ECLIPSE: Exploring the dark proteome of ESKAPE pathogens through the sequence similarity network of the Protein Universe Atlas

The paper introduces ECLIPSE, a network-based computational framework that maps ESKAPE pathogen proteomes onto the Protein Universe Atlas to identify and prioritize evolutionarily conserved, structurally defined "dark" protein families as novel antimicrobial targets, demonstrating its efficacy through the discovery of a unique beta-barrel fold in *Pseudomonas aeruginosa*.

The Gist

Imagine the world of bacteria, specifically the notorious "ESKAPE" group (a gang of super-bugs like Pseudomonas aeruginosa and Staphylococcus aureus), as a massive, bustling city. For decades, scientists have been mapping this city, trying to understand every building, street, and citizen to find a way to stop the criminals (the bacteria) from causing harm.

However, there's a problem: a huge part of this city is in total darkness.

The Problem: The "Dark Proteome"

In this bacterial city, every protein (the tiny machines that keep the bacteria alive) is like a building. Scientists have a giant library of blueprints (known protein families) for most buildings. But for about 4% of the buildings in these super-bug cities, the lights are off. We don't know what they are, what they do, or even if they exist in our blueprints.

These are called "Dark Proteins." They are labeled "Hypothetical" (meaning "we think they exist, but we have no idea what they are"). Because we can't see them, we can't target them with new antibiotics. The bacteria are hiding their most dangerous secrets in these dark corners.

The Solution: ECLIPSE (The Flashlight)

The authors of this paper, Surabhi Lata and Dirk Heinz, built a new tool called ECLIPSE. Think of ECLIPSE not as a simple flashlight, but as a high-tech drone with a thermal camera that can map the entire city at once, even the parts where the lights are out.

Here is how it works, step-by-step:

1. The Big Map (The Protein Universe Atlas)

Imagine a giant, global map of every building ever built in the entire bacterial world. This is the "Protein Universe Atlas." It connects buildings that look similar, even if they are far apart.

• Old Way: Scientists used to look at one building and ask, "Does this look like any building I know?" If the answer was "No," they gave up.
• ECLIPSE Way: ECLIPSE looks at the neighborhood. Even if a building is unique, it might be connected to a street of other unique buildings. ECLIPSE finds these "dark neighborhoods" where no one has a blueprint.

2. Sorting the Neighborhoods

Once ECLIPSE finds these dark neighborhoods, it asks two important questions:

• Is this neighborhood unique to our specific criminal gang? (e.g., Is it only in Pseudomonas?)
• Is this neighborhood shared by the whole gang? (e.g., Is it in Pseudomonas, Klebsiella, and Staphylococcus?)

This helps them decide which dark buildings are the most suspicious. A building that exists only in the bad guys' city is a great target. A building that exists in humans too is a bad target (because killing it might hurt us).

3. The "Suspicion Score" (DPPS)

ECLIPSE gives every dark building a Suspicion Score (called the Dark Proteome Prioritisation Score). It looks at four clues:

1. How dark is it? (Does it have zero known blueprints?)
2. How common is it? (Is it in almost every strain of the bacteria?)
3. Is it exclusive? (Does it only appear in the bad bacteria, not in good ones?)
4. Is it stable? (Does the score hold up even if we change the rules slightly?)

The buildings with the highest scores are put on a "Top 10 Most Wanted" list.

The Big Discovery: Finding a New Weapon

To prove their tool works, the scientists picked the #1 suspect from their list (a protein called Component 95203).

• The Reveal: Using a super-powerful AI (AlphaFold), they built a 3D model of this dark protein. It turned out to be a Beta-Barrel—a shape that looks like a hollow tube made of beta-sheets.
• The Context: They found this protein living next door to a "traffic cop" (a regulator protein) that controls how the bacteria form biofilms (slime layers that protect them from antibiotics).
• The Implication: This suggests the dark protein is likely a door or a pump on the bacteria's outer wall, helping it survive or communicate. Because it's a unique shape found in many super-bugs but never seen before, it's a perfect candidate for a new antibiotic. If we can design a drug to jam this specific door, we might stop the bacteria without hurting humans.

