OrthoGather: a local platform for orthology-based proteome and proteomics comparisons and Gene Ontology enrichment

OrthoGather is a locally hosted web application designed to streamline comparative proteomic analysis and Gene Ontology enrichment across species by integrating orthology inference with flexible functional annotation, thereby lowering technical barriers for researchers without computational expertise.

The Gist

Imagine you are trying to understand how different families (species) react to the same storm (a stimulus, like a drug or disease). You have a list of people who got sick in Family A, and another list for Family B. You want to know: Did they get sick for the same reason? Did they use the same tools to fight back?

The problem is that Family A speaks a language you understand perfectly (they have detailed manuals for every person), but Family B has very few manuals, and many of their people are listed only as "Unknown." Comparing them directly is like trying to match a detailed resume with a blank piece of paper.

OrthoGather is a new, easy-to-use tool that solves this problem. Think of it as a universal translator and family reunion planner for proteins.

Here is how it works, broken down into simple steps:

1. The "Family Tree" Builder (Orthogroups)

In biology, proteins that share a common ancestor are like cousins. Even if they look slightly different or have different names in different species, they are related.

• The Old Way: Researchers had to use complex computer code (like a command line) to build these family trees. It was like trying to build a house using only a hammer and a blueprint written in a language you don't speak.
• The OrthoGather Way: You just type in the names of the species you want to study (like E. coli or Mycobacterium). OrthoGather automatically downloads their "family photos" (proteomes) and builds the family tree for you. It groups all the "cousin" proteins together into Orthogroups.

2. The "Smart Inference" Trick

This is the magic part.

• The Problem: If a protein in a poorly studied species (let's call him "Bob") has no manual, you don't know what he does.
• The Solution: OrthoGather looks at Bob's cousin, "Steve," who lives in a well-studied species. Steve has a detailed manual saying, "I am a firefighter." Because Bob and Steve are in the same family group (Orthogroup), OrthoGather assumes Bob is also a firefighter.
• The Result: You can now understand the function of proteins in "mystery" species by borrowing the knowledge from their well-known relatives.

3. The "Party Planner" (Comparative Analysis)

Once the families are grouped, OrthoGather helps you see who is at the party and who is missing.

• It creates visual charts (called UpSet plots) that look like a Venn diagram on steroids. They show you:
  • Which proteins are unique to one species?
  • Which proteins are shared by everyone?
  • Which proteins are shared only by specific groups?
• You can filter this list. For example, "Show me only the proteins that changed when we gave the bacteria a specific drug."

4. The "Theme Detector" (Gene Ontology Enrichment)

Now that you know who changed, you want to know what they are doing.

• OrthoGather takes your list of "changed" proteins and asks: "Are these proteins mostly firefighters, or mostly chefs?"
• It runs a statistical test to see if certain themes (like "fighting antibiotics" or "building cell walls") appear more often in your list than you would expect by random chance.
• It spits out easy-to-read bar charts and downloadable reports that you can put straight into a scientific paper.

Why is this a big deal?

Before OrthoGather, doing this kind of comparison required you to be a computer expert, know how to code, and spend hours cleaning up messy data files. It was like needing a degree in mechanics just to change a tire.

OrthoGather is like a self-driving car for this process:

• It's Local: You run it on your own computer, so your data stays private.
• It's Visual: You click buttons and see pretty graphs, not lines of code.
• It's Smart: It fills in the gaps for poorly studied organisms using the knowledge of well-studied ones.

In a nutshell: OrthoGather takes the messy, confusing task of comparing biological "families" across different species and turns it into a simple, visual story, allowing scientists to discover how life adapts to challenges without needing a PhD in computer science.

Technical Summary

Based on the provided preprint, here is a detailed technical summary of the paper describing OrthoGather.

1. Problem Statement

Comparative proteomic analysis faces significant hurdles due to the incomplete and inconsistent functional annotation of proteins, particularly in prokaryotes and non-model eukaryotes.

• Annotation Gaps: Many proteins lack reliable Gene Ontology (GO) terms, making it difficult to interpret the biological significance of differential expression studies or curated protein sets across species.
• Technical Barriers: While orthology inference tools (e.g., OrthoFinder) offer a robust solution to map functionally equivalent proteins across species, they typically require command-line execution, complex manual data preparation, and the interpretation of difficult-to-read outputs. This creates a high barrier to entry for researchers lacking advanced computational training.

