Software

DEB2

higbyfme — Fri, 23 Aug 2019 18:16:48 +0000

DEB2

higbyfme Fri, 08/23/2019 - 13:16 RSS:

DEB2 is a medication indication and adverse effect knowledgebase derived from five publicly available sources: the VA’s National Drug File-Reference Terminology, MEDLINE, the US Food and Drug Administration’s drug product labels (via the SIDER2 database), the MedlinePlus consumer health information website, and DrugBank, a manually-curated drug target database. All medications, indications, and adverse effects in DEB2 are represented using the RxNorm and SNOMED-CT terminologies.

Indications and adverse effects in DEB2 were extracted from the constituent sources using natural language processing and other ontology methods. The DEB2 knowledgebase requires indications and adverse effects to be corroborated by two or more of its constituent drug information sources, resulting in improved accuracy and precision over the original DEB1, and other popular medication databases.

The DEB2 Dataset contains 6431 indications and 11,539 adverse effects. It includes 1163 distinct drugs and 1601 distinct clinical features (symptoms, findings, diseases, etc.). Based on physician review, DEB2 has an estimated precision of 86% (+/- 3.1%) overall, with indications being slightly more accurate and adverse effects slightly less. Indications and adverse effects for the most commonly-prescribed medications were shown to have a precision of 94% (+/- 3.5%). Drug relationships found in more sources were generally found more likely to be correct.

The DEB2 Full Data Superset is also available for download. It includes all drug-indication and drug-adverse effect pairs found in any of the five sources (including those found in only 1 source and therefore not corroborated elsewhere). The DEB2 Full Data Superset includes over 131,000 pairs (consisting of 31,203 indications and 98,745 adverse effects). While it was not reviewed separately, we suspect the superset should have a higher recall and lower precision than the primary DEB2 Dataset.

DEB2 is available for free download. It is intended for research and is not a substitute for professional clinical advice. We believe this resource can be useful for research in the areas of pharmacovigilance, clinical data mining, clinical phenotyping, and clinical decision support systems, among others. It can be used unaltered, or as a valuable starting point for developing manually-curated indication and adverse effect datasets.

Please cite the reference below when using this resource. Contact joshua.smith@vumc.org with any questions.

*The DEB2 manuscript has been submitted for publication. If this website has not been updated with the correct reference, please contact me for instructions on how to cite DEB2 in your work.

Browse DEB2 online [coming soon]

Download links:

DEB2.txt - The DEB2 knowledgebase, containing medication indications and adverse effects found in at least two sources. Tab-Separated text file; first line contains column names; double-quotes around any field values with commas.

DEB2.xlsx - The DEB2 knowledegbase in Microsoft Excel format. Color-coded for human readability.

DEB2_full_data_superset.txt - The raw data superset from which DEB2 was constructed, containing medication indications and adverse effects found in any of the sources. The vast majority of drug-CF pairs in this file are from only one source. Tab-Separated text file; first line contains column names; double-quotes around any field values with commas.

README.txt - description of data elements in the above files.

DEB2 by Dr. Joshua C. Smith, et al., is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Reference:

Smith JC, Chen Q, Denny JC, Roden DM, Johnson KB, Miller RA. DEB2: An Improved Drug Evidence Base for Biomedical Research. Manuscript submitted for publication. Currently under review.

Tags

phenotyping

Lockdown Auth

AMIA 2015 NLP Application Tutorial

Visitor — Sun, 15 Nov 2015 19:46:35 +0000

AMIA 2015 NLP Application Tutorial

Software

Visitor Sun, 11/15/2015 - 13:46 RSS:

AMIA 2015 NLP Application Tutorial

Below are the slides from the AMIA 2015 NLP Tutorial.

Josh Denny: NLP applied to EHR phenotyping and medical education:

josh amia NLP tutorial 2015.pdf

Hua Xu: Introduction to pipeline NLP and Clinical Language Annotation, Modeling, and Processing Toolkit (CLAMP)

2015_AMIA_NLP_Tutorial_CLAMP.pdf

Lockdown Auth

ICD-9 to PheWAS code map, version 1.2

Visitor — Mon, 10 Aug 2015 19:14:19 +0000

ICD-9 to PheWAS code map, version 1.2

PheWAS, Software

Visitor Mon, 08/10/2015 - 14:14 RSS:

Download the mapping between ICD-9 codes and PheWAS codes, version 1.2 at https://phewascatalog.org/phecodes .

Lockdown Auth

KMCI - KnowledgeMap Concept Indexer

Visitor — Fri, 07 Aug 2015 20:52:37 +0000

KMCI - KnowledgeMap Concept Indexer

Center Resources, Software

Visitor Fri, 08/07/2015 - 15:52 RSS:

The KnowledgeMap Concept Indexer (KMCI) is the underlying natural language processing engine used in the KnowledgeMap and Learning Portfolio website, and has been used for many clinical and genomic research studies. It identifies biomedical concepts, mapped to Unified Medical Language System concepts, from natural language documents and clinical notes.

KMCI employs part-of-speech information to develop a shallow sentence parse, and performs variant generation and normalization using the SPECIALIST Lexicon and related tools. The KMCI system was designed particularly for poorly-formatted documents containing ad hoc abbreviations and underspecified concepts (e.g., the document phrase “ST” implying the “ST segment” of an electrocardiogram instead of abnormal finding “ST elevation”). Using probabilistic information and concept co-occurrence data derived from PubMed, KMCI can map ambiguous strings such as “CHF” to the UMLS concept C0018802 “Congestive heart failure” in an echocardiogram report but to the concept C0009714 “Congenital hepatic fibrosis” in a document discussing infantile polycystic kidney disease (a genetically related condition to congenital hepatic fibrosis).