Why This Matters

For years, we've been trying to find new antibiotics by looking at things we already know. But the bacteria are evolving faster than we can keep up.

ECLIPSE changes the game. Instead of looking at the known world, it shines a light into the unknown. It tells us: "Hey, there are 120,000 secret buildings in the bacteria's city that we've completely ignored. Let's go investigate the most suspicious ones first."

By illuminating these "dark" proteins, ECLIPSE gives scientists a treasure map to find the next generation of life-saving drugs before the bacteria escape us completely.

Technical Summary

1. Problem Statement

The rise of multidrug-resistant ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp.) poses a critical global health threat. Despite advances in genomic sequencing, a significant portion of bacterial proteomes remains "dark"—functionally uncharacterized and annotated only as "hypothetical proteins."

• The Gap: Conventional homology-based annotation methods (e.g., BLASTp, HMMs) fail to detect distant evolutionary relationships, leaving many proteins without functional context.
• The Consequence: These "dark" proteins may harbor novel virulence factors or metabolic pathways essential for pathogenicity and antibiotic resistance, representing untapped potential for new antimicrobial targets.
• The Challenge: Existing pipelines lack a systematic, scalable approach to identify and prioritize these uncharacterized families across the massive pangenomes of ESKAPE pathogens.

2. Methodology: The ECLIPSE Framework

The authors introduce ECLIPSE (ESKAPE Connectome Linkage and Inference for Proteome Sequence Exploration), a network-based computational framework. It moves beyond pairwise sequence comparison by embedding pathogen proteomes into the global Protein Universe Atlas (ATLAS), a massive sequence similarity network derived from the AlphaFold Database (AFDB).

The workflow consists of three main analytical layers:

A. Network Mapping and "Darkness" Classification

• Input: Protein FASTA files from pathogen strains (Case study: 635 Pseudomonas aeruginosa strains, totaling ~3.46 million sequences).
• Mapping: Sequences are mapped to the ATLAS network (UniRef50 clusters) using MMseqs2.
• Brightness Metric: Proteins are classified based on the "functional brightness" of their connected components in the network.
  • Bright: Components with high fractions of annotated homologs.
  • Dark: Connected components where the median brightness is 0% (no annotated homologs exist anywhere in the known protein universe).

B. Taxonomic Diversity Analysis

To distinguish between lineage-specific dark proteins and those broadly conserved across ESKAPE, the authors calculate Normalized Shannon Indices:

1. ESKAPE Proportion: Fraction of component members belonging to ESKAPE genera.
2. ESKAPE Genus Evenness: Diversity of ESKAPE genera within the component.
3. ESKAPE Relative Evenness: Distribution of the component across AMR vs. non-AMR taxa globally.

• Stratification: Dark components are split into two tracks:
  • Track A (Pathogen-Specific): Components exclusive to a single genus (e.g., Pseudomonas).
  • Track B (ESKAPE-Enriched): Components enriched in ESKAPE genera (≥50% membership) but not exclusive to one.

C. Dark Proteome Prioritisation Score (DPPS)

A composite scoring framework ranks candidates for experimental follow-up. It integrates four to five orthogonal sub-scores (normalized to [0,1]):

• S1 (Darkness): Functional darkness (1.0 for all retained candidates).
• S2/S2b (Pathogen Evidence):
  • Track A: S2b (Combined PA evidence) = max(Atlas species proportion, Query strain fraction). This corrects for annotation gaps where P. aeruginosa is labeled as Pseudomonas sp. in databases.
  • Track B: S2 (Atlas PA proportion).
• S3 (AMR-Clade Specificity): Inverse of taxonomic evenness (1.0 if restricted to ESKAPE genera globally).
• S4 (Strain Coverage): Empirical conservation across the input query strains (annotation-independent).
• S5 (ESKAPE Enrichment): (Track B only) Combines ESKAPE proportion and genus evenness.

