DupyliCate: mining, classifying, and characterizing gene duplications

DupyliCate is a versatile, high-throughput Python tool designed to identify, classify, and characterize gene duplications across diverse species, as demonstrated through its application to various plant and non-plant genomes and specific evolutionary case studies.

The Gist

Imagine you are an archivist in a massive, chaotic library. This library contains the instruction manuals (genomes) for every living thing on Earth, from tiny bacteria to giant oak trees. Over millions of years, the librarians (evolution) have made photocopies of these manuals. Sometimes they copy just one page, sometimes a whole chapter, and sometimes they accidentally photocopy the entire book and stuff it back into the same shelf.

These photocopies are called gene duplications. They are the raw material for evolution. If you copy a page, you can keep the original safe while scribbling new ideas on the copy. This is how life invents new tricks, like a plant learning to make a new poison to fight off bugs or a flower changing its color.

But here's the problem: The library is messy. The photocopies are scattered, some are torn, some are identical, and some are so old they look nothing like the original. Finding these copies and figuring out what they do is a nightmare for scientists.

Enter DupyliCate.

Think of DupyliCate as a super-smart, high-speed robot librarian built by scientists Shakunthala Natarajan and Boas Pucker. Its job is to walk through the library, find all the photocopies, sort them into neat piles, and tell you exactly what kind of copy they are.

How DupyliCate Works (The Robot's Toolkit)

1. The "Selfie" Test (Finding Copies):
   The robot looks at a gene and asks, "Who do you look like?" It compares every gene in the genome against every other gene. If Gene A looks 90% like Gene B, they are likely copies.

   • The Analogy: Imagine you have a room full of people. The robot asks everyone to take a selfie and compare it to everyone else's. If two people look like twins, the robot tags them as a "duplicate pair."

2. Sorting the Piles (Classification):
   Once it finds the copies, it sorts them into different bins based on where they are sitting on the shelf:

   • Tandem: Two copies sitting right next to each other (like two books glued together).
   • Proximal: Copies sitting close by, but with a few other books in between.
   • Dispersed: Copies that got scattered to completely different shelves (different chromosomes).
   • The Analogy: If you find two copies of a recipe, are they in the same cookbook (Tandem), in the same drawer but different books (Proximal), or is one in the kitchen and the other in the garage (Dispersed)?

3. The "Goldilocks" Threshold (Not too strict, not too loose):
   One of DupyliCate's coolest features is that it doesn't use a "one size fits all" rule. Different libraries have different messiness levels.

   • The Analogy: If you are looking for twins in a room of identical clones, you need a very strict rule. If you are looking for cousins in a room of strangers, you need a looser rule. DupyliCate uses a special metric (called BUSCO) to measure how "clony" a specific species is and automatically adjusts its sensitivity. It's like the robot saying, "Okay, this species is very messy, so I'll lower my standards to catch all the copies."

4. Checking the "Voice" (Expression Analysis):
   Finding the copy is step one. Step two is: Is the copy actually doing anything?

   • The Analogy: Imagine you find a photocopy of a song. Is it being played on the radio? Or is it just sitting in a drawer gathering dust (a "pseudogene")? DupyliCate can check if the gene is "singing" (active) or silent. This helps scientists guess if the copy has evolved a new job or if it's just broken.

5. The "Family Tree" (Orthology):
   If you compare two different species (like a human and a mouse), DupyliCate can figure out which gene in the mouse is the "cousin" of the human gene. It builds a family tree to see who descended from whom.

Why This Matters (The Real-World Magic)

The paper shows DupyliCate doing some impressive detective work:

• The "Weed" Detective: It analyzed rice and its wild relatives. It found that some wild rice weeds had massive numbers of gene copies, suggesting they might have recently doubled their entire genome (like photocopying the whole library at once). This helps us understand why some weeds are so tough.
• The "Flower Color" Mystery: The robot looked at the FLS genes (which help make flower pigments) in the Brassicales family (which includes mustard, cabbage, and capers). It discovered that some plants have lost these genes, while others have exploded with new copies. It's like finding out why some flowers are blue and others are yellow based on their family history.
• The "Sunscreen" Regulators: It tracked MYB genes, which act like a manager telling plants how to make sunscreen (flavonols) to protect against UV rays. It found that these managers split into different teams in different types of plants, explaining how plants adapted to different environments.
• Beyond Plants: It even worked on bacteria, yeast, and worms, proving it's a universal tool, not just for the plant world.

The Bottom Line

Before DupyliCate, finding these gene copies was like trying to find a specific needle in a haystack while wearing blindfolds. You had to use different tools for different jobs, and they often missed things or got confused by messy data.

DupyliCate is the robot that puts on the glasses, sweeps the whole library, sorts the needles by type, and hands you a report saying:

• "Here are the twins."
• "Here are the cousins."
• "Here are the broken copies."
• "And here is how they are related to the original."

It's a high-speed, flexible, and smart tool that helps scientists understand the history of life by reading the photocopies evolution left behind.

Technical Summary

1. Problem Statement

Gene duplications (paralogs) are a primary driver of evolutionary novelty and diversification in biosynthetic pathways. However, identifying and analyzing these duplications presents significant challenges:

• Complexity of Detection: Distinguishing between singletons and duplicates is difficult due to functional redundancy, extensive small-scale duplications, gene loss, rearrangement, and multidomain protein structures.
• Limitations of Existing Tools:
  • Many tools focus primarily on ortholog identification or specific duplication types (e.g., only whole-genome or only segmental).
  • Most tools operate on standardized database formats (Ensembl/NCBI), lacking flexibility for diverse General Feature Format (GFF) files from various sources.
  • Existing methods often rely on fixed, non-adaptive thresholds that do not account for species-specific duplication landscapes (e.g., varying ploidy levels).
  • Few tools integrate downstream analyses like expression divergence, synteny, and Ka/Ks calculations into a single, cohesive workflow.
  • Many tools identify duplicates only in pairs, failing to capture complex arrays or clusters of genes.

