BABAPPASnake: a workflow for episodic selection analysis with robustness-aware summaries

BABAPPASnake is an integrated, reproducible workflow that unifies orthogroup construction, alignment, phylogeny, and selection testing to facilitate robust, sensitivity-aware episodic selection analysis with comprehensive robustness summaries.

The Gist

Imagine you are trying to solve a massive, ancient mystery: How did certain genes in mosquitoes evolve to help them fight off infections?

In the past, solving this mystery was like trying to build a house where every brick had to be ordered from a different store, assembled by a different contractor, and glued together by a different person. You'd have to manually move data from one computer program to another, risking lost notes, broken connections, and a final result that no one could fully trust because the process was so messy.

Enter BABAPPASnake.

What is BABAPPASnake?

Think of BABAPPASnake not as a single tool, but as a super-smart, automated project manager for evolutionary biologists. It's a "workflow" (a set of instructions) that takes a messy pile of genetic data and guides it through every single step of the investigation, from start to finish, without needing a human to constantly babysit the process.

The name comes from a cute story: the authors named it after their young son's word for a butterfly ("BABAPPA"), because the software helps scientists "fly" through complex data just like a butterfly flits from flower to flower.

How Does It Work? (The "Six Paths" Analogy)

Usually, when scientists analyze genes, they pick one way to do it. It's like asking one person to taste a soup and deciding if it's salty. If that person has a cold, the whole conclusion is wrong.

BABAPPASnake does something smarter. It says, "Let's not just ask one person; let's ask six different experts, using six different methods, to taste the soup."

1. The Setup: It gathers the genetic "ingredients" (proteins and DNA) from mosquitoes.
2. The Six Paths: It runs the analysis through six different "kitchens" (combinations of different software tools and cleaning methods).
   • Path 1: Method A + Clean Style 1
   • Path 2: Method A + Clean Style 2
   • ...and so on, up to Path 6.
3. The Comparison: Instead of just giving you one answer, it shows you the results from all six paths.
   • Did all six agree the gene is under "selection" (evolving quickly)? Great! That's a strong, reliable finding.
   • Did only one path say yes, while the other five said no? Caution! That result is likely a fluke or a mistake in that specific method.

This is what the paper calls "Robustness-Aware." It doesn't just tell you what happened; it tells you how sure you can be about it.

The Real-World Test: The Mosquito Melanization Module

To prove it works, the authors tested this system on a specific group of four mosquito genes known for helping the insect's immune system turn dark (a process called melanization) to trap parasites.

• The Result: The workflow found that these genes were indeed evolving in interesting ways.
• The Twist: It showed that while the "core" part of the immune system was evolving strongly, the "catalytic" (helper) part was less active.
• The Lesson: Because the software ran all six paths, the scientists could see that this pattern was mostly consistent, but not perfect. This prevented them from making a bold, overconfident claim. Instead, they could say, "This looks like a real trend, but we need to keep studying it."

Why Does This Matter?

In the old days, scientists might have run one analysis, found a "cool" result, and published it as a fact. If they had used a slightly different tool, they might have found nothing. This led to a lot of confusion in science.

BABAPPASnake changes the game by:

• Automating the boring stuff: It handles the file conversions and error-checking so humans can focus on the biology.
• Exposing the truth: It highlights where the data is shaky and where it is solid.
• Saving time: It allows researchers to run complex, multi-step experiments with a single command.

The Bottom Line

BABAPPASnake is like a safety net for evolutionary research. It ensures that when scientists say, "We found a new evolutionary pattern," they aren't just guessing based on one lucky roll of the dice. They have checked the dice six different ways, and the result holds up.

It turns a fragmented, risky process into a reliable, reproducible journey, helping us understand how life evolves with much greater confidence.

Technical Summary

1. Problem Statement

Episodic selection analysis in molecular evolution is currently hindered by technical fragmentation. Researchers typically assemble disjointed tools for ortholog discovery, codon alignment, phylogeny reconstruction, exploratory scans, and branch-site testing. This fragmentation leads to:

• Reproducibility Risks: Critical analytical decisions are scattered across manual scripts and formats, making it difficult to replicate studies.
• Unreported Sensitivity: Results often vary significantly based on ortholog selection rules, alignment engines, trimming states, and model settings. Without explicit reporting of these variations, conclusions are frequently overinterpreted as "workflow-invariant" truths rather than method-dependent signals.

2. Methodology: BABAPPASnake Workflow

BABAPPASnake is an integrated, rule-ordered workflow designed to unify the episodic selection pipeline into a single reproducible framework. It is built on Snakemake logic and supports both guided interactive and non-interactive execution.