KMCI has performed favorably in comparison to MetaMap and has been validated in a variety of clinical and education contexts (see publications). Later additions to KMCI include the ability to detect negated terms (e.g., "no chest pain) via a Perl implementation of NegEx.

For further information, please contact Josh Denny

Lockdown Auth

PheWAS R Package

Visitor — Mon, 20 Jul 2015 05:00:00 +0000

PheWAS R Package

PheWAS, Genomics/Pharmacogenomics, Software

Visitor Mon, 07/20/2015 - 00:00 RSS:

This package contains methods for performing PheWAS. Please contact PheWAS@vumc.org. if you encounter any errors or apparent bugs. The documentation is done natively in R. The command ?PheWAS once the package is loaded will direct you to the package description, including references to each function and an example. The command vignette("PheWAS-package") will display the package vignette with further "How to's". It can also be accessed online inside the github here.

If you would like to participate in development or report a bug, you can find the github repository at https://github.com/PheWAS/PheWAS. The newest information about the package is always available there!

There is one reliable method for installing the most up-to-date version of the package. Please use the devtools package you can download here by running the following in R:


install.packages("devtools")
library(devtools)
install_github("PheWAS/PheWAS")
library(PheWAS)

To run the example code in R, one can use the R standard example function after loading the package:


example(PheWAS)

Lockdown Auth

MEDI--an Ensemble MEDication Indication Resource

Visitor — Tue, 01 Jan 2013 06:00:00 +0000

MEDI--an Ensemble MEDication Indication Resource

Genomics/Pharmacogenomics, PheWAS, Center Resources, Software

Visitor Tue, 01/01/2013 - 00:00 RSS:

MEDI (MEDication Indication) is an ensemble medication indication resource for primary and secondary uses of electronic medical record (EMR) data. MEDI was created based on multiple commonly used medication resources (RxNorm, MedlinePlus, SIDER 2, and Wikipedia ) and by leveraging both ontology and natural language processing (NLP) techniques.

MEDI contains 3,112 medications and 63,343 medication-indication pairs. The MEDI high precision subset (MEDI-HPS) includes indications found within either RxNorm or at least two of the three other resources. MEDI-HPS contains ~13,400 unique indication pairs regarding 2,136 medications. The average (±standard deviation) number of indications for each medication in MEDI-HPS is 6.22±6.09. The mode for each medication is 2 while the median is 4. The estimated precision MEDI-HPS is >92% based on physician review.

The download of MEDI is for free. Right click and select the download linked file or similar option in your browser to save the files on your computer.

MEDI.csv

MEDI_HPS.csv (High Precision Set)

MEDI_UMLS.csv (including RxCUI, UMLS, ATC, NDDF concepts and their indications)

MEDI_wPrevalence_Published.csv

MEDI by Dr. Wei-Qi Wei & Dr. Joshua Denny is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.

Key references:
1. WQ Wei, RM Cronin, H Xu, TA Lasko, L Bastarache, JC Denny, Development and evaluation of an ensemble resource linking medications to their indications, Journal of the American Medical Informatics Association. 2013;20:954-961 doi:10.1136/amiajnl-2012-001431
2. WQ Wei, JD Mosley et al. Validation and Enhancement of a Computable Medication Indication Resource (MEDI) Using a Large Practice-based Dataset. AMIA Annual Symposium, 2013, Washington DC.

Lockdown Auth

SecTag -- Tagging Clinical Note Section Headers

Visitor — Sun, 21 Mar 2010 05:00:00 +0000

SecTag -- Tagging Clinical Note Section Headers

Center Resources, Genomics/Pharmacogenomics, Software

Visitor Sun, 03/21/2010 - 00:00 RSS:

Clinical notes are often divided into sections, or segments, such as "history of present illness" or "past medical history." These sections often have subsections as well, such as the "cardiovascular exam" section of the "physical exam." One can gain greater understanding of clinical notes by recognition of the section in which a concept lives. For instance, both a "past medical history" and the "family medical history" sections can contain a list of diseases, but the context decribes very different import to the patient about whom the note was written. Section tagging is an important early step in natural language processing applications applied to clinical notes.

To improve recognition of section headers, we have developed SecTag. SecTag recognizes note section headers using NLP, Bayesian, spelling correction, and scoring techniques. The algorithm can auto-train through multiple iterations on a single corpus.

Since some codes are borrowed from Logical Observation Identifiers Names and Codes (LOINC®), users must have either a valid LOINC or UMLS license:

LOINC license - visit http://loinc.org/downloads
UMLS license - http://www.nlm.nih.gov/databases/umls.html

Download SecTag vocabulary

sec_tag.zip

Primary references:

Denny JC, Spickard A 3rd, Johnson KB, Peterson NB, Peterson JF, Miller RA. Evaluation of a method to identify and categorize section headers in clinical documents.J Am Med Inform Assoc. 2009 Nov-Dec;16(6):806-15.

Denny JC, Miller RA, Johnson KB, Spickard A 3rd. Development and evaluation of a clinical note section header terminology. AMIA Annu Symp Proc. 2008 Nov 6:156-60.

Tags

sectag, NLP

Lockdown Auth