Biological Interpretation

• ID: MICROB-011
• Type: System Component
• Audience: Students, researchers, analysts, and practitioners
• Theme: Translating microbiome outputs into defensible biological insight

Introduction

Biological interpretation is the stage where microbiome analysis results are translated into meaningful biological statements.

At this point, the Microbiome Analysis System may contain:

• feature tables
• taxonomic profiles
• functional profiles
• diversity results
• differential analysis results
• plots
• metadata summaries
• quality-control reports

These outputs are not the final interpretation by themselves.

Biological interpretation connects analytical evidence back to the biological question, study design, sample metadata, and known limitations.

Why Biological Interpretation Matters

Microbiome analysis can generate many tables and figures.

Without interpretation, these outputs may remain disconnected observations.

Biological interpretation helps answer questions such as:

• What do the observed microbial patterns suggest?
• Are the results consistent with the study question?
• Which findings are descriptive?
• Which findings are exploratory?
• Which findings are statistically supported?
• Which findings require caution?
• What limitations affect interpretation?
• What should be reported as the main message?

A strong interpretation stage prevents the analysis from becoming a collection of disconnected outputs.

Position in the Microbiome Analysis System

Biological interpretation receives evidence from multiple analytical stages.

flowchart TB
  A[Taxonomic Profiling] --> E[Biological Interpretation]
  B[Diversity Analysis] --> E
  C[Functional Profiling] --> E
  D[Differential Analysis] --> E
  F[Metadata and Study Design] --> E
  G[Quality Control] --> E
  E --> H[Reproducible Reporting]

flowchart TB
  A[Taxonomic Profiling] --> E[Biological Interpretation]
  B[Diversity Analysis] --> E
  C[Functional Profiling] --> E
  D[Differential Analysis] --> E
  F[Metadata and Study Design] --> E
  G[Quality Control] --> E
  E --> H[Reproducible Reporting]

Interpretation should never be separated from metadata, study design, and quality-control context.

Interpretation Is Evidence Integration

Biological interpretation is not simply describing the largest bars in a plot.

It is an evidence-integration process.

The analyst should consider:

• study objective
• sample type
• comparison groups
• sequencing strategy
• metadata completeness
• quality-control outcomes
• taxonomic patterns
• functional patterns
• diversity patterns
• differential results
• statistical support
• biological plausibility
• limitations

The goal is to produce statements that are supported by the analysis and honest about uncertainty.

From Results to Claims

A useful way to interpret microbiome results is to separate observations, interpretations, and claims.

Observation:
  A taxon has higher relative abundance in one sample group.

Interpretation:
  This pattern may indicate a community composition difference between groups.

Claim:
  This taxon is associated with the condition in this dataset, subject to study limitations.

A claim should be weaker or stronger depending on the evidence.

For toy workflow data, no biological claims should be made.

Interpretation Strength

Not all findings have the same strength.

A simple interpretation framework is:

Descriptive:
  The result summarizes what is seen in the data.

Exploratory:
  The result suggests a possible pattern that needs further testing.

Supported:
  The result is consistent across analyses and supported by appropriate statistics.

Confirmed:
  The result is supported by independent validation or experimental evidence.

Most microbiome workflow outputs are descriptive or exploratory unless the study design and validation support stronger conclusions.

Common Interpretation Questions

During interpretation, ask:

• What was the original biological question?
• Which outputs directly address that question?
• Are the sample groups clearly defined?
• Are the results consistent across taxonomic, diversity, functional, and differential analyses?
• Could sequencing depth, batch effects, or metadata gaps explain the pattern?
• Are the taxonomic labels reliable enough for the claim being made?
• Are functional findings measured or inferred?
• Are p-values and effect sizes interpreted together?
• Are limitations clearly stated?

These questions help keep interpretation defensible.

Example Interpretation Support Scripts

The following scripts provide a lightweight MAS-side interpretation support workflow.

They do not create biological conclusions automatically. Instead, they collect key outputs from previous chapters and generate an interpretation evidence table and draft interpretation notes.

The workflow uses two scripts:

scripts/R/11a-build-interpretation-evidence.R
scripts/R/11b-draft-interpretation-notes.R

The first script checks for major MAS outputs and builds an evidence index.

The second script creates a plain Markdown interpretation draft that can be edited by the analyst.

