A neural named entity recognition and multi-type normalization tool for biomedical text mining
BERN is a BioBERT-based multi-type NER tool that also supports normalization of extracted entities. This repository contains the official implementation of BERN. You can use BERN at https://bern.korea.ac.kr, or host your own server by following the description below. Please refer to our paper (Kim et al., IEEE Access 2019) for more details. This project is done by DMIS Laboratory at Korea University.
[Updates]
* Check out BERN2, an improved version of BERN with much faster and more accurate inference! *
Fixed our gene normalizer to respond to issues between 2020-03-12 and 2020-03-13
Done - Server down due to air conditioning problems in our server room 2019-10-10 - 2019-10-11 7:55 AM (UTC-0)
Fixed our disease normalizer 2019-08-19, 2019-08-10 and 2019-08-02 issues
Done - Server check 2019-07-18 8:20 AM - 1:30 PM (UTC-0)
Overview of BERN.
The description below gives instructions on hosting your own BERN. Please refer to https://bern.korea.ac.kr for the RESTful Web service of BERN.
Note that you will need at least 66 GB of free disk space and 32 GB or more RAM.
Clone this repo
cdgit clone https://github.com/dmis-lab/bern.git
Install python packages
pip3 install -r requirements.txt --user
Install GNormPlus & run GNormPlusServer.jar
cd GNormPlusJava
wget -O ./crfpp-0.58.tar.gz https://drive.google.com/uc?id=0B4y35FiV1wh7QVR6VXJ5dWExSTQ
tar xvfz crfpp-0.58.tar.gz
cp -rf CRF++-0.58/* CRF
cd CRF
sh ./configure
make
sudo make install
cd ..
chmod 764 Ab3P
sed -i ‘s/= All/= 9606/g’ setup.txt; echo “FocusSpecies: from All to 9606 (Human)”
sh Installation.sh
rm -r CRF++-0.58
rm crfpp-0.58.tar.gz
wget —load-cookies /tmp/cookies.txt “https://docs.google.com/uc?export=download&confirm=$(wget —quiet —save-cookies /tmp/cookies.txt —keep-session-cookies —no-check-certificate ‘https://docs.google.com/uc?export=download&id=1g-JlhqeDIlZX5YFk8Y27_M8BXUXcQRSX‘ -O- | sed -rn ‘s/.confirm=([0-9A-Za-z_]+)./\1\n/p’)&id=1g-JlhqeDIlZX5YFk8Y27_M8BXUXcQRSX” -O GNormPlusServer.jar && rm -rf /tmp/cookies.txt
nohup java -Xmx16G -Xms16G -jar GNormPlusServer.jar 18895 >> ~/bern/logs/nohup_gnormplus.out 2>&1 &
* Install tmVar2 & run tmVar2Server.jar
cd ~/bern
wget ftp://ftp.ncbi.nlm.nih.gov/pub/lu/Suppl/tmVar2/tmVarJava.zip
unzip tmVarJava.zip
cd tmVarJava
wget -O ./crfpp-0.58.tar.gz https://drive.google.com/uc?id=0B4y35FiV1wh7QVR6VXJ5dWExSTQ
tar xvfz crfpp-0.58.tar.gz
cp -rf CRF++-0.58/* CRF
cd CRF
sh ./configure
make
sudo make install
cd ..
chmod 764 CRF/crf_test
sh Installation.sh
rm -r CRF++-0.58
rm crfpp-0.58.tar.gz
wget —load-cookies /tmp/cookies.txt “https://docs.google.com/uc?export=download&confirm=$(wget —quiet —save-cookies /tmp/cookies.txt —keep-session-cookies —no-check-certificate ‘https://docs.google.com/uc?export=download&id=1kQYzLHLFLsU9qKpRRGjXkIYmaYK6bPJm‘ -O- | sed -rn ‘s/.confirm=([0-9A-Za-z_]+)./\1\n/p’)&id=1kQYzLHLFLsU9qKpRRGjXkIYmaYK6bPJm” -O tmVar2Server.jar && rm -rf /tmp/cookies.txt
wget https://repo1.maven.org/maven2/org/xerial/sqlite-jdbc/3.20.0/sqlite-jdbc-3.20.0.jar
wget https://repo1.maven.org/maven2/edu/stanford/nlp/stanford-corenlp/3.5.2/stanford-corenlp-3.5.2.jar
nohup java -Xmx8G -Xms8G -jar tmVar2Server.jar 18896 >> ~/bern/logs/nohup_tmvar.out 2>&1 &
* Download normalization resources and pre-trained BioBERT NER models
cd ~/bern/scripts
sh download_norm.sh
sh download_biobert_ner_models.sh
* Run named entity normalizers
cd ..
