Unlocking DrugBank: Parsing and Visualizing Mechanistic Data

Introduction

The DrugBank database is the global gold standard for drug information, containing rich data on chemical structures, pharmacology, and molecular targets. However, the raw data is provided as a massive, complex XML file that is difficult to use in statistical analysis.

dbparser solves this problem. It transforms that complex XML into a Drugverse Object (dvobject)—a structured collection of R data frames ready for analysis.

In this tutorial, we will demonstrate how to:

Parse the database.
Explore the data structure.
Visualize key insights using dplyr and canvasXpress.

1. Loading Data

The Parsing Workflow

In a real-world scenario, you would parse the full XML database (downloaded from DrugBank) using parseDrugBank().

library(dbparser)

# Parse the full XML file (this creates a dvobject)
# drugbank_db <- parseDrugBank("drugbank_full_database.xml")

Loading Sample Data

For this vignette, we will load curated sample datasets included in the package. These represent specific tables that would be found inside the full dvobject.

suppressPackageStartupMessages({
  library(tidyr)
  library(dplyr)
  library(tibble)
  library(canvasXpress)
  library(dbparser)
})

# Load sample tables representing parts of the parsed dvobject
drugs <- readRDS(system.file("drugs.RDS", package = "dbparser"))
drug_groups <- readRDS(system.file("drug_groups.RDS", package = "dbparser"))
drug_targets_actions <- readRDS(system.file("targets_actions.RDS", package = "dbparser"))

2. Analysis: The Drug Landscape

What does the universe of approved drugs look like? Let’s analyze the composition of the database by drug type (Small Molecule vs. Biotech).

# Prepare data: Count drugs by type
type_stat <- drugs %>% 
  group_by(type) %>% 
  summarise(Count = n()) %>% 
  arrange(desc(Count)) %>% 
  column_to_rownames("type")

# Visualize
canvasXpress(
  data             = type_stat,
  graphType        = "Bar",
  title            = "Composition of DrugBank: Drug Types",
  showSampleNames  = FALSE,
  legendPosition   = "right"
)

3. Analysis: Approval Status

Drugs are categorized into groups such as “approved”, “investigational”, or “experimental”. How does the complexity (Biotech vs. Small Molecule) differ across these groups?

# Prepare data: Cross-tabulate Type vs Group
group_stat <- drugs %>% 
  full_join(drug_groups, by = "drugbank_id") %>% 
  group_by(type, group) %>% 
  summarise(count = n(), .groups = 'drop') %>% 
  pivot_wider(names_from = group, values_from = count, values_fill = 0) %>% 
  column_to_rownames("type")

# Visualize with a Stacked Bar Chart
canvasXpress(
  data           = group_stat,
  graphType      = "Stacked",
  graphOrientation = "horizontal",
  title          = "Drug Types by Approval Status",
  xAxisTitle     = "Number of Drugs",
  legendPosition = "bottom",
  xAxis2Show     = FALSE
)

4. Analysis: Molecular Mechanisms

One of DrugBank’s most valuable features is the detailed information on how drugs interact with their targets (Proteins, Enzymes, etc.). Are drugs mostly inhibitors, agonists, or antagonists?

# Prepare data: Top 10 most common Mechanisms of Action
targetActionCounts <- drug_targets_actions %>% 
    group_by(action) %>% 
    summarise(Count = n()) %>% 
    arrange(desc(Count)) %>% 
    slice_head(n = 10) %>% 
    column_to_rownames("action")

# Visualize
canvasXpress(
  data            = targetActionCounts,
  graphType       = "Bar",
  graphOrientation = "vertical",
  colorBy         = "Count",
  title           = "Top 10 Mechanisms of Action",
  xAxisTitle      = "Number of Interactions",
  showSampleNames = FALSE,
  legendPosition  = "none"
)

5. Next Steps: Integrated Pharmacovigilance

Now that you have mastered the mechanistic data in DrugBank, you can combine it with real-world data!

dbparser now supports OnSIDES (Adverse Events) and TWOSIDES (Drug-Drug Interactions).

Check out the Integrated Pharmacovigilance Vignette to learn how to merge these databases to perform polypharmacy risk analysis.