11a: Build the Interpretation Evidence Table

Save this script as:

scripts/R/11a-build-interpretation-evidence.R

###############################################################################
# Microbiome Analysis System
# 11a-build-interpretation-evidence.R
#
# Purpose:
#   Build an evidence index from MAS analysis outputs.
#
# Usage:
#   Rscript scripts/R/11a-build-interpretation-evidence.R
###############################################################################

library(readr)
library(dplyr)
library(tibble)

interpretation_dir <- "data/interpretation"
report_dir <- "data/reports"

dir.create(interpretation_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(report_dir, recursive = TRUE, showWarnings = FALSE)

expected_outputs <- tibble(
  evidence_area = c(
    "data_acquisition",
    "quality_control",
    "feature_generation",
    "taxonomic_profiling",
    "diversity_analysis",
    "functional_profiling",
    "differential_analysis"
  ),
  expected_file = c(
    "data/reports/data-acquisition-summary.tsv",
    "data/reports/qc-readiness-report.tsv",
    "data/reports/feature-table-check-report.tsv",
    "data/reports/taxonomic-profile-report.tsv",
    "data/reports/diversity-analysis-report.tsv",
    "data/reports/functional-profile-report.tsv",
    "data/reports/differential-analysis-report.tsv"
  ),
  interpretation_role = c(
    "documents whether data were acquired and organized",
    "documents whether FASTQ inputs passed lightweight QC checks",
    "documents whether the feature table is structurally usable",
    "summarizes taxonomic profile readiness",
    "summarizes alpha and beta diversity outputs",
    "summarizes functional profile readiness",
    "summarizes differential comparison outputs"
  )
)

evidence_index <- expected_outputs %>%
  rowwise() %>%
  mutate(
    status = ifelse(file.exists(expected_file), "FOUND", "MISSING"),
    file_size_bytes = ifelse(file.exists(expected_file), file.info(expected_file)$size, NA_real_)
  ) %>%
  ungroup()

write_tsv(
  evidence_index,
  file.path(interpretation_dir, "interpretation-evidence-index.tsv")
)

summary_report <- evidence_index %>%
  count(status, name = "n_outputs")

write_tsv(
  summary_report,
  file.path(report_dir, "interpretation-evidence-summary.tsv")
)

message("Created:")
message("  ", file.path(interpretation_dir, "interpretation-evidence-index.tsv"))
message("  ", file.path(report_dir, "interpretation-evidence-summary.tsv"))

Run it from the MAS project root:

Rscript scripts/R/11a-build-interpretation-evidence.R

This creates:

data/interpretation/interpretation-evidence-index.tsv
data/reports/interpretation-evidence-summary.tsv

11b: Draft Interpretation Notes

Save this script as:

scripts/R/11b-draft-interpretation-notes.R

###############################################################################
# Microbiome Analysis System
# 11b-draft-interpretation-notes.R
#
# Purpose:
#   Create a draft biological interpretation notes file from MAS outputs.
#
# Usage:
#   Rscript scripts/R/11b-draft-interpretation-notes.R
###############################################################################

library(readr)
library(dplyr)
library(glue)

interpretation_dir <- "data/interpretation"
report_dir <- "data/reports"

dir.create(interpretation_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(report_dir, recursive = TRUE, showWarnings = FALSE)

evidence_file <- file.path(interpretation_dir, "interpretation-evidence-index.tsv")
notes_file <- file.path(interpretation_dir, "biological-interpretation-notes.md")
report_file <- file.path(report_dir, "biological-interpretation-report.tsv")

if (!file.exists(evidence_file)) {
  stop(
    "Missing interpretation evidence index: ",
    evidence_file,
    "\nRun: Rscript scripts/R/11a-build-interpretation-evidence.R"
  )
}

evidence <- read_tsv(evidence_file, show_col_types = FALSE)

found_outputs <- evidence %>%
  filter(status == "FOUND") %>%
  pull(evidence_area)

missing_outputs <- evidence %>%
  filter(status == "MISSING") %>%
  pull(evidence_area)

found_text <- if (length(found_outputs) > 0) {
  paste0("- ", found_outputs, collapse = "\n")
} else {
  "- No evidence outputs were found"
}

missing_text <- if (length(missing_outputs) > 0) {
  paste0("- ", missing_outputs, collapse = "\n")
} else {
  "- No expected evidence outputs are missing"
}

notes <- glue(
"# Biological Interpretation Notes

## Interpretation Status

This file is a draft interpretation support document generated from MAS workflow outputs.

It should be edited by the analyst before reporting.

## Evidence Outputs Found

{found_text}

## Evidence Outputs Missing

{missing_text}

## Draft Interpretation Framework

### 1. Biological Question

State the biological question that the microbiome analysis is intended to answer.

### 2. Study Design Context

Summarize the sample type, comparison groups, sequencing strategy, and relevant metadata.

### 3. Quality-Control Context

Summarize whether data acquisition and quality-control checks support downstream interpretation.

### 4. Taxonomic Patterns

Describe major taxonomic patterns, dominant taxa, or community composition observations.

### 5. Diversity Patterns

Describe alpha diversity and beta diversity patterns. Avoid overinterpreting ordination plots.

### 6. Functional Patterns

Describe functional profile patterns if functional outputs are available. Distinguish functional potential from functional activity.

### 7. Differential Results

Summarize differential results as candidate observations. Include effect sizes, adjusted p-values where appropriate, and limitations.