sh load_dicts.sh
* Run BERN server
echo $CUDA_VISIBLE_DEVICES
export CUDA_VISIBLE_DEVICES=0
nohup python3 -u server.py —port 8888 —gnormplus_home ~/bern/GNormPlusJava —gnormplus_port 18895 —tmvar2_home ~/bern/tmVarJava —tmvar2_port 18896 >> logs/nohup_BERN.out 2>&1 &
tail -F logs/nohup_BERN.out
* Usage* PMID(s) (HTTP GET)* http://\<YOUR_SERVER_ADDRESS>:8888/?pmid=\<a PMID or comma seperate PMIDs>&format=\<json or pubtator>* Example: http://\<YOUR_SERVER_ADDRESS>:8888/?pmid=30429607&format=json&indent=true* Example: http://\<YOUR_SERVER_ADDRESS>:8888/?pmid=30429607&format=pubtator* Example: http://\<YOUR_SERVER_ADDRESS>:8888/?pmid=30429607,29446767&format=json&indent=true* Raw text (HTTP POST)* POST Address: http://\<YOUR_SERVER_ADDRESS>:8888* Set key, value of a body as follows:
import requestsimport jsonbody_data = {"param": json.dumps({"text": "CLAPO syndrome: identification of somatic activating PIK3CA mutations and delineation of the natural history and phenotype. PURPOSE: CLAPO syndrome is a rare vascular disorder characterized by capillary malformation of the lower lip, lymphatic malformation predominant on the face and neck, asymmetry, and partial/generalized overgrowth. Here we tested the hypothesis that, although the genetic cause is not known, the tissue distribution of the clinical manifestations in CLAPO seems to follow a pattern of somatic mosaicism. METHODS: We clinically evaluated a cohort of 13 patients with CLAPO and screened 20 DNA blood/tissue samples from 9 patients using high-throughput, deep sequencing. RESULTS: We identified five activating mutations in the PIK3CA gene in affected tissues from 6 of the 9 patients studied; one of the variants (NM_006218.2:c.248T>C; p.Phe83Ser) has not been previously described in developmental disorders. CONCLUSION: We describe for the first time the presence of somatic activating PIK3CA mutations in patients with CLAPO. We also report an update of the phenotype and natural history of the syndrome."})}response = requests.post('http://<YOUR_SERVER_ADDRESS>:8888', data=body_data)result_dict = response.json()print(result_dict)```
[
{
“denotations”: [
{
“id”: [
“MESH:C567763”,
“BERN:262813101”
],
“obj”: “disease”,
“span”: {
“begin”: 0,
“end”: 13
}
},
{
“id”: [
“MIM:171834”,
“HGNC:8975”,
“Ensembl:ENSG00000121879”,
“BERN:324295302”
],
“obj”: “gene”,
“span”: {
“begin”: 53,
“end”: 58
}
},
{
“id”: [
“MESH:C567763”,
“BERN:262813101”
],
“obj”: “disease”,
“span”: {
“begin”: 133,
“end”: 146
}
},
{
“id”: [
“MESH:D014652”,
“BERN:256572101”
],
“obj”: “disease”,
“span”: {
“begin”: 158,
“end”: 174
}
},
{
“id”: [
“MESH:C567763”,
“BERN:262813101”
],
“obj”: “disease”,
“span”: {
“begin”: 193,
“end”: 231
}
},
{
“id”: [
“MESH:C567763”,
“BERN:262813101”
],
“obj”: “disease”,
“span”: {
“begin”: 234,
“end”: 288
}
},
{
“id”: [
“MESH:C567763”,
“BERN:262813101”
],
“obj”: “disease”,
“span”: {
“begin”: 589,
“end”: 593
}
},
{
“id”: [
“MIM:171834”,
“HGNC:8975”,
“Ensembl:ENSG00000121879”,
“BERN:324295302”
],
“obj”: “gene”,
“span”: {
“begin”: 748,
“end”: 758
}
},
{
“id”: [
“CUI-less”
],
“mutationType”: “ProteinMutation”,
“normalizedName”: “p.F83S;CorrespondingGene:5290”,
