Skip to content

UC-2.5 — Descriptive Statistics of Samples Across Databases

Module: 2 – Exploratory Analysis: Ranking the Functional Potential of Samples and Compounds
Visualization type: Interactive box-and-scatter (jitter) plot of unique KO counts
Primary inputs: BioRemPP, HADEG, and KEGG results tables with sample-level KO annotations
Primary outputs: Distribution of unique KO counts per sample for each database

Scientific Question and Rationale

Question: How does the statistical distribution of KO annotation counts compare across different samples and annotation databases, and what insights might this provide about their scope and focus?

This use case provides a comparative statistical summary of unique KEGG Orthology (KO) counts across all analyzed biological samples under different annotation sources. By switching between BioRemPP, HADEG, and KEGG, the visualization shows how KO annotation counts are distributed for each database. A combined box plot and jittered scatter plot allows users to see both the overall distribution (median, quartiles, spread) and the position of individual samples, which may offer insight into database-specific annotation coverage, variability, and potential outliers.

Data and Inputs

  • Primary data sources:
  • BioRemPP results table
  • HADEG results table
  • KEGG results table
  • Key columns (per selected database):
  • sample – identifier for each biological sample
  • ko or Gene – KEGG Orthology identifier (or equivalent KO column)
  • Accepted format: semicolon-delimited text tables (.txt or .csv)
  • Entity of interest: unique KO counts per sample for the selected database

Analytical Workflow

  1. Database Selection
    The user selects a data source (e.g., "BioRemPP") from an interactive dropdown menu.

  2. Dynamic Data Loading
    The corresponding raw results table for the selected database is loaded and parsed from its semicolon-delimited format.

  3. Column Resolution
    The script automatically identifies the correct columns for sample and the KO identifier, treating both ko and Gene as valid KO columns.

  4. Data Processing and Aggregation
    The data is filtered to remove invalid entries (e.g., missing sample or KO identifier). For the selected database, the script computes, for each unique sample, the number of distinct KO identifiers, yielding a per-sample unique_ko_count.

  5. Rendering
    The visualization is constructed using the per-sample unique_ko_count values:

  6. A box plot is rendered to summarize the distribution (median and quartiles, with potential whiskers and outliers).
  7. A jittered scatter plot is overlaid, where each point represents one sample; the vertical position encodes its unique_ko_count, and the horizontal position is slightly jittered to reduce overlap and improve readability.

How to Read the Plot

  • Dropdown Menu
    Use the dropdown to switch between the BioRemPP, HADEG, and KEGG datasets. The distribution updates dynamically for the selected source.

  • Box (Central Distribution)
    The box represents the interquartile range (IQR)—the central 50% of the unique_ko_count values for the selected database.

  • The horizontal line inside the box marks the median.

  • Whiskers and optional points outside the whiskers indicate the spread and possible outliers.

  • Jittered Points (Individual Samples)
    Each dot corresponds to a single Sample.

  • Its vertical position encodes the unique_ko_count for that sample.

  • Its horizontal position is slightly randomized (jittered) to prevent overlap between samples with similar counts.

  • Hover Information
    Hovering over a point reveals detailed information, such as:

  • sample name,
  • exact unique KO count, and
  • rank of that sample within the selected database.

Representative Output

The image below illustrates a representative output generated by this use case using the example dataset.

Click on the image to enlarge and explore details.

Representative output for UC-2.5

Interpretation and Key Messages

  • Comparing Distributions Across Databases By toggling between BioRemPP, HADEG, and KEGG, users can compare how KO annotation counts are distributed:
  • A higher median in one database may suggest broader average annotation coverage per sample in that source.

  • A wider IQR or longer whiskers could indicate more variability in KO annotation counts among samples for that database.

  • Identifying Database-Specific Outliers Samples that appear as outliers (points far above or below the box) in one database but not in others may reflect a specific focus of that annotation source, warranting closer examination of which KO categories drive the difference.

  • Assessing Annotation Scope and Focus Databases with narrow, tightly clustered distributions (short boxes and whiskers) may have more uniform annotation depth or a focused scope, while broader distributions may correspond to more heterogeneous annotation coverage. Comparing spread and central tendency across databases can provide insights into differences in curation strategy and scope.

Reproducibility and Assumptions

  • Input Format
    The analysis assumes semicolon-delimited tables for each database, each containing at least sample and a KO identifier column (ko or Gene).

  • KO Column Handling
    The code dynamically resolves whether the KO identifiers are stored under ko or Gene and treats both as equivalent sources of KO IDs.

  • Uniqueness Definition KO annotation count is defined as the count of unique KO identifiers per sample. Multiple occurrences of the same KO within a sample (e.g., via multiple genes or compounds) are counted only once.

  • Jitter and Determinism
    The horizontal jitter applied to sample points is seeded to produce a stable layout for a given dataset, so the same input data yields identical visual arrangements across runs. The box plot itself is a deterministic statistical summary of the unique_ko_count distribution.

Activity diagram of the use case

Click on the image to enlarge and explore details.

Activity diagram of the use case