### 8. Integrated Interpretation

Connect taxonomic, diversity, functional, and differential evidence into cautious biological statements.

### 9. Limitations

Document limitations including sample size, metadata completeness, sequencing strategy, toy data, batch effects, or method assumptions.

### 10. Reporting Statement

Write a concise report-ready statement supported by the evidence.

## Important Note

For the MAS toy example data, no biological conclusions should be made. The purpose is workflow testing only.
"
)

writeLines(notes, notes_file)

report <- tibble::tibble(
  metric = c(
    "evidence_outputs_found",
    "evidence_outputs_missing",
    "notes_file",
    "interpretation_status"
  ),
  value = c(
    length(found_outputs),
    length(missing_outputs),
    notes_file,
    "DRAFT_NOTES_CREATED"
  )
)

write_tsv(report, report_file)

message("Created:")
message("  ", notes_file)
message("  ", report_file)

Run it from the MAS project root:

Rscript scripts/R/11b-draft-interpretation-notes.R

This creates:

data/interpretation/biological-interpretation-notes.md
data/reports/biological-interpretation-report.tsv

Running the Complete Interpretation Support Example

If you are continuing from previous chapters, generate the interpretation support files:

Rscript scripts/R/11a-build-interpretation-evidence.R
Rscript scripts/R/11b-draft-interpretation-notes.R
cat data/reports/interpretation-evidence-summary.tsv
cat data/reports/biological-interpretation-report.tsv

Then open and edit:

data/interpretation/biological-interpretation-notes.md

The generated Markdown file is a draft. It should be reviewed and rewritten by the analyst.

Evidence Table

The interpretation evidence index records whether expected workflow outputs are available.

Example structure:

evidence_area   expected_file   status  interpretation_role
quality_control data/reports/qc-readiness-report.tsv    FOUND   documents whether FASTQ inputs passed lightweight QC checks
taxonomic_profiling data/reports/taxonomic-profile-report.tsv   FOUND   summarizes taxonomic profile readiness
diversity_analysis  data/reports/diversity-analysis-report.tsv  FOUND   summarizes alpha and beta diversity outputs

This table helps the analyst avoid interpreting results without checking whether the supporting evidence exists.

Writing Biological Interpretation

A strong interpretation paragraph usually includes:

• the main observed pattern
• the evidence supporting the pattern
• the biological context
• the level of confidence
• the limitation or caution

A useful structure is:

The analysis observed [pattern] in [samples or groups].
This was supported by [evidence source].
Biologically, this may suggest [cautious interpretation].
However, interpretation is limited by [limitation].

This structure helps avoid unsupported claims.

Examples of Careful Language

Prefer careful language such as:

The results suggest...
The pattern is consistent with...
This observation may indicate...
This finding should be interpreted cautiously because...
Additional validation would be required to confirm...

Avoid overstated language such as:

This proves...
This microbe causes...
This pathway is active...
This taxon is responsible for...

Microbiome results are often associative and context-dependent.

Integrating Evidence Across Outputs

A stronger interpretation connects multiple outputs.

For example:

A taxonomic shift may be more convincing if it is supported by relative abundance profiles, diversity patterns, and differential analysis results.

However, consistency across outputs does not automatically prove causality.

It improves coherence, but conclusions still depend on study design and validation.

Interpretation Limitations

Common limitations include:

• small sample size
• missing metadata
• weak group definitions
• unbalanced groups
• batch effects
• sequencing depth differences
• marker-gene resolution limits
• incomplete taxonomic assignment
• inferred rather than measured function
• lack of experimental validation
• exploratory rather than confirmatory design

These limitations should be stated clearly in the final report.

MAS Biological Interpretation Outputs

At the end of this stage, MAS should have:

• interpretation evidence index
• interpretation evidence summary
• draft biological interpretation notes
• biological interpretation report
• analyst-reviewed interpretation statements
• documented limitations

flowchart LR
  A[Analysis Outputs] --> B[Evidence Index]
  B --> C[Draft Interpretation Notes]
  C --> D[Analyst Review]
  D --> E[Report-Ready Interpretation]

flowchart LR
  A[Analysis Outputs] --> B[Evidence Index]
  B --> C[Draft Interpretation Notes]
  C --> D[Analyst Review]
  D --> E[Report-Ready Interpretation]

Key Takeaways

Biological interpretation turns analysis outputs into defensible insight.

A strong interpretation stage ensures that:

• claims are supported by evidence
• results are connected to the biological question
• metadata and study design are considered
• limitations are documented
• descriptive findings are not overstated
• toy workflow outputs are not biologically interpreted

Interpretation is where microbiome analysis becomes useful, but it is also where unsupported claims can easily enter the report.

What Comes Next

The next chapter examines Reproducible Reporting, where the full microbiome workflow is assembled into a transparent, reusable, and shareable report.