“obj”: “mutation”,
“span”: {
“begin”: 857,
“end”: 866
}
},
{
“id”: [
“BERN:257523801”
],
“obj”: “disease”,
“span”: {
“begin”: 906,
“end”: 928
}
},
{
“id”: [
“CUI-less”
],
“obj”: “gene”,
“span”: {
“begin”: 1009,
“end”: 1024
}
},
{
“id”: [
“MESH:C567763”,
“BERN:262813101”
],
“obj”: “disease”,
“span”: {
“begin”: 1043,
“end”: 1047
}
}
],
“elapsed_time”: {
“ner”: 0.611,
“normalization”: 0.218,
“tmtool”: 1.281,
“total”: 2.111
},
“project”: “BERN”,
“sourcedb”: “PubMed”,
“sourceid”: “29446767”,
“text”: “CLAPO syndrome: identification of somatic activating PIK3CA mutations and delineation of the natural history and phenotype. PURPOSE: CLAPO syndrome is a rare vascular disorder characterized by capillary malformation of the lower lip, lymphatic malformation predominant on the face and neck, asymmetry, and partial/generalized overgrowth. Here we tested the hypothesis that, although the genetic cause is not known, the tissue distribution of the clinical manifestations in CLAPO seems to follow a pattern of somatic mosaicism. METHODS: We clinically evaluated a cohort of 13 patients with CLAPO and screened 20 DNA blood/tissue samples from 9 patients using high-throughput, deep sequencing. RESULTS: We identified five activating mutations in the PIK3CA gene in affected tissues from 6 of the 9 patients studied; one of the variants (NM_006218.2:c.248T>C; p.Phe83Ser) has not been previously described in developmental disorders. CONCLUSION: We describe for the first time the presence of somatic activating PIK3CA mutations in patients with CLAPO. We also report an update of the phenotype and natural history of the syndrome.”,
“timestamp”: “Thu Jul 04 06:15:27 +0000 2019”
}
]
# Start GNormPlusServercd ~/bern/GNormPlusJavanohup java -Xmx16G -Xms16G -jar GNormPlusServer.jar 18895 >> ~/bern/logs/nohup_gnormplus.out 2>&1 &# Start tmVar2Servercd ~/bern/tmVarJavanohup java -Xmx8G -Xms8G -jar tmVar2Server.jar 18896 >> ~/bern/logs/nohup_tmvar.out 2>&1 &# Start normalizerscd ~/bern/sh load_dicts.sh# Check your GPU number(s)echo $CUDA_VISIBLE_DEVICES# Set your GPU number(s)export CUDA_VISIBLE_DEVICES=0# Run BERNnohup python3 -u server.py --port 8888 --gnormplus_home ~/bern/GNormPlusJava --gnormplus_port 18895 --tmvar2_home ~/bern/tmVarJava --tmvar2_port 18896 >> logs/nohup_BERN.out 2>&1 &# Print logstail -F logs/nohup_BERN.out
List processes (every 5s)
watch -n 5 "ps auxww | egrep 'python|java|node' | grep -v grep"
Periodic HTTPS GET checker
Permission setting
chmod +x scripts/bern_checker.sh
crontab (every 30 min)
crontab -e*/30 * * * * /home/<YOUR_ACCOUNT>/bern/scripts/bern_checker.sh >> /home/<YOUR_ACCOUNT>/bern/logs/bern_checker.out 2>&1
Add a new issue to https://github.com/dmis-lab/bern/issues
@article{10.1093/bioinformatics/btz682,
author = {Lee, Jinhyuk and Yoon, Wonjin and Kim, Sungdong and Kim, Donghyeon and Kim, Sunkyu and So, Chan Ho and Kang, Jaewoo},
title = “{BioBERT: a pre-trained biomedical language representation model for biomedical text mining}”,
journal = {Bioinformatics},
year = {2019},
month = {09},
issn = {1367-4803},
doi = {10.1093/bioinformatics/btz682},
url = {https://doi.org/10.1093/bioinformatics/btz682},
}
```