analysis based on structure learning.

Publications

Determining molecular predictors of adverse drug reactions with causality analysis based on structure learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets