CoChat

Machine learning-based identification and rule-based normalization of adverse drug reactions in drug labels

Open access
Tiftikci, Mert, Özgür, Arzucan, He, Yongqun, Hur, Junguk · 2019 · BMC Bioinformatics

The DOI matches its Crossref record; 32 of the 33 references it cites were found in a registry or an index, and 1 are not checked yet.

Abstract

Background: Use of medication can cause adverse drug reactions (ADRs), unwanted or unexpected events, which are a major safety concern. Drug labels, or prescribing information or package inserts, describe ADRs. Therefore, systematically identifying ADR information from drug labels is critical in multiple aspects; however, this task is challenging due to the nature of the natural language of drug labels. Results: In this paper, we present a machine learning- and rule-based system for the identification of ADR entity mentions in the text of drug labels and their normalization through the Medical Dictionary for Regulatory Activities (MedDRA) dictionary. The machine learning approach is based on a recently proposed deep learning architecture, which integrates bi-directional Long Short-Term Memory (Bi-LSTM), Convolutional Neural Network (CNN), and Conditional Random Fields (CRF) for entity recognition. The rule-based approach, used for normalizing the identified ADR mentions to MedDRA terms, is based on an extension of our in-house text-mining system, SciMiner. We evaluated our system on the Text Analysis Conference (TAC) Adverse Drug Reaction 2017 challenge test data set, consisting of 200 manually curated US FDA drug labels. Our ML-based system achieved 77.0% F1 score on the task of ADR mention recognition and 82.6% micro-averaged F1 score on the task of ADR normalization, while rule-based system achieved 67.4 and 77.6% F1 scores, respectively. Conclusion: Our study demonstrates that a system composed of a deep learning architecture for entity recognition and a rule-based model for entity normalization is a promising approach for ADR extraction from drug labels.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1186/s12859-019-3195-5

Cites

32 of 33 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Illustration of MedDRA PT and LLT hierarchy with example. A single medical concept is represented by a PT which could be described with many different ways that could have different lexical variants and synonyms. With the help of LLTs, it is pretty beneficial to classify a given text to one of PTs
Fig. 1 Illustration of MedDRA PT and LLT hierarchy with example. A single medical concept is represented by a PT which could be described with many different ways that could have different lexical variants and synonyms. With the help of LLTs, it is pretty beneficial to classify a given text to one of PTsCC BY · PubMed Central
Fig. 2 Overall workflow. This figure illustrates our overall workflow in the present study. Drug labels included in the TAC dataset were analyzed to identify ADRs and normalized them through MedDRA v20. Pre-processing was needed only when the deep learning architecture was used
Fig. 2 Overall workflow. This figure illustrates our overall workflow in the present study. Drug labels included in the TAC dataset were analyzed to identify ADRs and normalized them through MedDRA v20. Pre-processing was needed only when the deep learning architecture was usedCC BY · PubMed Central
Fig. 3 Character representation generation with CNN. This component can only work on the same sized inputs; therefore, inputs are extended with paddings which are inconsequential in the feature extraction. Also, this representation only shows calculation of one filter on the Convolution part, but CNN used in the model have 30 filters
Fig. 3 Character representation generation with CNN. This component can only work on the same sized inputs; therefore, inputs are extended with paddings which are inconsequential in the feature extraction. Also, this representation only shows calculation of one filter on the Convolution part, but CNN used in the model have 30 filtersCC BY · PubMed Central
Fig. 4 Graphical representation of LSTM module for understanding the operations. This represent a single input in a given sequence, sentence in our task. In this representation input is denoted by X which enters from below. Features that are extracted for the given input is represented by h and cell state is represented by c
Fig. 4 Graphical representation of LSTM module for understanding the operations. This represent a single input in a given sequence, sentence in our task. In this representation input is denoted by X which enters from below. Features that are extracted for the given input is represented by h and cell state is represented by cCC BY · PubMed Central
Fig. 5 Bi-LSTM component with variational dropout (depicted by colored & dashed connections). Bi-suffix in the component name stands for the bi-directional which means there exist two identical LSTM modules running on a given input on different directions. Concatenation of extracted features of LSTMs are the output of this component. Intuition behind this is to utilize the information exist in the rest of a given sequence since single LSTM extracts latent information using only elements in the sequence before that one
Fig. 5 Bi-LSTM component with variational dropout (depicted by colored & dashed connections). Bi-suffix in the component name stands for the bi-directional which means there exist two identical LSTM modules running on a given input on different directions. Concatenation of extracted features of LSTMs are the output of this component. Intuition behind this is to utilize the information exist in the rest of a given sequence since single LSTM extracts latent information using only elements in the sequence before that oneCC BY · PubMed Central
Fig. 6 Deep learning model for NER. Complete structure of the architecture to display how described components are put together as a single named entity recognizer
Fig. 6 Deep learning model for NER. Complete structure of the architecture to display how described components are put together as a single named entity recognizerCC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 33 references

Research works printed but not linked: 4

  1. Gurulingappa H, Fluck J, Hofmann-Apitius M, Toldo L. Identification of adverse drug event assertive sentences in medical case reports. In: First international workshop on knowledge discovery and health care management (KD-HCM), European conference on machine learning and principles and practice of knowledge discovery in databases (ECML PKDD); 2011. p. 16–27.
  2. Bird S, Loper E, Klein E. Natural language processing with Python: analyzing text with the natural language toolkit; 2009.
  3. Mikolov T, Sutskever I, Chen K, Corrado GS, Dean J. Distributed representations of words and phrases and their compositionality. In: Advances in neural information processing systems; 2013. p. 3111–9.
  4. Gal Y, Ghahramani Z. A theoretically grounded application of dropout in recurrent neural networks. In: Advances in neural information processing systems; 2016. p. 1019–27

Cited by

See all 24

Metadata

Authors
Tiftikci, M., Özgür, A., He, Y., & Hur, J.
Affiliations
  • Boğaziçi University (TR)
  • University of Michigan (US)
  • University of North Dakota (US)
Corresponding author
Hur, Junguk
Venue
BMC Bioinformatics
Published
2019-12
Volume
20
Issue
S21
Pages
707-707
ISSN
1471-2105 (ISSN-L)
OpenAlex
W2995450470
Cited by
23
References
32
Citations by year
2026: 4 · 2025: 2 · 2024: 4 · 2023: 3 · 2022: 5 · 2021: 3 · 2020: 1 · 2019: 1
FWCI
0.83
Type
journal article
Language
en
Open access
Licence
Locations
  • BMC Bioinformatics (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Deep Blue (University of Michigan) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
National Institute of General Medical Sciences (P20GM113123)
Topics
  • Health Sciences › Medicine › Health Informatics › Biomedical Text Mining and Ontologies
  • Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Toxicology › Pharmacovigilance and Adverse Drug Reactions
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
Keywords
adverse drug reactions, drug labels, prescribing information, entity recognition, text mining, BiLSTM, conditional random fields, FDA drug labels
MeSH terms
Machine Learning, Deep Learning, Drug Labeling, United States, United States Food and Drug Administration, Neural Networks, Computer, Data Mining, Drug-Related Side Effects and Adverse Reactions

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Machine learning-based identification and rule-based normalization of adverse drug reactions in drug labels. https://app.cochat.ai/papers/10.1186/s12859-019-3195-5
The paper
Tiftikci, M., Özgür, A., He, Y., & Hur, J. (2019). Machine learning-based identification and rule-based normalization of adverse drug reactions in drug labels. BMC Bioinformatics. https://doi.org/10.1186/s12859-019-3195-5

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.