DrugBank Database XML Parser

Mohammed Ali, Ali Ezzat

2019-08-28

The main purpose of the dbparser package is to parse the DrugBank database which is downloadable in XML format from this link. The parsed data can then be explored and analyzed as desired by the user. The dbparser package further provides the facility of saving the parsed data into a given database.

Getting Started – Loading the Data

Following is sample code that parses the DrugBank database, then loads the drugs info, drug groups info and drug targets actions info.

## load dbparser package
library(dbparser)
library(dplyr)
library(ggplot2)
library(XML)

## parse data from XML and save it to memory
get_xml_db_rows("drugbank_database.xml")

## load drugs data
drugs <- parse_drug()

## load drug groups data
drug_groups <- parse_drug_groups()

## load drug targets actions data
drug_targets_actions <- parse_drug_targets_actions()

Saving into a database

The parsed data may be saved into a given database. Databases supported by dbparser include MS SQL Server, MySQL and any database supported by DBI package. Following is an example of saving the parsed data into a MySQL database.

## open a connection to the desired database engine with an already
## created database
open_db(xml_db_name =  "drugbank.xml", driver = "SQL Server",
server = "MOHAMMED\\\\SQL2016", output_database = "drugbank")

## save 'drugs' dataframe to DB
parse_drug(TRUE)

## save 'drug_groups' dataframe to DB
parse_drug_groups(TRUE)

## save 'drug_targets_actions' dataframe to DB
parse_drug_targets_actions(TRUE)

## finally close db connection
close_db()

Exploring the data

Following is an example involving a quick look at a few aspects of the parsed data. First we look at the proportions of biotech and small-molecule drugs in the data.

## view proportions of the different drug types (biotech vs. small molecule)
drugs %>% 
    select(type) %>% 
    ggplot(aes(x = type)) + 
    geom_bar() + 
    guides(fill=FALSE)     ## removes legend for the bar colors

Below, we view the different drug_groups in the data and how prevalent they are.

## view proportions of the different drug types for each drug group
drugs %>% 
    rename(parent_key = primary_key) %>% 
    full_join(drug_groups, by = 'parent_key') %>% 
    select(type, text) %>% 
    ggplot(aes(x = text, fill = type)) + 
    geom_bar() + 
    theme(legend.position= 'bottom', 
          axis.text.x = element_text(angle=45, vjust=0.5)) + 
    labs(x = 'Drug Group', 
         y = 'Quantity', 
         caption="created by ggplot") + 
    coord_flip()

Finally, we look at the drug_targets_actions to observe their proportions as well.

## get counts of the different target actions in the data
targetActionCounts <- 
    drug_targets_actions %>% 
    group_by(text) %>% 
    summarise(count = n()) %>% 
    arrange(desc(count))

## get bar chart of the 10 most occurring target actions in the data
p <- 
    ggplot(targetActionCounts[1:10,], 
           aes(x = reorder(text,count), y = count, fill = letters[1:10])) + 
    geom_bar(stat = 'identity') +
    labs(fill = 'action', 
         x = 'Target Action', 
         y = 'Quantity', 
         caption = 'created by ggplot') + 
    guides(fill = FALSE) +    ## removes legend for the bar colors
    coord_flip()              ## switches the X and Y axes

## display plot
p