Scoring Output:

• DPPS Calculation: Weighted sum of sub-scores.
• Tiering: Candidates are ranked into four tiers. Tier I (DPPS ≥ 0.75) represents the highest priority.
• Robustness Check: A Monte Carlo sensitivity analysis (500 weight perturbations) ensures that Tier I assignments are stable and not artifacts of specific weight choices.

3. Key Results

The framework was applied to the Pseudomonas aeruginosa (PA) panproteome:

• Scale of the Dark Proteome:
  • 9% of PA sequences reside in functionally dark communities.
  • 4% (120,985 proteins) reside in completely dark connected components (no annotated relatives at any evolutionary distance).
• Prioritization Outcomes:
  • After filtering for sequence length (≥300 aa) and redundancy, 7 Tier I candidates were identified (2 Pathogen-Specific, 5 ESKAPE-Enriched).
  • Robustness: Top candidates (e.g., Component 92818 and 31745) showed Tier I stability scores >0.80, confirming their rankings are robust to weight variations.
  • Correction of Annotation Gaps: The S2b sub-score successfully rescued candidates that were highly conserved across query strains but had zero annotation in the Atlas due to non-canonical naming (e.g., Pseudomonas sp. vs. P. aeruginosa).
• Case Study: Component 95203 (Top ESKAPE-Enriched Candidate)
  • Structure: AlphaFold2 modeling predicted a novel 18-stranded antiparallel $\beta$-barrel with short $\alpha$-helices.
  • Homology: Foldseek identified distant structural similarities to A. baumannii outer membrane channels and P. aeruginosa AlginateE, but with distinct functional divergence.
  • Family: Identified as belonging to the DUF1302 family, which has no experimentally characterized structural homologs in the PDB.
  • Genomic Context: Consistently co-localized with LuxR-type transcriptional regulators and DUF1329 proteins, suggesting a role in quorum sensing and biofilm formation.
  • Conservation: Present in 629 of 635 PA strains and enriched across ESKAPE pathogens.

4. Key Contributions

1. Novel Framework: Introduction of ECLIPSE, the first scalable pipeline to systematically identify and prioritize "dark" protein families in ESKAPE pathogens using global sequence similarity networks rather than pairwise homology.
2. Taxonomic Stratification: Development of a Shannon-index-based method to distinguish between pathogen-specific and pan-ESKAPE dark proteins, allowing for targeted drug discovery strategies.
3. Robust Scoring: The DPPS framework integrates multiple biological signals (darkness, conservation, taxonomic restriction) and includes a Monte Carlo stability analysis to ensure candidate reliability.
4. Annotation Gap Correction: The S2b metric effectively mitigates database annotation biases (e.g., Pseudomonas sp. vs. P. aeruginosa), ensuring biologically relevant candidates are not discarded.
5. Structural Insight: Demonstration that "dark" proteins often possess defined structural folds (e.g., novel $\beta$-barrels) and genomic contexts, validating them as tractable targets for experimental characterization.

5. Significance

• Therapeutic Potential: ECLIPSE identifies evolutionarily conserved, structurally defined, and functionally uncharacterized proteins that are likely essential for pathogenicity. These represent high-value targets for the development of broad-spectrum antimicrobials against critical WHO priority pathogens.
• Scalability: The pipeline is modular and pathogen-agnostic. It can be applied to any ESKAPE pathogen (e.g., K. pneumoniae, S. aureus) by simply adjusting input parameters, facilitating a "pan-ESKAPE" approach to target discovery.
• Open Science: The source code, datasets, and analysis notebooks are freely available, enabling the broader scientific community to explore the dark proteome of other bacterial pathogens.

In conclusion, ECLIPSE bridges the gap between genomic data and functional understanding, transforming "hypothetical proteins" into prioritized, structurally characterized candidates for the next generation of antimicrobial therapies.