Core Architecture & Logic

The workflow operates through a structured decision tree:

1. Orthogroup Discovery: It runs parallel discovery routes using RBH (Reciprocal Best Hits) and OrthoFinder.
2. Backend Selection: It maps the query to OrthoFinder groups via BLASTP and compares strict 1:1 ortholog counts.
   • If one route yields more strict 1:1 orthologs, that backend is selected.
   • In case of a tie, RBH is prioritized.
   • If both yield zero support, the workflow halts explicitly.
3. CDS Integration & QC: If CDS (Coding Sequence) data is provided, the workflow performs quality control, including intron clipping (lowercase masking), ORF/frame checks, and mapping CDS to proteins. If CDS is missing, it pauses for staged resumption.
4. Multi-Engine Pathways: The workflow expands into six parallel pathways (Method × Trim State):
   • Alignment Methods: BABAPPAlign, MAFFT, PRANK.
   • Trim States: Raw vs. ClipKIT.
5. Inference Chain (Per Pathway):
   • Phylogeny: IQ-TREE for tree reconstruction (with optional outgroup rooting).
   • Recombination Screening (Optional): HyPhy GARD screening is available as a conservative preprocessing layer. It reports breakpoint evidence but does not automatically reroute downstream analysis to fragments by default.
   • Exploratory Scans: HyPhy aBSREL and MEME to identify candidate branches/sites.
   • Follow-up Testing: Branch-site codeml (PAML) testing with dynamic foreground nomination and Benjamini-Hochberg (BH) correction.
   • Ancestral Reconstruction: ASR to extract sequence changes on selected branches.
6. Output Generation: The system aggregates results into pathway-specific summaries, cross-pathway robustness matrices, consensus narratives, and machine-readable provenance logs.

3. Key Contributions

• Integrated Framework: Replaces fragmented toolchains with a single, executable workflow that handles orthogroup construction, alignment, phylogeny, and selection testing.
• Robustness-Aware Interpretation: Instead of a single "best" result, the workflow generates comparative outputs across multiple method-trim combinations. It explicitly categorizes results into reproducibility classes (e.g., high, moderate, method-sensitive, trim-sensitive).
• Staged Execution: Features a "stop-and-resume" capability, particularly for CDS integration, allowing users to manage large datasets in stages.
• Conservative Recombination Handling: Introduces an optional GARD screening module that reports recombination status without forcing automatic downstream rerouting, preserving the default full-length inference chain unless explicitly modified.

4. Results: Empirical Demonstration

The authors validated the workflow using a four-gene mosquito melanization-associated module (SPCLIP1, CLIPA8, CLIPB14, CLIPB15), which includes core and catalytic pathway tiers.

• Workflow Performance: The analysis successfully processed 4 genes across 6 pathways (3 methods × 2 trim states).
• Signal Detection:
  • In the primary pathway (BABAPPAlign + ClipKIT), 17 exploratory candidates yielded 14 branch-site follow-up-positive branches after BH correction.
  • 48 BEB codons (Posterior Probability ≥ 0.95) were identified, with a strong skew toward the core tier (43 codons) compared to the catalytic tier (5 codons).
• Robustness Analysis:
  • Directional Asymmetry: The trend of higher selection signals in the core tier vs. catalytic tier was observed in 4 out of 6 pathways.
  • Sensitivity: One pathway showed an inversion (catalytic > core), and one was a tie, demonstrating that results are method-sensitive.
  • Statistical Significance: While the directional trend was suggestive, exact asymmetry tests were non-significant ($p > 0.16$), reinforcing the workflow's conclusion that the pattern is hypothesis-generating rather than statistically decisive.
• Global Context: Global one-ratio codeml values ($\omega = 0.186–0.297$) indicated purifying selection, confirming that the detected episodic signals occur on constrained backgrounds.

5. Significance

• Paradigm Shift in Reporting: BABAPPASnake moves the field away from "single-path" overstatement toward transparency. By explicitly reporting which signals are robust across methods and which are artifacts of specific alignment or trimming choices, it prevents false positives in evolutionary inference.
• Practical Utility: It provides a "real-data" use case showing how to handle biological complexity (e.g., mosquito immunity genes) while maintaining rigorous analytical standards.
• Future-Proofing: The workflow is designed to be extensible, with future plans for fragment-aware routing post-GARD and structural mapping of candidate codons.
• Reproducibility: By providing machine-readable provenance and archived run bundles, it ensures that episodic selection studies can be fully audited and reproduced.

In summary, BABAPPASnake is not a new inference engine but a robustness-aware orchestration layer that integrates established tools (OrthoFinder, IQ-TREE, HyPhy, PAML) to provide a comprehensive, transparent, and reproducible framework for detecting episodic selection.