2. Methodology

OrthoGather is a locally hosted web application designed to streamline the workflow from data acquisition to functional interpretation. It integrates orthology inference with GO enrichment in a user-friendly interface.

• Architecture & Stack:

  • Backend: Python 3.7.12 using the Flask 2.2.5 framework.
  • Frontend: HTML and JavaScript.
  • Core Engines:
    • OrthoFinder (v3.0.1): Executed locally via Python's subprocess module to infer orthogroups from input proteomes.
    • GOATOOLS (1.4.12): Used for Gene Ontology enrichment analysis (Fisher's exact test with Benjamini–Hochberg FDR correction).
    • Visualization: matplotlib (3.2.2) and UpSet plots for data visualization.
  • Data Sources: Retrieves reference proteomes from UniProt and GO annotations from the European Bioinformatics Institute (EBI).

• Workflow Components:

  1. Input Modes:
     • New Analysis: Users search for species/UniProt IDs; the system automatically downloads FASTA files and runs OrthoFinder.
     • Preselected Dataset: Includes a pre-loaded dataset of 47 proteomes from 17 bacterial species (e.g., cystic fibrosis pathogens) for immediate testing.
     • External Data Upload: Allows users to upload existing OrthoFinder results (compressed Orthogroups directory).
  2. Comparative Orthogroup Analysis:
     • Species Filtering: Selects specific species to filter orthogroups, generating UpSet plots to visualize orthogroup distribution and protein overlaps.
     • Protein Filtering: Allows users to input custom lists of UniProt IDs (e.g., differentially expressed proteins) to identify shared and species-specific orthogroups.
  3. Gene Ontology Enrichment:
     • Validation: Automatically matches species to available GOA files; excludes species without annotations to prevent statistical bias.
     • Coverage Assessment: Quantifies the proportion of annotated proteins per orthogroup to assess data completeness.
     • Enrichment Logic: Defines a foreground (proteins of interest) and background (control set). Crucially, it supports an "Include Complete Orthogroups" option, which expands the analysis to include all proteins in an orthogroup if at least one member is selected. This allows functional inference for unannotated proteins based on their annotated orthologs.

3. Key Contributions

• Accessibility: Transforms complex, command-line-based orthology workflows into a browser-based, locally hosted application requiring no programming skills.
• Functional Inference via Orthology: Solves the "annotation gap" problem by leveraging the "guilt-by-association" principle. If a protein in a poorly annotated species belongs to an orthogroup with a well-annotated protein in another species, the tool infers function based on the orthogroup's collective annotation.
• Integrated Workflow: Unifies proteome selection, orthogroup clustering, comparative visualization (UpSet plots), and statistical enrichment analysis into a single pipeline.
• Reproducibility: Generates publication-ready, downloadable outputs (graphs, Excel/CSV data files) and runs locally, ensuring data privacy and reproducibility without reliance on external cloud servers for processing.

4. Results & Validation

The authors validated OrthoGather using a differential proteomics dataset involving Mycobacterium smegmatis (Msm) treated with sub-lethal rifampicin.

• Case Study Application:
  • Data: 596 differentially expressed proteins from Msm were compared against a background of 3,244 detected proteins.
  • Cross-Species Comparison: The tool successfully integrated Msm with five other bacterial species (A. baumannii, B. cenocepacia, E. coli, M. abscessus, P. aeruginosa).
  • Annotation Handling: The tool automatically handled species with missing GOA files (e.g., B. cenocepacia was excluded from enrichment but retained in orthogroup structure).
  • Coverage: By including E. coli (high-quality annotation), the median GO annotation coverage across orthogroups reached 70.52%.
  • Enrichment Output: Using an FDR threshold of <0.001 and depth level 2, the tool identified significant GO terms associated with rifampicin resistance, generating clear bar plots and detailed statistical tables.

5. Significance

OrthoGather addresses a critical gap in bioinformatics by democratizing access to comparative proteomics.

• Bridging the Gap: It enables researchers studying non-model organisms or poorly annotated genomes to perform high-level functional analysis by leveraging data from well-annotated reference species.
• Broad Applicability: While demonstrated on proteomics, the tool is applicable to transcriptomics, whole-genome sequencing, and the identification of orthologs for specific genes.
• Scientific Impact: By lowering technical barriers, it facilitates hypothesis generation and guides downstream experimental validation, particularly in fields like antimicrobial resistance and host-pathogen interactions where cross-species comparison is vital.

Availability: The source code, documentation, and tutorials are available on GitHub and Zenodo.