umls-docker

A step-by-step guide and scripts to build a MySQL docker image with UMLS, UML-Interface and UMLS-Similarity.

Prepare the environment to create the Docker image:

Install Docker Engine and Docker Compose

Clone this repo:

git clone https://github.com/cbadenes/umls-docker.git

Move into the root directory:
```
cd umls-docker
```
Download the UMLS Metathesaurus Full Subset (File Size: compressed- 5GB, uncompressed-32.8GB) and extract the contents into this directory. You will probably have to use Metamorphosys to generate the bbdd dumps. Move folders META/ and NET/ into this directory.
Download the UMLS-Interface API and extract the contents into this directory. Folder UMLS-Interface-1.51 is created.
Download the UMLS-Similarity API and extract the contents into this directory. Folder UMLS-Similarity-1.47 is created.

Create a Docker image with both the UMLS Knowledge and the related services:

With Docker Daemon running execute command:
```
docker build -t umls-mysql -f Dockerfile .
```
If you have an ARM architecture (e.g. Mac M1), use the dockerfile-arm descriptor:
```
docker build -t umls-mysql -f Dockerfile-arm .
```
It will take a few minutes to create the image.

Load UMLS data and prepare interfaces

Deploy the image via docker compose:
```
docker compose up -d
```
The UMLS database will not be available instantly. Once the container is running the database will load and MySQL will be restarted. To check if database is available you can search for the log trace mysqld: ready for connections by:
```
docker compose logs
```

Edit the file META/populate_mysql_db.sh to replace corresponding lines with following:

- MYSQL_HOME=/usr 
- user=root 
- password= 
- db_name=umls
- In each lne starting with $MYSQL_HOME/bin/mysql replace  
    -vvv -u $user -p $password 
with 
    -vvv --local-infile

Edit the file NET/populate_netmysql_db.sh to replace corresponding lines with following:

- MYSQL_HOME=/usr 
- user=root 
- password= 
- db_name=umls
- In each lne starting with $MYSQL_HOME/bin/mysql replace  
    -vvv -u $user -p $password 
with 
    -vvv --local-infile

Once the file are updated you can connect to the container via:
```
docker exec -it umlsdb /bin/bash
```

Edit the file /etc/mysql/my.cnf and add the following properties:

[mysqld]
 key_buffer=600M
 table_cache=300
 query_cache_limit=3M
 query_cache_size=100M
 read_buffer_size=200M
 myisam_sort_buffer_size=200M
 bulk_insert_buffer_size=100M
 join_buffer_size=100M

Enables execution permissions for /umls_init.sh by means of:
```
chmod +x umls_init.sh
```
And finally load the UMLS information via:
```
./umls_init.sh
```
This process can take hours.

Calculates similarities between UMLS terms

Enter the container by means of:
```
docker exec -it umlsdb /bin/bash
```
Calculates the similarity between terms (e.g. paracetamol and diarrhea):
```
ulms-similarity.pl paracetamol diarrhea
```
it returns 0.0833<>paracetamol(C0000970)<>diarrhea(C0011991)
Or calculate the similarity between CUI codes (e.g. C0000970 and C0011991):
```
ulms-similarity.pl C0000970 C0011991
```
it returns 0.0833<>C0000970(paracetamol)<>C0011991(Diarrhea NOS)