2. Methodology: DupyliCate

DupyliCate is a high-throughput Python-based pipeline designed to identify, classify, and characterize gene duplications across diverse datasets.

Core Workflow Steps:

1. Input Validation & Standardization: The tool accepts GFF3 annotation files and corresponding FASTA sequences (assembly, CDS, or polypeptides). It includes helper scripts (AGAT_wrapper.py, FASTA_fix.py) to correct GFF errors and standardize FASTA headers to match GFF attributes, ensuring compatibility with various file "flavors."
2. Quality Control (QC): Uses BUSCO (Benchmarking Universal Single-Copy Orthologs) to assess genome completeness and estimate pseudo-ploidy.
3. Species-Specific Threshold Determination:
   • Auto-Mode (BUSCO-based): Calculates a self-normalized bit score cutoff based on the 95th percentile of BUSCO single-copy genes to segregate singletons from duplicates. This adapts to the specific evolutionary history of the organism.
   • Manual Mode: Allows users to set custom normalized bit score and self-similarity thresholds.
4. Self-Alignment & Segregation: Performs local self-alignment (using DIAMOND, BLAST, or MMseqs2) to identify hits. Genes are segregated into singletons or duplicates based on the determined thresholds.
5. Classification & Clustering: Duplicates are classified into:
   • Tandem: Adjacent genes.
   • Proximal: Genes within a specific genomic distance (user-defined).
   • Dispersed: Genes on different chromosomes/scaffolds.
   • Mixed: Genes belonging to multiple categories (in "Overlap" mode) or merged groups (in "Strict" mode).
   • Note: The tool outputs "Duplicate Relationships" files to maintain links between genes in complex arrays.
6. Orthology & Synteny (Reference-Based): If a reference genome is provided:
   • Performs forward alignment against the reference.
   • Uses a multi-evidence approach (local alignment + global alignment via MAFFT + gene tree inference via FastTree2) to assign orthologs.
   • Calculates an Orthologous Group Confidence Score (OGCS).
   • Performs synteny analysis for small-scale duplicates.
7. Downstream Analysis:
   • Ka/Ks Calculation: Computes selection pressures (nonsynonymous/synonymous ratios) using Nei-Gojobori or Modified Yang-Nielsen methods.
   • Expression Analysis: Correlates expression data (TPM) to infer functional fates (subfunctionalization, neofunctionalization, or pseudogenization).
   • Visualization: Generates "Duplication Landscape Plots" (Kernel Density Estimates of bit scores) and pairwise expression heatmaps.

3. Key Contributions

• Flexible Input Handling: Unlike many predecessors, DupyliCate can process heterogeneous GFF/FASTA formats without requiring pre-standardized database inputs.
• Adaptive Thresholding: The introduction of a BUSCO-based auto-thresholding method allows for species-specific identification of duplicates, addressing the issue of fixed thresholds failing in polyploid or highly divergent genomes.
• Array Identification: It moves beyond pairwise identification to detect and maintain relationships within complex gene arrays and clusters.
• Integrated Multi-Omics Workflow: It unifies duplication mining, orthology assignment, synteny analysis, Ka/Ks computation, and expression divergence analysis in a single pipeline.
• GeMoMa Integration: Includes a feature to generate structural annotations for unannotated genomes using a related species as a hint, facilitating analysis of non-model organisms.

4. Results

• Benchmarking:
  • Compared against doubletrouble and DupGen_finder using Arabidopsis thaliana (Col-0).
  • DupyliCate (manual threshold) produced results comparable to existing tools but classified 100% of input genes (0 unclassified), whereas competitors left thousands of genes unaccounted for.
  • The BUSCO-based threshold run identified ~27.5% of genes as duplicates in A. thaliana, aligning with biological expectations of post-polyploidy gene loss.
  • Runtime: DupyliCate showed comparable runtimes to doubletrouble and DupGen_finder (when accounting for necessary pre-processing steps the others require users to perform manually).
• Proof of Concept (PoC):
  • Correctly identified known duplications: SEC10 tandem duplication in A. thaliana Nd-1, a 4-gene tandem array in Gardenia jasminoides, and proximal duplications in sugarbeet nematode tolerance loci.
  • Successfully analyzed complex genomes (e.g., Vitis species) and non-plant models (E. coli, S. cerevisiae, C. elegans).
• Case Studies:
  • FLS Evolution in Brassicales: Analyzed 12 species, revealing expansion of FLS3 and FLS4 lineages and identifying pseudogenes via expression analysis.
  • MYB Evolution: Analyzed 153 plant species, confirming the presence of MYB12/MYB111 orthologs in dicots and their absence in non-land plants, with high validation rates against phylogenetic trees.
• Parameter Sensitivity: Identified --score, --self_simcut, and --seq_aligner as the most sensitive parameters influencing results.

5. Significance

DupyliCate addresses critical gaps in comparative genomics by providing a robust, flexible, and automated solution for gene duplication analysis. Its ability to handle diverse input formats and adapt thresholds to specific species makes it particularly valuable for studying non-model organisms and complex polyploid genomes. By integrating evolutionary metrics (Ka/Ks), functional inference (expression), and structural context (synteny), it offers a comprehensive toolkit for researchers investigating the mechanisms of evolutionary innovation, gene family expansion, and functional divergence. The tool is open-source and available via GitHub, Docker, and Conda, lowering the barrier to entry for high-throughput duplication analysis.